Welcome back to AI Operator: the weekly newsletter for integrating AI into every vertical of your internet business — one SOP, system, or honest reflection every week that you can steal, install, and run by the end of the week. Last week we added 401 subscribers, taking us to 689 total AI Operators. If someone forwarded this to you, subscribe here.

I just emerged from a day at the beach completing my Q1 2026 business reflection.

And today I want to share with you the 12 biggest lessons I learned to distill them for myself & hopefully help you apply some to your own business.

But before we dive into the lessons, I’ll walk you through a quick crash course overview of my process for this reflection in case you want to do it yourself.

And if you want the full AI prompt to walk you through this reflection, like this post and comment the word RECAP and I’ll send it your way.

Now let’s dive in:

By the way if you’re new here — my name is Dickie Bush, and I run a portfolio of writing education businesses that have done over $25M in revenue over the past 5 years:

• Premium Ghostwriting Academy — 1:1 coaching to help you land your first or next ghostwriting client in 16 weeks

• AI Writing Skool — a community of writers staying on the cutting edge of how to write and build businesses with AI

• Start Writing Online — live bootcamps teaching you the fundamentals of writing online

• Velocity — a done-for-you agency to turn email into your most profitable channel

• Write With AI — the leading paid newsletter for turning AI into your personal writing assistant

• Ghostbase — the all-in-one software platform to land and fulfill ghostwriting clients

• Typeshare — everything you need to start writing online

The goal of this newsletter is to break down exactly how we’re implementing AI across every department within every one of these businesses — so you can take what we’re doing and apply it to your own.

Why do a periodic business review

The quick sales pitch on why you should take a few hours every week, month, and quarter to do this:

1. It documents the journey. I know future me is wishing I spent more time writing down what was going on at different points of my business career. So these “recaps” allow me to tap back into what I was learning at the time.

2. It leads to content ideas. The most authentic & useful content you can create is crystallizing and distilling your own lessons. This reflection is the foundation for my creative idea generation process.

3. It prevents you from going down the wrong path on autopilot. Spending too much time in the nitty-gritty details always leads me to spin my wheels without progress. By zooming out, I can reflect on the higher-level direction to make sure we’re still on course.

I’ve been doing these recaps for both my personal & business life for the past 5 years, and I can’t recommend them highly enough.

So if you’re on board to run one of these yourself, here’s how to do it.

How to do a periodic business review

The process is extremely simple and involves just two questions:

1. What happened?

2. What were my lessons and realizations?

For the first question, I start by making a “What Happened” list — everything I can remember from the period. Meetings, decisions, problems solved, experiments run, things that surprised me, things that frustrated me, etc. I prime myself with my calendar, to-do list, and Slack history, then get all of the raw material out of my head and onto the page.

From there, I go line by line and ask:

• What did this teach me?

• What would I do differently, knowing what I know now?

• What belief did this confirm or change for me?

And as a result, I emerge with a list of iterations I can make or lessons I can share with others. And that’s the list I’m about to share with you now.

(Note: all of the strategic goal-setting, initiative prioritization, and resource allocation for the business happen in a different process. If you’d like me to write about that process, leave a comment and let me know.)

With that primer complete, let’s dive into my lessons from the quarter:

12 Lessons I Learned In 1Q2026

1. Force B players to become A players or get them out

I recently stumbled upon this quote from Frank Slootman, and I can’t stop thinking about it:

“Mediocrity is the silent killer. Organizations are not getting killed by their C players. Everybody knows who they are, and performance eventually is addressed. The people who kill organizations are your B players. It’s the scourge of the enterprise because there are many and they are generally accepted. Often, they are seen as not bad enough to fire, but not good enough to keep. They are the ultimate passengers.

B players need to be pared: they either become A players, or they become C players and get flushed out. You can help by raising standards, by refusing mediocre outcomes.”

He put words to something I’ve felt for the past few years. The clear underperformers on your team are always easy to let go. But it’s the average middling performer: the one with a decent attitude and okay performance that sticks around without any major contribution and ultimately holds an organization back from growing.

This doesn’t mean every single position is going to be an A-player. That’s unrealistic. But you need to feel that your B-players are DEAD SET on leveling up and compensating for their skill deficiency with work ethic. Otherwise, they are not going to ultimately perform at the level required.

And the longer you allow mediocrity to linger, the more you are communicating and rewarding a lower standard for everyone else.

2. Run a periodic Keeper Test for every person in your organization

I recently learned about The Keeper Test from former Netflix CEO Reed Hastings, and it goes like this:

You should look at each person in your organization and ask: “If this person told me they were leaving for a competitor tomorrow, would I fight hard to keep them, or would I feel a sense of relief?”

If the answer is relief, you should likely part ways with that person.

After I heard this framework, I immediately ran every person in my organization through the test. I emerged with ~85% of our team passing the Keeper Test, and the other 15% failing.

Those 85% are people that I am happy to pour effort, coaching, and attention into, because I strongly believe I will work with them for the long-term.

The other 15%, however, lined up perfectly with people who I’ve held back from fully investing into, because subconsciously, I didn’t think my effort would compound. And this is worst-case for everyone involved. They won’t feel me fully investing in them, and they won’t fully invest back either. The relationship quietly underperforms on both sides.

So over the last 3 months, we’ve parted ways with those people. That was what was best for me, for them, and for the organization. And each of them was probably ~6 months or more overdue. Had I been running this algorithm sooner, and acting on the results, we all would be better off.

3. Your team’s performance will always fall to the minimum standard that is tolerated

The standard on your team is whatever you allow to slide without consequence.

• A closer skips their pipeline review three days in a row and nobody says anything? That is now the standard.

• A department head sends a sloppy EOD report and it gets accepted without comment? New standard.

• A setter misses their daily activity number and the week just moves on? Same thing.

Most people say a standard one time and expect it to be upheld. And it is, for the first few days. Then, it slowly drifts, 50 small reinforcing events in a row, each of which too small to notice in the moment. And 6 months later, you check in and realize performance is not even close to what you expected.

So remember: at any given time, you are either reinforcing a standard, or letting it slowly erode. There is no in between.

4. The best compensation structure is 50% base / 50% performance tied to profit

I learned this one the hard way in the first quarter.

One of our manager’s compensation plans was broken in two ways:

1. The fixed base salary was ~80% of their compensation plan

2. The performance incentive was tied to revenue, not profit

As a result of the first, they had little incentive to give discretionary effort. And this showed up when their department had a slow month. We did not see their level of intensity go up to close the gap, because their compensation was not going to be impacted either way.

As a result of the second, their incentives were to drive top line revenue, rather than healthy profitable growth. They advocated for paying everyone on their team more, running more events with big advertising budgets, etc. Their team’s efficiency metrics were lagging, but they overlooked it, instead optimizing for volume.

So I’m taking that lesson and will apply it for my new manager hires going forward.

First, I want them to have a base pay amount that protects their downside, but leaves them dissatisfied if that is all they are earning. This unlocks the discretionary effort to drive their team’s performance.

Second, I want their performance bonuses tied to the profit of their department, so they treat themselves like they’re an “internal CEO” of that business.

5. The first step to improve anything is to measure it publicly, daily

Every area of the business that sees a big “leap” forward in progress has one thing in common:

We track the most important metrics, in public, as a team, every day.

That daily KPI report becomes a mini “async meeting” hub where we discuss the numbers, celebrate wins when things go well, and remove bottlenecks when things underperform.

On the other hand, every area of the business that struggles does not have that same daily feedback loop:

• We either measure it in a sheet but don’t talk about it publicly in Slack

• Or we only share the numbers once per week, and therefore have 7 fewer iteration cycles compared to analyzing it daily

So one of my key initiatives going into the next quarter is to make sure every number that we want to improve is brought into the open every day, by the person who owns it, with a collaborative brainstorm to help them optimize it.

We will measure the success of that initiative by how active that conversation is on that thread every day.

6. You need a 5x return on a fixed cost to make it worth pursuing

For years I had the faulty assumption that “breaking even” on a fixed cost makes it worth it. This was incorrect.

To illustrate, let’s say we’re paying $7,000 per month for a TikTok growth agency.

The incorrect math is to think “alright, if this generates more than $7,000 in revenue, that’s a positive return.” Wrong.

Here’s why:

First, you at least need to break even on gross profit, not revenue. So let’s say you’re a 70% gross margin service business. You need $10,000 in revenue to generate $7,000 in gross profit. So right away, your first “breakeven” point is the investment divided by your gross profit margin. That’s now 1/70% = 1.4x

Second, with each fixed cost you add on, you need more overall revenue to cover all of them. And in our type of business, CAC only goes up as you acquire more customers. So you need to model in a higher cost to acquire, call it an additional 20%. That takes us to 1.7x revenue just to break even.

But obviously we don’t just want to break even. We want a clear and obvious return that generates free cash flow that allows us to reinvest in other things or pay ourselves. So I want to at least double the money I’m putting in to make it worth it, which takes us to 3.4x.

But that fails to take into account the cognitive switching cost of managing another thing, which has the potential to worsen your returns elsewhere. So again, you want to add in an additional margin of safety.

And finally, you have to recognize the mental stress of having higher fixed costs in the first place and the demands that places on your business. Having $300,000 in fixed costs means you wake up every day with a -$10,000 bank account. So as someone who is relatively risk-averse, I prefer to have a healthy margin of safety on any internal business investment.

Put all of those things together, and I now look at any fixed cost as needing a 5x return for us to be worth pursuing/holding onto. This is true for employees, advertising channels, masterminds, software tools, and anything else that is a recurring fixed cost in the business.

7. When you run an organic business, there are diminishing returns on repeated events

We paid for this one with our February challenge.

We ran basically the same event as we did in November, with some slight packaging upgrades & a handful of fresh deliverables. We invested the same amount of money, sent the same number of emails, and ran the same number of ads.

But revenue came in about 50% lower than the first one.

In hindsight, this makes sense, for a few reasons:

• 75% of the audience we marketed to had already seen or attended a version of this event (with a different name)

• Then another chunk of past attendees bought it just to get a few of the upgrades, but did not participate with the same level of intention as they had in November

As a result, we had fewer people joining the challenge, fewer people attending live, and fewer people taking the next step to work with us.

And this had me step back and reflect on all of the other events we’ve run once, then repeated.

I can now look back and see how many times this lesson has held. Unless our organic audience was growing RAPIDLY (as it was in 2021/2022, at the peak of Ship 30’s growth), then repeating an event led to diminishing marginal returns. Every webinar, every product launch, every bootcamp outperformed the first time we did it relative to every other time.

Now, these are still positive returns. And occasionally it’s worth repeating something even if it underperforms the prior time.

But going forward, it’s wiser to spend more time on the creative and packaging side of the new events we run, making sure that someone who has joined a similar event in the past still thinks it’s a no-brainer to sign up for the new one.

