I built a 10-person team this weekend. I didn’t hire anyone.
Ten specialists, three teams: strategy, marketing, design. They sit inside a folder on my laptop. Every one of them has a brief, a job description, a tone. I can call any of them in by name.
Everything on this site (the home page, the about page, the article you’re reading) went through one or more of them.
I keep getting asked who designed the site, who writes the SEO, who runs the content strategy. The honest answer is: I do, and I don’t. The longer answer is: I built a team of AI specialists, and the output looks nothing like what I’d ship alone.
If you’re using AI as a tool and wondering why your output still feels generic, this is the gap.
The unlock isn’t a better prompt. It’s a better team.
Five moves, each one a direct violation of how most operators are using AI today.
Stop asking one AI to be everything
| Standard advice | What actually works |
|---|---|
| Give ChatGPT a long prompt. It can handle anything. | One AI per role. Ten roles, ten AIs. |
| Cram all context into one conversation. | Each specialist has its own context, its own voice. |
You ask one general-purpose model to wear ten different hats in one conversation. It does all ten badly. Strategy bleeds into copywriting. SEO advice contradicts brand voice. The output is generic because the role is generic.
The fix is the same fix every growing company makes when it scales past 5 people: specialists. Not one ten-armed generalist. Ten specialists, each with one job and one voice.
I have ten of them. Each one is a separate .md file in a folder my AI agent can read. When I need a tagline, I call /writing-style. When I need a keyword strategy, I call /seo. When I need to know what to write next, I call /content-pillars. They don’t bleed into each other.
Don’t ask one AI to be ten things. Build ten AIs, each with one thing.
Brief them like new hires, not like search queries
| Standard advice | What actually works |
|---|---|
| Write a clear prompt with the task. | Write a job description with the role. |
| Add context every time you need it. | Build the context into the specialist once. |
The shift: stop writing prompts. Start writing briefs.
Every one of my specialists has the same structure as a job description. Who they are. What they care about. What they don’t. How to brief them. Reference material with live examples of good and bad work.
The /writing-style specialist isn’t “tell me how to write.” It’s a structured doc: my writing voice, the 8 rules, 9 banned words, and a library of approved copy to reference.
That’s not a prompt. That’s an employee handbook.
The result: I don’t have to re-explain context every conversation. The specialist already knows.
Write your AI a job description, not a query.
Organize them into teams, not a roster
| Standard advice | What actually works |
|---|---|
| Keep a list of useful prompts. | Build an org chart with specialists and teams. |
| Remember which prompt to use when. | Call the team wrapper, let it route to the right specialist. |
A list doesn’t scale. I have ten specialists. If I had to remember which one to call for any given problem, I’d default to none of them.
So I organized them like an actual org:
Strategy team (4 specialists): Vision, GTM Plan, Market Research, Competitor Analysis. The ones I call when I’m asking what and why.
Marketing team (3 specialists): Writing Style, Content Pillars, SEO. The ones I call when I’m asking what to ship.
Design team (3 specialists): Design System, Graphic Design, Product Design. The ones I call when I’m asking how to ship it.
Each team has a wrapper: /strategy, /marketing, /design. When I’m not sure who I need, I call the wrapper and it pulls the right specialist.
This is the same logic any growing company uses. Pods, squads, teams. The point isn’t the headcount. It’s the mental shortcut. I think “I need marketing input” and I know exactly where to go.
Specialists need a chain of command. Even AI ones.
Give them taste, not tasks
| Standard advice | What actually works |
|---|---|
| Give the AI a clear task with success criteria. | Give the AI a sense of what good looks like. |
| One task, one prompt, one output. | One specialist, infinite tasks, consistent voice. |
This is the move most people are missing. Tasks are downstream. Taste is upstream.
The /writing-style specialist isn’t told “write me a LinkedIn post.” It’s told what good output looks like, what bad output looks like, the exact rhythm I want, the words I never use, and examples of approved copy. When a task arrives, the specialist already knows what to aim for.
Same for design. /design-system doesn’t get told “design me a card.” It gets told the color tokens, the typography scale, the spacing rhythm, the dark-mode rules. Then any card it designs respects the system without me having to micromanage.
Same for strategy. /vision knows what the site is for, who it’s serving, what success looks like in 6 months. So when I ask “should I write this?” it can answer in the context of the actual mission, not in the abstract.
You can’t shortcut this. Taste-first specialists take longer to build than task-first prompts. But they get sharper with every brief.
Tasks expire. Taste compounds.
Manage them, don't operate them
| Standard advice | What actually works |
|---|---|
| Iterate on your prompt until the output is right. | Give the specialist feedback, ask for v2. |
| You are the prompt engineer. | You are the editor and the director. |
That’s how you use a tool. It’s not how you run a team.
Once I had ten specialists, my job changed. I stopped writing copy and started briefing. I stopped optimizing prompts and started giving feedback.
When /writing-style gives me a draft I don’t love, I don’t rewrite the prompt. I tell it what’s off, the same way I’d tell a copywriter, and ask for v2. When /design proposes a layout that fights the brand, I don’t redesign. I point at the design system rule it broke and ask it to retry.
The shift is small but the change is huge. I’m not the bottleneck on every output anymore. I’m the bottleneck only on decisions and direction.
Manage the team. Don’t operate the tool.
The Specialist Stack
10 specialists, 3 teams, 1 operator
- 01 Vision · Strategy
- 02 GTM Plan · Strategy
- 03 Market Research · Strategy
- 04 Competitors · Strategy
- 05 Writing Style · Marketing
- 06 Content Pillars · Marketing
- 07 SEO · Marketing
- 08 Design System · Design
- 09 Graphic Design · Design
- 10 Product Design · Design
The frame shift I didn’t expect: I stopped thinking in tasks and started thinking in briefs.
Tasks are small (“write a LinkedIn caption”). Briefs are big (“here’s the context, here’s what good looks like, here’s the constraint”). Briefs take five minutes to write and produce ten times the output. Tasks take thirty seconds and produce one piece of work that’s almost-but-not-quite right.
I now write more briefs than I write content. The team produces the content. I direct.
I stopped thinking in tasks and started thinking in briefs.
If you’re an operator who’s been using AI as a smarter Google, try the team version. Pick one role, your weakest one, and write it a job description. Not a prompt. A job description. Then call that specialist by name for a week.
The output won’t sound generic anymore. It’ll sound like the specialist you briefed.
This site is the proof. Ten specialists built it. None of them exist.
A tiny signal that this resonated. No account needed. Just a tap.