8. Periodic in-person immersions are non-negotiable when you run a remote business

I’m convinced you can make more progress with 3 days in-person than 30 days remote.

This is especially true for new initiatives. Anything that requires a lot of back and forth, strategizing, brainstorming, etc. is extremely energizing in-person, but extremely draining on Zoom.

And going out to Arizona with my business partners reminded me of this lesson that I’ve definitely learned a few times before. It makes me bullish on eventually having an in-person work/gym/cafe setup here in Miami. This would be a private space where I can invite my friends and team members who are building things to come spend time together in person.

9. There is no such thing as failure — only Mistake Dividends

I can quickly name a handful of strategic mistakes I’ve made in the last year trying to scale my business:

• Over-hiring / adding excess capacity too quickly in an effort to scale

• Spending money on unprofitable ad campaigns rather than doubling down on what was working organically

• Running a repeated event too many times rather than coming up with completely new material

• Outsourcing our content production too much rather than continuing to write new things

• Rolling out an internal or external policy change without thinking through literally every pushback scenario

• And on and on

But rather than look at these as “negative” mistakes, I’ve reframed them to be a positive. Because now, I think of failure as getting paid “Mistake Dividends.”

Let me explain:

If you think about making an investment in something, a new initiative, channel, business, team, etc. you can think about it as buying an asset. And when you buy an asset, two things can happen:

• It can work out. If so, great, you made a good strategic choice

• If it doesn’t work out, you still collect an asset in the form of Mistake

Obviously the first has a clear positive return. And if you allow it to, the second scenario, despite being “failure,” can pay you back a positive dividend in multiple ways:

• You can avoid making that mistake again in the future (saving you future time/money)

• You can tell the story of that mistake to other people (like I’m doing now) so hopefully they can avoid it too

• And in the future you can laugh at how inexperienced or naive you were at various stages of your business journey

All three of these are net-positives in the long run. And you only get them if you try something uncertain and it doesn’t work out.

So this reframe helps me take on things with fixed downside, knowing they will pay me in the future one way or another.

10. Decide what to work on with pen & paper, then execute with AI

The single biggest productivity hack I’ve found:

Spend the first 90 minutes of my day without my phone, just a pen and paper journal.

I’m still using my own brain for 100% of my:

• Content idea outlining & creative thinking

• Business strategy & resource allocation

• Personal goal setting and vision definition

And this silent morning time is the only place I can trust that I will fully “think” and use my own judgment.

I find that my brain function is more effective when I start my day removed from all technology/AI/internet.

And I feel my productivity and output is highest when I then arm myself with the most advanced technology ever created to execute what my human brain decides to execute.

11. Every big leap felt easy — “forced” is now a signal I’m on the wrong track

We started two new verticals this quarter and both worked from literally the moment we started offering them.

We clearly tapped into a vein of demand for people who want to learn live, on rotating topics, and then talk about them over time within a community. We saw massive uptake and positive response on our Bootcamp and AI Writing Skool community offers.

These felt so “smooth” that we felt stupid for not having come to the realization sooner. I’m not saying they were easy. They took a ton of work, effort, energy, and hours from our entire team. But that effort faced little resistance, like a knife cutting through soft butter.

And in looking back, this is true for basically every big leap in business we’ve had in the past. So now when I feel like something is becoming forced, or the effort is not flowing smoothly, it’s a signal that I need to change my approach. This could be either pivoting the initiative entirely, our approach to it, or the team members involved in it.

And I need to spend more time bringing awareness to that, because there is no better feeling than having your entire team in alignment on something that is clearly working.

12. When you can build anything, choosing what to build is the highest-leverage skill

This is the current mantra I am obsessed with:

“AI allows you to build anything. Which means choosing what to build is now more important than ever before.”

I feel strongly about my ability to execute on any initiative, given the tools and team at my disposal.

Which means I want to be more thoughtful about what I deploy those resources on. This doesn’t mean procrastinating waiting for the “perfect thing.” It means being deliberate with exactly what you want, how you want it done, and the reasons you want to do it. Because with all of those clearly defined, getting there has never been easier.

Aaaand that’s it. That’s every lesson I learned during the first quarter of 2026.

1. Force B players to become A players or get them out

2. Run a periodic Keeper Test for every person in your organization

3. Your team’s performance will always fall to the minimum standard that is tolerated

4. The best compensation structure for managers is 50% base / 50% performance tied to profit

5. The first step to improve anything is to measure it publicly, daily

6. You need a 5x return on a fixed cost to make it worth pursuing

7. When you run an organic business, there are diminishing returns on repeated events

8. Periodic in-person immersions are non-negotiable when you run a remote business

9. There is no such thing as failure — only Mistake Dividends

10. Decide what to work on with pen & paper, then execute with AI

11. Every big leap felt smooth — “forced” is now a signal I’m on the wrong track

12. When you can build anything, choosing what to build is the highest-leverage skill

Here’s to a big rest of the year ✊

Want The Full Business Reflection Prompt? (For Free)

If you want to run this process yourself, I built something for you — for free.

It’s a two-part guided prompt that walks you through the entire reflection:

• Part 1: Builds your “What Happened” list by pulling from your calendar, Slack, and memory across 7 categories

• Part 2: Goes line by line through that list and extracts the real lessons hiding inside it

The output is a clean numbered list of principles — in your own words, from your own experience — ready to carry into the next quarter.

All you have to do to get it:

1. Subscribe to AI Operator

2. Like this post by clicking the heart at the bottom

3. Leave a comment below or reply to this email with the word RECAP and I’ll send it over to you

Talk soon,

Dickie

P.S. — The frameworks in this newsletter are the same ones we teach live inside AI Writing Skool. If you’d rather build alongside a community and get your questions answered in real time, come join us here.

If you enjoyed this, here’s what to read next

Welcome back to AI Operator: the weekly newsletter for integrating AI into every vertical of your internet business — one SOP, system, or honest reflection every week that you can steal, install, and run by the end of the week. Last week we added 102 subscribers, taking us to 288 total AI Operators. If someone forwarded this to you, subscribe here.

If you’re using AI as a business owner but you haven’t armed your LLM with 50+ pages of business context, you are playing on hard mode.

In fact, you might as well not use AI at all—because you’re probably spending more time filling in knowledge gaps or correcting incorrect output than if you just did things manually.

If that sounds familiar, then keep reading.

What I’m about to share is the single most impactful AI upgrade I’ve made in the last 6 months. And I want to help you make the same upgrade in the next hour.

Now, I could write you a 2,000-word intro story about how I came to this realization, but today I’m going to cut straight to the point.

My time spent using AI is up 100% in the last few months, but my output is up more than 500%.

I’m using AI more, and I’m getting way better with it.

And it’s not that I got better at prompting or that the models got better. Those helped, but the real needle-move was one “context architecture” change. I answered 150 questions to build a set of context documents that give the AI a deep understanding of my business. And as a result, every single output I get requires significantly less editing, correction, or additional input from me.

So today I’m going to walk you through each of those 8 documents, what they are, why you need them, and examples of what they look like.

And if you want the full system: fill-in-the blank templates, real examples from my business, and copy-paste prompts that interview you one question at a time to build each document, like this post and leave a comment (or reply) to this email with the word CONTEXT and I’ll send it over to you.

Now let’s dive into these 8 documents:

1. The Money Model Builder — what you sell, how it’s priced, the full money model

2. The Perfect Avatar Map — who you sell to, their pains, fears, beliefs, and objections

3. The Belief Ladder — the sequential chain of beliefs your prospect must adopt before they buy

4. The Acquisition Blueprint — how strangers become customers, funnel by funnel

5. The North Star Brief — your mission, vision, values, and operating principles

6. The Org Chart Builder — who does what, reporting to whom

7. The Tech Stack Inventory — every tool in your business and what it does

8. The “What Good Looks Like” Vault — real examples of your best content, copy, and voice

Create each of these, upload them to one Project in your LLM of choice, and watch what happens to your conversations & outputs.

💰Document 1: The Money Model Builder

• What it is: A complete breakdown of what you sell, at what price points, in what order.

• What’s inside: Your offer name, core promise, transformation, deliverable stack, pricing, and upsell/downsell/continuity paths

• Why you need it: Without this, your AI is blind to what you sell. So every time you ask it to write something relevant to your offer, it lacks the context required to write anything specific. You can also make better strategic decisions around your offer & pricing by asking it to brainstorm ways to make it more efficient

🎯Document 2: The Perfect Avatar Map

• What it is: A deep profile of the person you sell to—everything from demographics to the thoughts they have when they’re considering your offer

• What’s inside: Their current circumstances, 5 tangible pains (and the emotional pain behind each one), 5 tangible desires (and how achieving them would feel), what they’ve already tried, 5 fears and limiting beliefs, the 5 necessary beliefs they need to adopt before they’ll buy, and the top 10 objections they have before purchasing — categorized by uncertainty, timing, spouse, and money.

• Why you need it: This is the doc that makes your AI sound like it actually understands your customer. Ad copy hits different when the AI knows the exact words your buyers use, the exact objections they raise, and the exact beliefs holding them back

🪜Document 3: The Belief Ladder

• What it is: The sequential chain of beliefs your prospect must adopt, in order, before they’ll buy your offer.

• What’s inside: Your destination belief (the single conclusion your prospect must reach), your unique mechanism (the proprietary method that makes the destination belief true), and the belief ladder itself — 5-7 rungs, each building on the last, from their current faulty belief to the belief that makes buying the obvious next step. Plus an argument summary that walks the whole chain in one paragraph.

• Why you need it: This is the doc that turns your AI from a random copy generator into a persuasion strategist. Without it, every ad, email, and sales script says different things. With it, every piece of content follows the same logical and emotional argument.

🧲 Document 4: The Acquisition Blueprint

• What it is: A map of every channel you use to get customers and the step-by-step funnel that turns a stranger into a buyer.

• What’s inside: All your acquisition channels (organic, paid, referrals, partnerships, SEO, community), then your core funnel mapped end to end — capture step, conversion step, and every bridge touchpoint in between (what happens, who does it, when).

• Why you need it: When you ask AI to help with your marketing & sales, it needs to know how your funnel actually works. Otherwise it gives you advice for a funnel you don’t have. With this loaded, it can spot gaps in your bridge, write emails that fit your sequence, and suggest improvements to your actual system.

⭐️ Document 5: The North Star Brief (Mission/Vision/Values)

• What it is: A one-page doc that captures what you’re building, why you’re building it, and the rules you operate by.

• What’s inside: Your vision (the long-term destination), your mission (the daily work), 3-5 core values (with what each one looks like in action), and 5-12 operating principles (the “how” behind the values — your actual decision-making frameworks).

• Why you need it: This is the doc that keeps the AI aligned with how you think. Without it, the AI defaults to generic business advice. With it, it makes recommendations that match your values and uses your language.

👥 Document 6: The Org Chart Builder

• What it is: A full roster of who’s on your team, what they do, and who they report to.

• What’s inside: Your department structure, then every person listed with their name, role, department, core responsibility, who they report to, and how long they’ve been with you. Plus a visual org chart the AI compiles for you.

• Why you need it: This is the doc that makes delegation and planning actually useful. Instead of “communicate this to your marketing team,” the AI can say “have Colin update the master content table and loop in Vitor for the YouTube cut.” It knows your people by name.

🛠️ Document 7: The Tech Stack Inventory

• What it is: A complete list of every tool your business runs on.

• What’s inside: 40+ business functions (email hosting, CRM, payment processing, scheduling, design, automations, etc.) — for each one, which tool you use and what you use it for.

• Why you need it: When the AI helps you build a workflow or write an SOP, it needs to know what tools you actually use. Without this, it suggests tools you don’t have or writes instructions for software you’ve never touched. With it, every system it builds plugs directly into your existing stack.

📝 Document 8: The “What Good Looks Like” Vault

• What it is: A collection of real examples of your best content, copy, emails, sales materials, and brand voice

• What’s inside: Best-performing short-form posts, long-form content, ad copy, landing page copy, emails, sales scripts, DM outreach, objection handling, client deliverables, testimonials — plus one example that perfectly nails your voice and one example of what your voice is NOT.

• Why you need it: This is the doc that kills the “this sounds like a robot wrote it” problem. Without examples, the AI defaults to generic. With them, it has a concrete reference for what good looks like in YOUR world — your tone, your structure, your style. The floor on every piece of content goes up immediately.

How To Set This Up (LLM Project Architecture Best Practices)

Once you have your 8 context docs built, here’s exactly how to set up your project:

• Step 1: Create one Project. Call it whatever you want — “Biz Operations,” “[Your Company] HQ,” whatever. This is home base.

• Step 2: Load all 8 context docs as files in that Project. Every chat you start inside this project now inherits all of that context automatically.

• Step 3: Create your persistent chats. These are ongoing conversations you come back to again and again:

  • One “CEO Strategy” chat — this is where you think through big decisions, bottleneck analysis, restructuring, prioritization. The AI has your full business context so it can actually be useful here.

  • One chat per department — Sales, Marketing, Product, Success, Ops, etc. Each one becomes a running thread where you make strategic decisions for that area of the business.

• Step 4: Use tactical one-off chats for everything else. Need to write an email? Draft an SOP? Brainstorm ad copy? Open a new chat inside the same project. It still has all the context — you just don’t need to keep the thread going.

I use strategic chats for things I want to come back to over time. I use tactical chats for things I use then move on from. Either way, both inherit the full context from the project.

Bonus Context Documents To Add

The 8 docs above cover the foundation. But the more context you give, the better. Here are a few additional documents & tools you can add to give the LLM even more context.

• Any spreadsheets or dashboards you reference regularly — download them and upload them to the project. Revenue trackers, ad dashboards, conversion trackers, etc. Eventually AI will be able to read these in real time, but for now, a snapshot gives them context on how things are currently going.

• Bank account or financial statements — if you want the AI to help you analyze where your money is going, give it the data. Upload a recent month’s transactions or P&L.

• Your current SOP library — if you have SOPs written for your team, drop them in. Now the AI can reference your actual processes instead of making up generic ones. We currently host everything in Notion so it can easily connect via MCP.

• Meeting notes or strategy docs — anything that captures recent decisions, priorities, or context that wouldn’t be in the 8 core docs.

• Connect Claude to Slack — this one allows you to scan Slack to get real time updates of what you’re building

• Anything else you think would be relevant — the more context you give, the better your results.

IMPORTANT! Remember these are living documents

Don’t treat your context docs like a one-time project. They should be updated in real time as your business changes.

• Hired someone new? Update the Team Structure doc.

• Changed your pricing? Update the Money Model.

• Launched a new funnel? Update the Acquisition Blueprint.

The more current your context is, the more useful every conversation becomes.

ALSO IMPORTANT! These prompts are not perfect

Once you start using the prompts to build out your context documents, remember that there may be some errors. I’ve built them to handle most scenarios and output things in a formulaic way, but don’t conclude “these are broken!” if the output is not 100% accurate on the first go.

Use these as a collaborative back & forth exercise to save your hours and hours of time in the future.

Want the Full Context Architecture Starter Kit? (For Free)

Now, if you want to set these up for yourself, I built something for you—for free.

For each of the 8 docs, you get:

• A fill-in-the blank template if you want to do it manually

• A real example from my business so you can see exactly what “done” looks like

• A copy-paste prompt that interviews you one question at a time and compiles the finished doc for you

All you have to do to get it:

1. Subscribe to AI Operator

2. Like this post by clicking the heart at the bottom

3. Leave a comment below or reply to this email with the word CONTEXT, and I’ll send it over to you

Looking forward to seeing what you build ✊

Why should you do this in the first place?

Once you have all of these documents set up, here are a couple things that will happen.

1. Content sounds like you on the first draft. No more “this sounds like a robot.” It has your real voice and examples loaded.

2. Strategy conversations actually go somewhere. You can have a real back-and-forth about your business because it understands your model, your team, and your funnels.

3. Copy references your real offer. Not “unlock your potential” — real copy about real deliverables for real people at real price points.

4. SOPs reference your tools and your people. “Send this via Slack to Tristan” — not “communicate the update to the relevant stakeholder.”

5. Bottleneck analysis is actually useful. Feed it your metrics and it spots where you’re leaking because it knows the full system.

6. Delegation briefs are ready to hand off. It knows who owns what, so the specs it writes are actually assignable.

7. And tons more that I’m missing. In short, it turns AI into a Genius Consultant with a deep understanding of your business.

So if you’re looking to take your AI output to the next level…

All you have to do to get the Context Architecture Quickstart Kit:

1. Subscribe to AI Operator

2. Like this post by clicking the heart button below

3. Leave a comment below or reply to this email with the word CONTEXT, and I’ll send it over to you

Talk soon,

Dickie

P.S. — The frameworks in this newsletter are the same ones we teach live inside AI Writing Skool. If you’d rather build alongside a community and get your questions answered in real time, come join us here.

Leave a comment

Share

If you enjoyed this, here’s what to read next

Welcome back to AI Operator: the weekly newsletter for integrating AI into every vertical of your internet business — one SOP, system, or honest reflection every week that you can steal, install, and run by the end of the week. Last week we added 38 subscribers, taking us to 186 total AI Operators. If someone forwarded this to you, subscribe here.

Toward the end of 2025, something big shifted in AI.

Models began to accelerate, seemingly overnight. And if you were active on X or LInkedIn, it felt like everyone was building something shiny, new, and impressive.

And during that time, I felt like I was falling behind.

I wasn’t using AI for anything meaningful. I had a few 1-off prompts here and there and used it like a fancy Google, but it wasn’t integrated into any of my workflows. And I knew I was leaving a lot of opportunity on the table.

Meanwhile, it seemed like my team was using it better than me.

My Co-Founder, Nicolas Cole, is a maestro at writing prompts. Our CMO & Head of Velocity, Daniel Bustamante, was dropping new Claude Skills & AI prompts to his newsletter left and right. And our team at Write With AI continued to stay on the cutting edge.

And there I was, a few months away from being doomed to the permanent underclass.

So to start the year, I went all in on becoming an AI Operator. I blocked off deep work time strictly for learning AI, signed up for the $200/mo Claude Max, and dug into my first real project: creating my personal productivity assistant.

A few days into this project, I had nothing to show for it. I spent more time debugging broken prompts and downloading context windows than actually getting work done. The magical tool that was supposed to speed up me was slowing me down.

But I quickly realized: this was not a technical problem. It was a thinking problem. I was trying to be “AI-first” without a proper understanding of what AI is good at.

And as we sit here a couple months later, I’ve been able to integrate AI into each vertical of our business.

And that realization, plus my current workflow, is what I want to walk you through today.

The One Word That Changed How I Understood AI

The shift happened when Claude released Skills.

Before that, all we had were “Prompts.” And I had an association that Prompts had to be these in-depth, master-level mini books that handle complex problems from start to finish. So that’s what I kept trying to build.

I tried to have one big prompt for my entire productivity system—the same window holding context for all of my projects, helping me plan my days, coordinating my weekly reviews, and helping me with time blocking—all things I was doing manually.

And it sucked.

Then I saw people building Skills—and the word itself changed something for me. A Skill isn’t a master prompt that handles a full, complex job. It’s a small, specific cartridge. One input, one set of instructions, one output. You slot it into a bigger system with a handful of other skills, each completing one step.

That’s when it clicked. I had been treating AI as one big assistant with a big job to do. Instead, I needed to create an army of 1000+ AI assistants, each handling one hyperspecific, small job as part of a bigger process.

And here’s the thing—I already think about business this way. I’ve given keynotes on systems thinking & bottleneck analysis. I consider myself good at this kind of work. I just never applied that same lens to AI.

And the moment I did:

• My output became reliably useful

• It became obvious where to use AI

• My prompts & skills became much easier to write

• And I could do a better job consulting my team on how to use it as well

So my key realization here:

The skill that separates those who dabble with AI and the true AI Operators comes down to one concept:

AI Systems Thinking 101

Let’s start with a high-level crash course in Systems Thinking.

Rather than spending 25+ hours working through the resources I’ll reference below, I want to hit on the most actionable parts of each & show you how they connect. These are the systems thinking “context” documents I’ve been feeding Claude, so it’s worth understanding them before we use them.

And those 3 resources are:

• Thinking in Systems by Donella Meadows

• How to Train Employees by Alex Hormozi

• The Goal by Eliyahu Goldratt

Let’s breeze through each.

Thinking in Systems by Donella Meadows

This book is a beast of a read: 500+ pages of academic systems theory. And in full transparency, I have not slogged through the entire thing.

But for integrating AI into your workflow, you really only need one concept: Nested Systems.

Here’s how I would explain it to a 5th grader:

Think about making a peanut butter and jelly sandwich. You might think that’s one thing — “make a sandwich.” But it’s actually a bunch of smaller things inside of it:

First you have to get all the stuff out. Then you have to open the bread. Then you spread the peanut butter. Then you spread the jelly. Then you put it together. Then you clean up.

And some of those things have even smaller things inside them. “Spread the peanut butter” is actually: open the jar, grab a knife, scoop some out, spread it on the bread, don’t rip the bread, etc.

And if you really wanted to get specific, you could explain how to open the jar: grab the jar with your right hand, grab the top with your left hand, twist it to the left, etc.

That’s systems thinking.

Every big thing is actually a bunch of medium things, and every medium thing is actually a bunch of small things. You just keep zooming in until you get to something so small you can just… do it.

So the first step to fully integrate AI into your workflows is to go from trying to give it big, vague systems to giving it small, specific subsystems.

Which presents the next question—how small do we need to break things down?

That brings us to the second resource.

How to Train Employees by Alex Hormozi

(Here’s the full video where I pulled ideas from)

If we can break systems down from one big thing to a bunch of smaller things, where do we stop? How small do we need to go?

It depends on the skill of the person executing the instruction.

Going back to our peanut butter and jelly example, let’s say we were working with a Master Peanut Butter & Jelly Sandwich maker. To that person, we could give vague instructions — “make me a good PB&J.” We don’t need to explain a single substep. They’ve got it.

Now, if instead we were teaching our 1st grade child to make their peanut butter and jelly for the first time, we might need to slow things way down.

• Open the jar.

• Use the back of the spoon, not the front.

• Wipe it down before you switch from peanut butter to jelly.

And if we were training a ROBOT with zero context on how the world works, how physics works, etc., we might need even more granular instructions for every inch they move along the way.

Now, this is true for sandwich making, and it’s also true in your business.

The more skilled the employee, the less granular the instructions can be.

The less skilled the employee, the more granular the instructions need to be

If you have an experienced head of marketing, you can say “go drive more leads.” Vague instructions with tons of subjective decisions, but you can trust their judgement to get it done.

If instead you have a beginner media buyer who has never worked with your business before, you will need to lay out more steps if you want the same outcome.

Now, how does this relate to AI?

Working with AI and working with employees works exactly the same way.

Most people treat AI like a senior employee. They give broad, vague instructions and expect it to fill in the gaps. But AI does not shine here. It will pretend it knows how to make decisions, but they are rarely the right ones.

Instead, you should treat AI like a smart but inexperienced new hire. When unsure, err on the side of breaking things down into more steps & giving more specific instructions. The more steps you have, and the more specific your instructions, the better your results will be.

And rather than think you have “one” employee to execute the entire process, you want to build an army of employees that execute each single step of the process, with detailed instructions to execute that step.

This is true both when you’re designing your systems (breaking things down into more subsystems) AND when you’re describing how to execute one of those subsystems (breaking one subsystem down into steps).

So keep this in mind:

If you’re getting bad outputs from AI, you’re giving instructions that are too vague for its skill level.

And this brings us to the last resource: which step should I train AI to complete for me?

The Goal by Eli Goldratt

(This is my favorite book on the Theory of Constraints. I highly recommend listening to the audiobook version. And for a full breakdown of how I think about this in a business context, you can watch this Keynote.)

Donella Meadows teaches us how to see the structure. Alex Hormozi teaches us how granular to get. And Goldratt teaches us where to focus first.

In The Goal, Goldratt tells the story of a factory manager trying to figure out why his plant can’t hit its targets — even though every individual department looks like it’s working hard.

The answer: it doesn’t matter how fast every station runs if one station is slow. The whole system backs up behind it.

That’s his Theory of Constraints, and it boils down to this:

• Every system is a sequence.

• The output of one step becomes the input of the next.

• Your system is only as fast as its slowest step — the bottleneck.

• All time spent improving anything other than the bottleneck is wasted time, because it just leads to more backup at the bottleneck.

And so to find the best place to integrate AI, you need to two things:

• You need the step to have clear inputs, instructions, and outputs

• Integrating AI into that step needs to save you more time than integrating it elsewhere

In other words, when you’re looking at where to integrate AI, don’t ask “where can I use AI?” Ask “what’s the slowest step in my system that has a clear input and output?”

That’s where AI goes first.

How These Pieces Connect

To summarize AI Systems Thinking 101:

1. Every big system is made of smaller systems

2. Break the system down to match the level of skill of whoever is executing it—the less skilled the executor, the more granular the instructions

3. Systems are linear, with the output of one system becoming the input to the next

4. Your system is only as fast as its slowest step

5. The best place to use AI is on the most time-intensive step that has a clear input & output

And once you see your workflows through this lens, integrating AI stops being some big, gnarly project and becomes a simple, repeatable process:

1. Identify your system and break it down into subsystems

2. Get granular enough that each step has a clear input, instruction, and output

3. Then, once you’ve mapped this out, find the bottleneck and streamline it with AI

Let’s walk through a live example from within my business to see it in practice.

What This Looks Like In Practice

With the theory out of the way, let’s apply it.

For our example, we’ll use the process of writing this newsletter to see where AI could best integrate.

My 10-Step Newsletter System

“Write a newsletter” is one system, but we know it will be easier to execute when we break it down into subsystems.

So we could break it down into 2 steps:

1. Write the newsletter

2. Publish the newsletter

But that still feels too big. Remember, we want to get as granular as possible to make it easier to take action.

We could further break down writing into:

1. Coming up with an idea

2. Identify the PISA Angle

3. Outlining the idea and all its sections

4. Writing the first draft

Then for publishing we could break that step down into:

1. Edit & polish for publishing

2. Write the subject line & subheadline

3. Create the visual assets that go with it

4. Upload the newsletter to Substack

5. Schedule the newsletter for publishing

6. Promote the newsletter link on Notes

And if we really wanted to, there are probably more subsystems within each of these. But these 10 are what I landed on after brainstorming with the Systems Thinking Prompt I’ll share with you in a second.

Now, to look at what we just accomplished:

• It’s easier for us to write a newsletter. Because we broke the process down into steps, it’s less intimidating. “Coming up with an idea” is way easier than writing a full newsletter. This means less procrastination and more action.

• It’s easier for us to see where AI can help. With these 10 subsystems, we have 10 opportunities to integrate AI to streamline a specific step.

With this full system outlined, where should we first integrate AI?

Finding The Bottleneck To Streamline With AI

First, let’s look at all 10 of these steps:

1. Coming up with an idea

2. Identify the PISA Angle

3. Outlining the idea and all its sections

4. Writing the first draft

5. Edit & polish for publishing

6. Write the subject line & subheadline

7. Create the visual assets that go with it

8. Upload the newsletter to Substack

9. Schedule the newsletter for publishing

10. Promote the newsletter link on Notes

And remember the mental model of where want to use AI:

• On a step with a clear input, a clear output, and clear set of instructions

• On the element that would take the most time for us to do manually

Looking at every one of these…

Now, if we filter for everything that has clear input, clear instructions, and clear output, AND we rank it from most time to least time, we get this list of AI integration opportunities:

And THIS is what it looks like to be an AI Operator.

I now have a clear list of opportunities, ranked in order from highest leverage to lowest leverage.

Is it worth it for me to set up some complex Claude Cowork/OpenClaw/Mac Mini agent to be able to click the schedule button on Substack for me?

No.

And that’s what most people are doing who waste time using AI on things that aren’t worth automating.

Instead, by following this process we’ve gotten better in two ways:

• We got more efficient at doing things manually by breaking down the big system into smaller subsystems

• We identified the highest leverage use of AI that would most streamline our current system

Trying to create these crisp, AI Operator handbook visual assets would take me ages to do manually. And they would look horrible, because that is not my skillset. So that presents the best opportunity for AI integration.

And so the first step I took to making this newsletter process more effective was using the Brand Asset Visualizer inside AI Writing Skool to build out these sleek classified document images you see at the top of each edition.

Then I used our Subject Line Generator skill from Substack Starter Kit to write me 3 high-converting and high-open-rate subject lines in under 30 seconds.

Then I built prompts to help me identify ideas and outline those ideas in under 30 minutes per week, which I outlined in last week’s AI Operator.

That’s roughly half of my newsletter workflow, now fully handled by AI.

I’m holding on to the parts where I expand the draft & how I edit & polish, because these have the least objective instructions. Trying to write a prompt that would handle these for me would not be a high-leverage use of time.

So—ready to try this for yourself?

The 2 Prompts To Integrate AI Into Your Workflow

This 2-prompt flow follows the exact process we just did manually:

1. Prompt 1 helps you define all of the elements of your system, tuned up or down based on how granular you want to get

2. Prompt 2 takes the decomposed system and helps you identify which step is best suited for AI integration

Let’s walk through them together.

Prompt 1: Turn Any Process Into A System

A few notes on how to get the most out of this prompt:

• As with any business prompt, this will perform better if your LLM is fully up to date on the context of your business. Stay tuned for next week’s AI Operator where I’ll show you how to create a full suite of business context files.

• This is not a 1-shot prompt. This requires back and forth in an interview format to achieve the clearest system steps and best results

• You can use this both for identifying where both employees AND AI can take over / where you can delegate

# Systems Decomposition Engine

You are an elite systems thinker and decomposition specialist. Your job is to take any business objective, project, or process and break it down into a clear, executable hierarchy of subsystems and steps — co-built with the user through an interview-style process.

You think using three mental models:

**Donella Meadows (Thinking in Systems)** — Every big thing is actually a bunch of medium things, and every medium thing is actually a bunch of small things. You just keep zooming in until you get to something so small you can just do it. Your first job is always to see the structure: what are the 3-7 medium things inside this big thing? You never jump from one big thing to a 47-step checklist. You map the subsystems first.

**Alex Hormozi (Operationalizing Behavior)** — How far you zoom in depends on how skilled the person executing is. A senior employee can take vague instructions and fill in the gaps. A brand new hire needs every substep spelled out. If you’re getting bad results, you’re giving instructions that are too vague. When in doubt, break it down further. The key test: could someone watching a video of the executor verify whether the step was done correctly? If not, the step isn’t specific enough.

**Eli Goldratt (The Goal)** — Every system is a sequence. The output of one step becomes the input of the next. Your system is only as fast as its slowest step. When you lay out the subsystems, order them by dependency — what must happen before what — because nothing downstream can move until the upstream step is done.

-----

## YOUR PROCESS

### Phase 1: Clarify Before You Build (Interview Style)

When the user gives you an objective, ask clarifying questions **one at a time.** Do not stack multiple questions. Wait for each answer before asking the next.

Your questions should come from this list (skip any the user has already answered):

1. **What is the system?** — What exactly are we breaking down? Get specific enough that you could explain it back in one sentence.
1. **What does done look like?** — What is the end state or output when this system is working (for a process) or complete (for a project)?
1. **What already exists?** — Is any part of this already built, or are we starting from zero? Are there tools, platforms, or constraints already in play?

Move through these quickly. The goal is to gather what you need in as few turns as possible, not to interrogate. If the user gives you a brain dump that answers multiple questions at once, don’t circle back — take it and move on.

### Phase 2: Lock the Subsystems First

Once you have clarity, produce **only the subsystems** — names and one-line purposes. No steps yet. Use 3-7 subsystems and err on the side of fewer to start.

**SYSTEM:** [Name of the system]
**PURPOSE:** [One sentence — what this system produces or achieves]

1. **[Subsystem Name]** — [One-line purpose]
1. **[Subsystem Name]** — [One-line purpose]
1. **[Subsystem Name]** — [One-line purpose]
1. **[Subsystem Name]** — [One-line purpose]
1. **[Subsystem Name]** — [One-line purpose]

Then ask the user to react: add, remove, reorder, rename. Do not move forward until the user confirms the subsystem list.

### Phase 3: Go Deeper (Only When Asked)

Once subsystems are locked, ask: **“Want me to go a level deeper on any of these?”**

When the user says yes, break that subsystem into its sequential steps. Each step should be a single action with a clear verb. Flag any steps that are handoff-ready (can be delegated to a team member or tool).

Then ask: **“Want me to go a level deeper on any of these steps, or do you want to zoom out and see the full system?”**

Continue this loop as many times as the user wants. **You never decide the depth — the user does.** Your job is to present the current level cleanly and wait for direction.

### Phase 4: Full View

At any point the user can ask you to show the complete system as one unified breakdown with all levels expanded.

-----

## RULES

### Decomposition Rules

- **3-7 subsystems first, err on fewer.** Never jump straight to steps. Always map the major subsystems before going granular. See the structure before touching the parts. Start with the fewest subsystems that capture the full system.
- **Order by dependency.** The output of one subsystem or step becomes the input of the next. Sequence them by what must happen before what.
- **Each step = one action, one verb.** A step should start with a verb and describe one discrete thing to do. If a step has “and” in it, it’s probably two steps.
- **Flag handoffs.** When the output of one step becomes the input for another person or tool, mark it. Handoffs are where systems break.

### Stopping Rule

When you go deeper, keep decomposing until each step is a single action that takes approximately **10-15 minutes or less** to execute. If a step would take longer than that, it can be broken down further — but only go there when the user asks.

### Depth Control

You never decide how deep to go. The user controls the zoom level. Your default output is always the highest useful level — the 3-7 subsystems. You go deeper only when told to. Every time you deliver a level, you ask if they want to go deeper or zoom out to see the full system.

### What You Are NOT

- You are **not a strategic advisor.** You don’t evaluate whether the objective is a good idea. You decompose it.
- You are **not a bottleneck identifier.** You don’t tell the user which step to prioritize. You give them the full map.
- You are **not a project manager.** You don’t assign timelines, owners, or deadlines unless explicitly asked.

-----

## EXAMPLE: The Granularity Dial

The same task broken down at three different levels of depth. This illustrates how the prompt works — subsystems first, then nested steps within each when you zoom in.

**Task: Make a peanut butter and jelly sandwich**

**Advanced (experienced — just needs the outcome):**

> Make a peanut butter and jelly sandwich

**Intermediate (needs the major subsystems):**

> 1. Gather supplies
> 2. Spread peanut butter
> 3. Spread jelly
> 4. Assemble
> 5. Clean up

**Beginner (needs nested steps within each subsystem — every step is observable and verifiable):**

> 1. Gather supplies
>    1.1 Open the bread bag and pull out two slices
>    1.2 Lay both slices flat on the counter
>    1.3 Get the peanut butter jar, jelly jar, and a knife
> 2. Spread peanut butter
>    2.1 Open the peanut butter jar by twisting the lid to the left
>    2.2 Pick up the knife and scoop out a marble-sized amount
>    2.3 Spread it across the entire surface of the first slice using the flat of the knife
>    2.4 Wipe the knife clean on the edge of the jar
> 3. Spread jelly
>    3.1 Open the jelly jar by twisting the lid to the left
>    3.2 Scoop out a marble-sized amount of jelly
>    3.3 Spread it across the entire surface of the second slice
> 4. Assemble
>    4.1 Pick up the first slice and place it peanut-butter-side down onto the jelly slice
> 5. Clean up
>    5.1 Close both jars and twist lids to the right
>    5.2 Put both jars back where you got them
>    5.3 Put the knife in the sink

Notice: every beginner step is a single physical action that someone could watch and verify. No judgment calls. No subjective decisions. That’s the standard for maximum depth

Prompt 2: Identify Where AI Can Best Help

With your system fully-defined, the goal of this prompt is to help you find the highest-leverage step to integrate AI.

The prompt works like this:

• You enter in your full system

• It takes a first guess at the AI-potential of each step based on the criteria we walked through above

  • Clear inputs? Yes/No

  • Clear instructions? Yes/No

  • Clear outputs? Yes/No

  • Manual time

• It then confirms using your judgment on each step, one by one

• Once complete, it will take all of the steps that have clear inputs, instructions, and outputs and stack rank them by time saved

• It will then guide you to create the prompt for that particular step

# AI Integration Evaluator

You are a world-class AI integration strategist. Your job is to take a clearly defined business system — broken into sequential steps — and help me identify the single highest-leverage place AI can own a step in that system.

You evaluate each step on 3 criteria:

1. **Clear Input** — Does this step have a defined input that exists before the step begins? Or does a significant portion of it need to come from the person’s head in the moment?
1. **Objective Instructions** — Can you write specific, non-subjective instructions for how to execute this step? Or does it require human judgment, taste, voice, or subjective decision-making?
1. **Clear Output** — Does this step produce a defined, verifiable output? Could you look at it and say “yes, this is done correctly” without subjective evaluation?

Plus one tiebreaker:

1. **Manual Time** — How long does this step take to do manually?

**Scoring biases:**

- **Err on the side of Yes for Objective Instructions.** Most people think their processes are subjective because they’ve never tried to make them objective — but they could. If objective instructions are *possible* (even if they haven’t been written yet), score it Yes.
- **Err on the side of Yes for Clear Input & Clear Output on steps further in the chain.** If the system was decomposed correctly, the output of one step becomes the input to the next. Steps later in the sequence almost always have clear inputs and outputs by definition — the decomposition already chained them together.

-----

## PHASE 1: RECEIVE THE SYSTEM

Ask me to paste my system — a numbered list of sequential steps or subsystems.

**If the input is not a decomposed system** (it’s vague, it’s a single sentence, it’s a goal without steps), tell me: “This needs to be broken down into sequential steps first. Go use the Systems Decomposition Engine to break this into subsystems and steps, then come back with the output.”

Once I paste a valid system, confirm you understand it by restating it back in one sentence, then move to Phase 2.

-----

## PHASE 2: FIRST PASS SCORING

Score every step in the system on all 3 criteria (Yes / No) plus estimated manual time. Present the full table:

|Step       |Clear Input|Objective Instructions|Clear Output|Manual Time|
|-----------|-----------|----------------------|------------|-----------|
|[Step name]|Yes/No     |Yes/No                |Yes/No      |[estimate] |

Then say: “Here’s my first pass. Let’s go through each one so you can confirm or correct.”

-----

## PHASE 3: CONFIRM EACH STEP (One at a Time)

Go through each step sequentially. For each one:

1. State the step name
1. **Refer back to your original table from Phase 2** to ensure your scores match exactly — never present mismatched information from your original estimate
1. Show your scores and a one-line explanation for each criteria score
1. Ask: **“Do you agree with these scores, or would you adjust any of them?”**

If the user agrees, move to the next step. If they adjust, update the score and confirm before moving on.

Do NOT batch these. One step at a time.

-----

## PHASE 4: IDENTIFY THE HIGHEST LEVERAGE STEP

Once all scores are confirmed, present the final table **sorted by manual time from most to least:**

|Step                |Clear Input|Objective Instructions|Clear Output|Manual Time|
|--------------------|-----------|----------------------|------------|-----------|
|[Longest step first]|Yes/No     |Yes/No                |Yes/No      |[time]     |
|…                   |…          |…                     |…           |…          |

Then identify the single step that:

- Scored **Yes** on all three criteria (Clear Input, Objective Instructions, Clear Output)
- Takes the **most manual time**

Present it as:

**Highest-leverage AI integration:**
**[Step name]** — [manual time] per cycle

This is the step where AI can fully own the work, the instructions can be made objective, and it saves the most time.

If multiple steps tie, present the top 2-3 and ask the user which one to build first.

If no steps scored Yes on all three, say so honestly and identify the step closest to qualifying — then explain which criteria would need to flip to Yes for AI to own it.

-----

## PHASE 5: BUILD THE PROMPT FOR THIS STEP

For the top step, help me define exactly how AI will execute it:

- **Input:** What exactly gets handed to AI to start this step?
- **Instructions:** What specific criteria, rules, frameworks, or constraints does AI need to execute well?
- **Output:** What exactly does AI produce when it’s done? What does “good” look like?

Walk through each of these one at a time. Ask me what the input is. Then ask what rules or criteria I’d include. Then ask what the finished output looks like. Draft the spec after each answer so I can react, then compile the full prompt at the end.

-----

## RULES

- If the system isn’t decomposed, redirect to the Systems Decomposition Engine. Do not try to evaluate a vague objective.
- Go one step at a time in Phase 3. Never batch confirmations.
- If my answer is vague (“the instructions are pretty clear”), push once: “Can you give me an example of what those instructions would say?”
- Do not assume — if you’re unsure whether something is Yes or No, ask.
- The goal is an honest evaluation, not an optimistic one. If something requires subjective judgment, score it No on Objective Instructions even if AI could technically attempt it.
- Only build the prompt for the single highest-leverage step. Do not try to build prompts for every step.

You Now Have Everything You Need To Integrate AI

Aaaand that’s it!

To recap everything we just walked through:

• You now understand AI Systems Thinking 101

  • Every big system is made of smaller systems

  • Break the system down to match the level of skill of whoever is executing it—the less skilled the executor, the more granular the instructions

  • Systems are linear, with the output of one system becoming the input to the next

  • Your system is only as fast as its slowest step

  • The best place to use AI is on the most time-intensive step that has a clear input & output

• Once you understand this, integrating AI stops being some big, gnarly project and becomes a simple, repeatable process:

  1. Identify your system and break it down into subsystems

  2. Get granular enough that each step has a clear input, instruction, and output

  3. Then, once you’ve mapped this out, find the bottleneck and streamline it with AI

• We walked through an example of breaking down my newsletter publishing process into a handful of subsystems

• Then, rather than waste time building AI systems for low-leverage activities, we evaluated each step based on how suitable it was for AI integration, then sorted it by the most time saved to least time saved

• Then, and only then, we moved on to creating AI prompts & skills that streamlined our workflow

This is what it looks like to be a top 1% AI Operator.

So, your next steps:

• Find a system in your business that currently takes a lot of your time

• Run it through the Systems Decomposition prompt to break it down into subsystems and steps

• Run that decomposed system through the AI Integration Evaluator prompt

• Choose the highest leverage place to integrate AI, then go build that prompt that executes it

• And then repeat the cycle for the next best opportunity, and on and on your AI flywheel spins

And as always, if you have any questions: hit reply to this email or leave a comment. I read and respond to every one of them.

See you next week,

Dickie

P.S. — The frameworks in this newsletter are the same ones we teach live inside AI Writing Skool. If you’d rather build alongside a community and get your questions answered in real time, come join us here.

Leave a comment

Share

If you enjoyed this, here’s what to read next

AI Operator

How To Generate Endless Business Content Ideas In Just 30 Minutes Per Week

Welcome back to AI Operator: the weekly newsletter for integrating AI into every vertical of your internet business — one SOP, system, or honest reflection every week that you can steal, install, and run by the end of the week. If someone forwarded this to you…

Read more

2 months ago · 19 likes · 1 comment · Dickie Bush

Welcome back to AI Operator: the weekly newsletter for integrating AI into every vertical of your internet business — one SOP, system, or honest reflection every week that you can steal, install, and run by the end of the week. If someone forwarded this to you, subscribe here.

In 2025, I was in a content rut.

For the first 4 years of operating my business, writing was my number one priority. The single highest leverage activity I could do every day was write down my ideas and share them on the internet. The more I wrote, the faster our email lists grew, which led to more customers, which led to more revenue.

It was simple.

But in 2025 the highest leverage activity changed. Rather than spend 4-5 hours per day writing, the business needed me in operator mode — making strategic decisions, leading the marketing & sales teams, and keeping the machine humming across every vertical.

Content was still important, but I had a team who could keep things going if I didn’t have as much time or mental bandwidth to write in a given week.

Fast forward to 2026 and the priorities are shifting again.

My number one priority is still operating the business. But we’ve made talent upgrades—true operators—at key department head & vertical positions that I can reliably depend on.

And now I’m plowing all that extra time & mental bandwidth into content.

But here’s the problem: I’d just spent a full year thinking like an operator, not a newsletter writer. And in my first few writing sessions, I did nothing but stare at a blank page. It felt like I was starting my writing journey over from scratch.

Now, it wasn’t that I had nothing to say. It was that I was using an outdated approach. When I was writing full time, I spent hours every day in idea mode — capturing observations, noticing patterns, writing things down as they happened. The ideas were always there because I was always looking for them.

But now my week was full of operating, not creating. By the time I sat down to write, I’d try to think of something from scratch, rather than talk about things I was already doing.

That’s when it hit me: I didn’t have an idea problem. I had a systems problem.

I had plenty to write about: I literally spent all day solving problems. But I didn’t have a system that helped me 1) recognize those problems could be pieces of content and 2) easily outline the idea for me based on work I had already done.

So I built a new system. And that’s what I want to share today.

By the end of this, you’ll have a complete system for turning the work you’re already doing into endless newsletter ideas — plus three AI prompts to run the whole thing on autopilot.

And it starts with a simple mindset shift that turns everything you’re already doing into content.

Selling Your Sawdust

I first heard this concept from Jack Butcher.

The idea comes from the woodworker who spends all day making things in his shop. At the end of the day, the floor is covered in sawdust—the natural byproduct of all the cutting, shaping, and sanding. To him, this sawdust is just scraps to sweep up and toss out.

But the sawdust, the byproduct of his work, is actually useful.

• Farmers buy it for animal bedding

• Campsites will buy it for fire starters

• Manufacturers use it in composite wood

To the right person, it’s valuable—yet the woodworker was throwing it away for free.

And the same is true with content.

Sharing the byproducts of your work—your sawdust—is the best type of content you can share as a business owner.

And this approach is better for 4 main reasons:

• It’s authentic. Since you’re only sharing things you actually use, it’s impossible for someone else to replicate it.

• It’s easy to write. Since you’ve already done the work, you simply need to document what you did.

• It solves real problems. Since it solved a problem for you and your business, it’s already proven to be valuable.

• It demonstrates expertise through action, not claims. And rather than try to convince people you know what you’re talking about, you can show them by sharing things they will find useful

And here’s the good news: your sawdust is already there. You just need a system to collect it.

What’s the best way to do that?

Easy—just follow these 3 steps.

3 Steps To Generate Endless Business Content Ideas

To come up with ideas, you’re going to follow 3 simple steps:

1. Make A “What Happened” Weekly Recap List

2. Turn Each Item Into A Proven Content Idea

3. Expand Each Content Idea Using The PISA Framework

And after I explain each of these, I’ll walk you through a chain of AI Prompts you can use to bake this process into your weekly workflow.

Let’s dive in:

Step 1 — Make A “What Happened” Weekly Recap List

The first step to generating ideas is getting all of the “raw material” of potential ideas out there.

And the best way to do this is to make a “What Happened List”

Here’s how it works.

1. Block 15-30 minutes at the end of the week. Sometimes I do this on Friday afternoons, other times Sunday mornings.

2. Prime your brain to remember what happened. I like to scan my Calendar, To-Do List, Slack, Email, Journal, and AI Chat history.

3. Write out as many items as you can remember without a filter. The most important part here is to brain dump without judgement. We will filter things down in the next step.

To further prime your brain, here’s a list of categories where you can generate things that happened.

• Meetings — conversations with your team, clients, mentors, or peers where something was discussed or decided

• Decisions — anything you approved, rejected, or changed course on, big or small

• Problems solved — things that were broken, stuck, or underperforming that you diagnosed and fixed

• Systems built — new processes, workflows, tools, automations, or playbooks you created or documented

• Lessons learned — realizations, mistakes, or mindset shifts that changed how you think or operate

• Experiments run — new approaches, tools, or ideas you tested for the first time

• Feelings had — energy shifts, frustrations, wins, or moments that stuck with you throughout the week

Nothing is off limits. Just get everything out of your head and onto the page.

To further drive this home, here’s an example list of my “What Happened” list from last week:

1. Wrote team-wide memo on March priorities

2. Built AI lead scoring assistant in Claude

3. Ran multiple sales team meetings to get them out a rut

4. Optimized our cold webinar funnel by attacking the show rate

5. Used AI to build a nurture content system for our sales team

6. Had a 1:1 call with one of my mentors to talk through the business

7. Experimented with going “no phone” for the first hour of the day

8. Let go of 3 team members who were undeperforming our standard

9. Watched the most recent Hormozi episode on how he uses AI

10. Started a brand new newsletter to talk about business ideas

As you can see, these all fall into different categories.

Some of them will make for great content ideas—and that takes us to the next step.

Step 2 — Turn Each Item Into A Proven Content Idea

With your list of things that happened, it’s time to turn them into content.

There are 3 angles that work well for business content, based on what you accomplished:

The SOP — A Step-By-Step Guide

Use this when you’ve achieved an outcome or solved a problem.

The goal is to walk someone through exactly what you did, in order. The reader should walk away with a specific solution to a specific problem.

SOP-worthy content usually happens when you built something or figured something out.

For example, right now I am walking you through a step-by-step guide to generate business ideas. This came from me creating a system for myself to generate those ideas. And now I’m sharing it with you. Meta!

Examples of SOP-worth content include:

• Templates & Fill-in-the-Blank Tools

• Checklists & Cheatsheets

• Recipes & Formulas

• Scripts & Prompts

• Guides & Playbooks

Anything you use to achieve an outcome for yourself or your clients is fair game for an SOP.

The Breakdown — An Overview of a System

These are similar to SOP-style content, but with a small difference.

SOPs give step by step instructions that everyone should apply in the same way. Here’s an outcome, here’s how to achieve it.

Breakdowns require the reader to adapt your process to their situation. They watch and identify the elements they can integrate, and discard the rest.

For example, let’s say I launched a new webinar funnel last week.

I could walk someone through a “Step by Step” process to launch their own webinar funnel. That would be an SOP-style of content.

OR, I could talk through our webinar funnel and all of the elements within it. Rather than show someone exactly how to set it up, I just show them ours and let them infer how to do it themselves.

Here are some examples of breakdown-worthy systems:

• Your AI tool stack

• Your landing page

• Your full tech stack

• Your hiring process

• Your content system

• Your morning routine

• Your launch sequence

• Your mission & values

• Your content swipe files

• Your onboarding process

• Your tracking spreadsheets

• Your decision-making frameworks

Any system you use in your business or create for your clients is fair-game for a Breakdown.

The Reflection — Sharing A Lesson You Learned

Lastly, sometimes you don’t have a full step-by-step guide to explain or a system to break down.

Instead, you just have a realization or lesson that 1) you want to distill for yourself and 2) would have saved you time/energy/effort/money had you known it sooner.

A single lesson usually follows this pattern: something happened, it changed how you think, and you can state the new principle clearly enough that someone could repeat it to a friend. The story is the vehicle. The principle is the point.

A list of lessons works well when you’ve had a week, month, or quarter full of distinct realizations that don’t connect into one unified system — but each one is worth sharing on its own.

Now whether you’re sharing an SOP, a breakdown, or a Reflection, you can expand it into a newsletter using the PISA framework.

Step 3 — Expand Each Content Idea Using The PISA Framework

Once you know what type of content you’re making, it’s time to turn your idea into a full piece.

The framework I use for every single piece of content — regardless of whether it’s an SOP, Breakdown, or Reflection — is PISA:

• Problem — Start with a tangible moment that you or your reader might experience. Describe the surface problem and the deeper frustration that comes with it. If your audience has felt this pain, they’ll keep reading.

• Insight — Share the “aha” moment. What realization made this the right way to solve it? What concept or principle guided you? This is what makes your content unique.

• Solution — Walk through exactly what to do, what you built, or what you learned.

  • SOP: Step-by-step instructions the reader can follow in the same order to get the same result.

  • Breakdown: A tour of the system — name each part, explain what it does, let the reader decide what to take.

  • Reflection: The lesson stated plainly, illustrated with your specific story. One principle the reader could repeat to a friend that prevents you from making that mistake in the future.

• Asset — Give the reader a deliverable that helps them implement the solution. A checklist. A prompt. A template. A framework. This is what leads to behavior change for the reader, and that is the goal of educational content.

Here’s how PISA would look across a handful of different content types:

SOP Example — How to generate endless business content ideas as a business owner

• Problem — Every week you sit down to write and stare at a blank screen. You’ve been solving real problems all week — meetings, decisions, systems, experiments — but none of it feels like content.

• Insight — You don’t have an idea problem. You have a systems problem. The content is already there in your daily work. You just don’t have a system to capture it.

• Solution — Step 1: make a “What Happened” list at the end of every week. Step 2: map each item to a content angle — SOP, Breakdown, or Reflection. Step 3: expand your best idea using the PISA framework.

• Asset — Three AI prompts that run each step automatically. Paste your week in, get a fully outlined newsletter out.

Breakdown Example — Our Sales Manager Daily Operating System

• Problem — Sales managers need a predictable operating playbook rather than operating off of vibes. No consistency, no visibility. I never knew what was actually happening until something broke.

• Insight — A sales manager doesn’t need creativity. They need a checklist. The job is the same every day. So I built the operating system and removed the guesswork entirely.

• Solution — Here are the 7 activities in our daily SOP: AM Check-In, Admin QC, EOD Feedback, BOD Revenue Hunting, Daily Team Meeting, 2x Call Reviews, Pipeline Power Hour. Each one has a time, an owner, and a clear output.

• Asset — The full Sales Manager Playbook — every SOP, every cadence, the complete daily operating system you can model for your own sales manager.

Reflection Example — I kept three underperformers for six months. Here’s what it cost me.

• Problem — I had doubts about all three for weeks. But I kept finding reasons to wait — give it one more week, don’t disrupt the team, wait for more data. Meanwhile I was absorbing their slack and the whole sales function was underperforming.

• Insight — The doubt IS the signal. I used to wait for proof. Now I know: the moment you’re unsure about someone, that’s already the answer. Waiting only costs you.

• Solution — Run the Keeper Test: if this person came to me as a new hire today, would I fight to keep them? If it’s not a clear yes, it’s a no. Optimize for people who inspire you, not just people who are competent.

• Asset — Three questions to run through for every person on your team: Would I hire them again today? Do they inspire or drain me? Would I feel relieved or gutted if they left tomorrow?

And just like that, you have a full newsletter outline that solves a real problem, using a solution you’ve already created.

Easy!

System Recap — How To Generate Endless Business Content Ideas

Before we dig into the prompts (the most important part!), here’s a recap of what we just covered:

1. You have plenty of content ideas, you just need a system for finding them

2. Sawdust is the best type of business content — sharing elements of your workflow that actually work

3. There are 3 types of high-performing business content — the SOP, the Breakdown, and the Reflection

4. You can expand each of those into a valuable newsletter using the PISA framework - Problem, Insight, Solution, Asset

Now, here’s how to use AI to help complete this process for you.

AI Prompts For Endless Business Content Ideas

Prompt #1: The Business Problem Interviewer

The first prompt I have for you here will help you generate your “What Happened list” from the week.

A couple quick nuances on this:

• This prompt works well even if you do not do most of your daily work “in” the LLM that you’re running the prompt. So it’s fine to run straight out of the box without anything connected

• It works EXTREMELY well if connect your tools (Calendar, to-do list/document app of choice, communication tool, etc.)

  • For example, I have Claude connected to Notion, Google Calendar, and Slack. Plus, I do a lot of my planning and prep work within Claude as well, so it has context on the things I’m building

• And it’s EVEN MORE overpowered if you give it a proper context document on your business and newsletter

  • If you’d like me to do an issue breaking down how to create context documents for your newsletter & business, please leave a comment or reply and let me know :)

You are a world-class strategic business interviewer helping me identify content ideas from my daily & weekly operations. 

Your first job is to help me create a "What Happened" list from my past week so I can turn my daily work into content ideas.

---

Step 0: Gather items from connected tools and our chat history

Before I list everything that happened, I want you to pull a list of ideas from your available connections to get us started

- If you can access my calendar, pull the meetings, events, and work blocks I had last week. If I do not have this connected, skip it.
- If you can access my to-do list, pull the tasks I completed. If I do not have this connected, skip it.
- If you can access my Slack, pull summaries of my most active channels to find the 3-5 activities that took the most of my time. If I do not have this connected, skip it.
- If we have completed work together in recent chat windows, include those activities from just the last 7 days. If you do not have chat history, skip it.

Step 1: Prompt Me To Brain Dump Anything Top of Mind

First, ask me to share everything I can remember from the week in one unfiltered stream. No structure needed. Whatever comes to mind: wins, frustrations, decisions, conversations, things I built, things that surprised me, things that felt hard. Prompt me with a trigger list of where to look and what to share. 

After I share, organize what I gave you into a rough list before moving to Part 2.

---

PART 2 — CATEGORY DEEP DIVE

Walk me through each of the following 7 categories one at a time.

For each category:

1. Show me the list of places to look that might spark more ideas
2. Ask me the category question
3. After I answer, review every item I gave you — if anything is vague or surface-level, ask one targeted follow-up question before moving to the next category

FOLLOW-UP RULE — this is important:
After I answer each category, check every item against this test: could someone read this and immediately understand what happened, why it mattered, and what the outcome or observation was? If not, push for one more level of detail before moving on.

Vague answer: "Had a sales meeting"
Strong answer: "Ran a sales team meeting to diagnose why reps were hitting below 60% for three weeks in a row — identified that the problem was clarity, not motivation, and built a daily non-negotiable checklist together"

Example follow-up: "You mentioned you worked on the funnel — what specifically did you change, what problem were you trying to solve, and what happened as a result?"

---

THE 7 CATEGORIES

1. Meetings & Conversations
Places to look: calendar invites, Zoom or call history, Slack DMs, voice memos, any recurring 1:1s or team standups
Question: "What meetings or conversations stood out this week — with your team, clients, mentors, or peers? What was discussed, decided, or realized?"
2. Decisions Made
Places to look: Slack threads where you gave direction, emails you sent, project management tools, anything you approved, rejected, or changed course on
Question: "What decisions did you make this week — big or small? Hiring, strategy, budget, process, product, team structure?"
3. Problems Solved
Places to look: support tickets, team escalations, fires that got put out, things that were broken and are now fixed, Slack channels where urgent issues were handled
Question: "What problems did you solve or make meaningful progress on this week? What was broken, what did you diagnose, and what did you do about it?"
4. Systems or Processes Built
Places to look: Notion docs or SOPs created, automations set up, new tools installed, templates made, workflows documented
Question: "Did you build, document, or install anything new this week — a process, a workflow, a system, a tool, or a playbook?"
5. Experiments Run
Places to look: A/B tests, new things you tried for the first time, changes you made to test a hypothesis, anything you're doing differently than last week
Question: "Did you try anything new this week — a new approach, a new tool, a new way of doing something? What did you test and what did you find?"
6. Lessons Learned or Realizations Had
Places to look: journal entries, things you told a colleague or friend, moments where something clicked, mistakes that taught you something, things you would do differently
Question: "What did you learn or realize this week? What shifted in how you think? What would you tell yourself two weeks ago?"
7. Anything Else Notable
Places to look: personal wins, energy shifts, things you read or heard that changed how you think, events that happened, conversations that stuck with you
Question: "Is there anything else worth capturing — a feeling, a small win, something you noticed, or something someone said that stayed with you?"

---

FINAL OUTPUT

Once we've gone through all 7 categories, compile the complete "What Happened" list.

Write each item as one specific sentence using this format:
[What happened] + [what the outcome, observation, or result was]

Format as one long numbered list grouped without grouping by category. Then ask if I would prefer it grouped by category or not. Remove anything that is too vague to turn into content even after the follow-up questions.

Prompt #2: Turn Your What Happened List Into Content Ideas

Once you have your “What Happened” list from Prompt #1, you will paste it into this second prompt.

This prompt maps every item to the right content angle and gives you a working title you could actually publish. It also surfaces your top 3 picks so you know exactly which one to take into Prompt #3.

You are a world-class business content strategist. I'm going to give you my "What Happened" list from this week — a specific, itemized list of everything I worked on.

Your job is to map each item to the best content angle and give me a working title I could actually publish.

THE 3 CONTENT ANGLES

**The SOP — A Step-By-Step Guide**
Use this when you achieved a specific outcome or solved a specific problem with a repeatable process.
Signal: "I figured out how to...", "I built a process for...", "I fixed...", "I solved..."
The reader follows the exact steps in order and gets the same result.
Example: "Ran a sales team meeting to diagnose why reps were hitting below 60% for three weeks — identified that the problem was clarity, not motivation, and built a daily checklist together"
→ "5 Steps To Reset A Sales Team That's Lost Its Edge"

**The Breakdown — An Overview of a System**
Use this when you built, adopted, or overhauled something you can describe at a high level — where the reader needs to see the whole picture and adapt what's relevant to their situation.
Signal: "I use...", "Our system for X involves...", "Here's everything that goes into..."
The reader sees all the parts, how they connect, and pulls what works for them.
Example: "Set up a 7-activity daily operating system for our new Sales Manager"
→ "The Sales Manager Daily Operating System We Built So Our Team Could Run Without Us"

**The Reflection — A Lesson or Realization**
Use this when something happened and the value is the perspective shift, not the steps. A win, a failure, a mindset shift, an observation.
Signal: "I realized...", "I used to think X but now I think Y", "This week taught me..."
The reader gets a principle they could repeat to a friend — one that would have saved them time, money, or energy.
Example: "Kept three underperformers for six months, then let them all go in one week"
→ "I Kept Three Underperformers For Six Months. Here's What It Cost Me."

For each item on my list, output:

**Item:** [one sentence description]
**Angle:** [SOP / Breakdown / Reflection]
**Working title:** [specific, punchy, publishable]
**Secondary option (if applicable):** [if a second angle is equally strong, give the alternative title]

If an item doesn't have enough substance to become a useful piece of content, flag it as "Not enough yet" and move on.

Once all items are mapped, give me your **Top 3 picks** — the three items with the highest content potential — with one sentence on why each one stands out.

Then ask: "Which one do you want to expand into a full newsletter? We'll take it into Prompt #3."

Prompt #3: The PISA Newsletter Builder

Use this once you’ve chosen a content idea from Prompt #2. Paste your selected item and angle, then let the interview run. The output is a fully outlined newsletter ready to write.

You are a world-class business content interviewer and newsletter strategist.

I'm going to give you a content idea and the angle I want to write it from. Your job is to interview me section by section using the PISA framework, then produce a complete newsletter outline I can write from directly.

My content idea is: [PASTE ITEM AND ANGLE FROM PROMPT #2 HERE]
Example: "SOP: How I rebuilt our sales team's daily checklist after three weeks of missed targets"

---

THE PISA FRAMEWORK

You will interview me on four sections in order. Do not move to the next section until I have given you enough to work with in the current one.

After each of my answers, do two things:
1. Reflect back what you heard in one sentence so I can confirm you understood it correctly
2. If my answer is surface-level or missing a key detail, ask one targeted follow-up before moving on

Use this test before moving on from each section: could a reader who has never met me understand exactly what happened, why it mattered, and what they should take from it? If not, push for one more level of detail.

Surface-level answer: "I realized I needed to fix the team"
Strong answer: "Three weeks of sub-60% attainment made me realize the problem wasn't effort — it was that reps didn't know what a good day looked like, so they defaulted to doing less"

---

THE 4 SECTIONS

**P — Problem**
What you are looking for: a concrete situation, what was broken or not working, and the emotional experience of having this problem. The reader should feel this in their gut.

Ask me: "Walk me through the specific moment or situation that made this a problem worth solving. What was happening, what wasn't working, and what did it feel like to be in that situation?"

Push for: a specific timeframe, a number or data point if relevant, and the emotional cost — not just the operational one.

Note on Pain vs. Outcome: The Problem section doesn't always have to open with suffering. Sometimes the problem is a gap between where you were and where you wanted to be. Lead with the gap, the tension, or the cost — whatever is most honest and most relatable.

**I — Insight**
What you are looking for: the "aha" moment — the realization or mindset shift that made the solution obvious. This is what makes the content unique. Tactics can be copied. Insights cannot.

Ask me: "What was the moment you realized how to solve this? What shifted in how you were thinking about the problem?"

Push for: the contrast between how you thought about it before vs. after. "I used to think X, but then I realized Y" is the structure you're looking for.

**S — Solution**
What you are looking for depends on the content angle passed in above:

- If this is an SOP:
Solution = the exact steps taken, in order, with enough specificity that a reader could replicate it. No vague summaries. Real details.
Ask me: "Walk me through exactly what you did, step by step. What did you do first, what came next, and what was the result at each stage?"
Push for: numbered steps, specific actions, tools or frameworks used, and at least one concrete detail per step. If I say "I sent an email," ask what the email said. If I say "I ran a meeting," ask what the agenda was.

- If this is a Breakdown:
Solution = an overview of the components, layers, or elements that make up the system. No strict order needed — the reader is meant to see the whole picture and pull what's relevant.
Ask me: "Walk me through all the pieces — what are the main components, how do they fit together, and what does each one do?"
Push for: names of each component, what role it plays, and one specific detail per piece that makes it concrete and not generic.

- If this is a Reflection:
Solution = the lesson distilled into a clear, transferable principle or set of principles. Not steps — just the realization and what it means for how you operate now.
Ask me: "What's the core lesson — and if you had to give someone one thing to walk away with, what would it be? How does this change the way you think or operate going forward?"
Push for: a before/after contrast, a specific example that illustrates the lesson, and a principle stated plainly enough that someone could repeat it to a friend.

**A — Asset**
What you are looking for: a deliverable the reader can steal and use immediately. Not a vague offer. A specific, standalone thing tied directly to this piece of content.

Ask me: "What is the one thing you could hand someone — a template, a prompt, a checklist, a script, a framework — that would let them implement this without starting from scratch?"

Push for: something concrete and specific to this content. The asset should solve the exact problem described in P, using the insight from I, following the approach in S. A generic "template" is not enough — name what it contains and what someone does with it the moment they open it.

Here are examples of strong asset types to draw from:
- Templates & Fill-in-the-Blank Tools
- Checklists & Cheat Sheets
- Recipes & Formulas
- Scripts & Prompts
- Guides & Playbooks
- Frameworks & Systems
- Swipe Files & Examples
- Logs & Trackers
- Numbers, Data & Research
- Curated Collections & Lists

---

FINAL OUTPUT

Once all four sections are complete, produce:

**5 Subject Line Options**

Hooks to draw from (use at least one per subject line):
- A ton of value for minimal time
- A ton of value for minimal cost
- How to solve your problem without much effort
- How to unlock a desirable outcome, instantly

Elements to incorporate:
- Unexpected comparisons or analogies
- Counterintuitive or controversial claims
- Promises of quick or dramatic results
- Hints at insider knowledge or secrets
- Numbers (e.g. "3 questions", "1 week", "7 years")
- Time-based anchors (e.g. "Last Tuesday", "In 10 minutes")

Techniques:
- Create a curiosity gap in each subject line without being cryptic
- Simple, conversational language (middle school reading level)
- Strategic capitalization and punctuation for emphasis
- Add a whisper at the end where relevant:
  - Trust whisper: "...from someone who's built a $25M internet business"
  - Obstacle whisper: "...without hiring anyone new"
  - Benefit whisper: "...and how to steal it"
  - Outcome whisper: "...and start running leaner by Friday"

For each of the 5 subject lines:
- The subject line itself
- Which hook it leads with
- Which whisper it uses (if any)
- One sentence on why it works

**Full PISA Outline**
Write each section as a header with 4-6 bullet points pulling directly from my answers. Write in my voice — direct, specific, operator-first, no filler. Each bullet should be a complete thought someone could expand into a full paragraph.

**Content Angle Note**
One sentence on the angle and how it should shape the writing:
- SOP: "Write each Solution step as an instruction the reader follows, not just something you did."
- Breakdown: "Write the Solution section as a tour — show them the system, name each part, and let them decide what to take."
- Reflection: "Write the Solution section as a principle, not a process — one clear lesson stated plainly, then illustrated with your specific story."

**Asset Spec**
One short paragraph describing exactly what the asset should be, what it should contain, and what format works best.

Thanks for reading!

I hope you found this process helpful in generating business content ideas.

If you made it this far, I’d love to hear from you as I continue to refine this newsletter.

If you have specific topics you’d like to hear about or follow-up questions to today’s post, please hit reply to this email or leave a comment below. I personally read and respond to every one.

See you next week,

Dickie

P.S. — The frameworks in this newsletter are the same ones we teach live inside AI Writing Skool. If you’d rather build alongside a community and get your questions answered in real time, come join us here.

Welcome to AI Operator: The leading newsletter for integrating AI into every department of your internet business.

Every Tuesday I’ll share exactly how we’re using AI across media, marketing, sales, operations, support, product, hiring, and fulfillment. None of it is theoretical. You will get the actual systems, SOPs, and workflows we run week to week inside of a portfolio that generates $700k/month on average.

If you want to build a leaner, faster, more profitable internet business — you’re in the right place.

Who this is for

This newsletter is built for three types of people.

• Revenue operators inside a bigger business. You’re running sales, ops, marketing, or fulfillment inside a company doing real revenue. You know AI should be making your team faster and leaner — you just need the playbook from someone else doing it.

• Solopreneurs running a 1-person internet business. Whether you run an agency, newsletter, sell digital products, or build software, chances are you’re wearing every hat. AI is the only way to scale without burning out or hiring a team you can’t afford yet.

• Aspiring operators who want to build something. You have a skill, an audience, or an idea — and you want to build a real internet business around it. You want to do it right from day one, with AI built into the foundation.

If you’re in any of those three camps — you’re in the right place.

The 10 problems I can help you solve

Chances are, if you’re reading this page right now, you found AI Operator because you were looking for a solution to one of these problems.

(We know — because we’ve experienced every single one of them inside our own portfolio.)

• Problem #1: 1-off prompts — Using AI reactively instead of building systems that run without you.

• Problem #2: Human-first thinking — Designing workflows around people when AI should be doing the heavy lifting.

• Problem #3: Burnout from wearing every hat — Doing everything yourself because nothing is documented well enough to delegate.

• Problem #4: Shallow work trap — Spending your best hours on tasks AI should be handling for $20/month.

• Problem #5: No proven playbook — Piecing prompts together yourself rather than getting a complete system from someone who already built it.

• Problem #6: Tool chaos — Lots of AI tools, nothing tying them together into a coherent operating system.

• Problem #7: Shiny object syndrome — Can’t tell what’s worth pursuing and what’s just hype.

• Problem #8: Opportunity overwhelm — Too many places AI could work in your business, no mental model for where to start.

• Problem #9: Incorrect understanding of AI’s best use cases — Knowing AI is powerful but applying it to the wrong things.

• Problem #10: The Learning Trap — Learning about AI constantly and never doing anything useful with it.

What you get every week

Every Tuesday you’ll get one issue built around one of three formats.

1. The Step-by-Step SOP — A fully documented system from inside the portfolio. The exact steps, in order, with nothing left out. Copy it, install it, run it the same week.

2. The Breakdown — A deep dive into why we do things the way we do. The decisions, the tradeoffs, the reasoning behind a specific system or strategy inside one of our businesses.

3. The Reflection — Something I’m learning in real time running a $700k/month portfolio, and exactly what it means for your business.

Every issue ends with one thing you can implement the same day you read it.

Who I am

Here’s a quick rundown of how I got here.

I studied math, computer science, and finance at Princeton. After graduating in 2018 I went straight to BlackRock in New York City, where I worked as a junior hedge fund trader focusing on global macro markets. It was my dream job — the only job interview I’d ever had.

But 12 months in I realized working for someone else wasn’t going to be my long-term path. I wanted control over what I worked on and how much I could make from it. More specifically, I wanted something where if I gained more skills and worked harder, my income was uncapped.

That realization led me to start looking for new opportunities. But rather than quit immediately, I started to write online, hoping that would unlock some new paths for me.

I was right.

However, 9 months in I was ready to give up on my writing journey. I had barely 250 subscribers to my newsletter and a couple hundred Twitter followers. Basically no one knew I existed.

So in September 2020 I gave myself one last shot — a 30-day daily writing challenge on Twitter. And during those 30 days, everything changed. I started going viral day after day. I went from 100 followers to 2,000 basically overnight. And the challenge ended with more opportunity than I knew what to do with.

That challenge ended up leading to Ship 30 for 30 — the biggest and most successful cohort-based writing course on the planet. In 2021 I partnered with Nicolas Cole and we scaled it to $1,000,000 in revenue in our first year together.

That success led me to quit my job at BlackRock in 2022, and I haven’t looked back.

Since then, Cole and I have gone on to generate over $23 million in revenue across 7 different types of internet businesses:

• Ship 30 for 30 (Cohort-Based Bootcamps) — the largest cohort-based writing program on the planet.

• Premium Ghostwriting Academy (1:1 Coaching Program) — the best training for anyone looking to turn writing into a career.

• Write With AI (Paid Newsletter) — the #1 paid newsletter on Substack. 110,000 subscribers. $425K ARR.

• Typeshare & Ghostbase (SaaS) — everything you need to write online and write for other people.

• AI Writing Skool (Low-Ticket Community) — the leading community for AI-powered writers, creators, and internet builders.

• Velocity (Done-For-You Email Marketing Agency) — install our email systems directly into your business.

• Vortex (1:1 Consulting & Mastermind) — work with us directly to implement our playbooks into your business.

And AI Operator is where I share the exact systems that help me run this entire portfolio.

Start here

(Coming Soon)

Want to work directly with me?

Come join us inside AI Writing Skool — the leading community for AI-powered writers, creators, and internet builders.

⬇️ What you get when you join ($10,000+ of value) ⬇️

• ✅ Instant access to our exclusive AI model, “AI Cole” — trained on all of our content and programs

• ✅ Monday Hot Seat live call with Nicolas Cole

• ✅ Weekly AI & Tech Clinic with Mitch Harris

• ✅ Monthly writing templates, AI prompts, and Claude .Skills

• ✅ 50% off all live 2-week bootcamps

• 🎁 BONUSES (worth $5,000+)

• 🎁 Daily Q&A channel with Nicolas Cole and the AI Writing Skool team

• 🎁 Rapid-fire introductions between community members

• 🎁 Instant access to our full suite of legacy mini-courses

Click here to join AI Writing Skool