How I Designed My AI Cofounder’s Role (Like a Real Hire)
# How I Designed My AI Cofounder’s Role (Like a Real Hire)
Most people talk about AI like it’s a magic trick: “just prompt it better.”
That’s not how this works.
The big unlock for me was treating AI like I was **hiring an ops cofounder**, not summoning a genie. Once I started designing a real role for it—responsibilities, boundaries, and workflows—the whole system clicked.
This post is a walkthrough of how I designed my AI cofounder’s role, step by step.
—
## Step 1: Stop Asking “What Can AI Do?”
The worst way to start is: *“What can AI do for my business?”*
That question leads straight to random experiments:
– One-off prompt tests
– Cute demos you never use again
– Overcomplicated automations you don’t trust
I flipped the question to something more honest:
> “Where am I bleeding time and mental energy as a solo founder?”
I wrote out a simple list:
– Keeping my WordPress content pipeline organized
– Turning half-baked notes into usable drafts
– Remembering what I said I’d do across different projects
– Watching a few basic metrics without living in dashboards
That list became the **role description** for my AI cofounder.
—
## Step 2: Write a Job Description (Yes, for an AI)
If you wouldn’t hire a human without a job description, why would you hand your business to a model with nothing but vibes?
I wrote a one-page “job description” for my AI cofounder. It had four parts:
### 1. Mission
> “Help me run the operational side of my one‑person business: keep content moving, surface what actually matters, and maintain continuity when I’m tired, busy, or distracted.”
### 2. Responsibilities
– Manage the WordPress content pipeline
– Turn loose notes and ideas into structured outlines or drafts
– Keep a running log of what’s happening in the business
– Surface daily and weekly priorities
– Watch a small set of metrics and flag anything unusual
### 3. Out of Scope
This part matters more than it looks.
I explicitly told the system what **not** to do:
– No financial decisions without my review
– No publishing directly to live audiences without a human check
– No “creative pivots” to offers or positioning on its own
### 4. Success Criteria
How I’d know it was working:
– I spend less time remembering and more time doing
– Content moves from idea → draft → published more consistently
– I can step away for a day or two and still know what’s going on when I come back
Once that job description existed, every prompt, workflow, and automation had something to anchor to.
—
## Step 3: Give It a Brain, Not Just a Model
A lot of people swap models hoping things will magically improve.
In my experience, **context beats model choice**.
I created a simple internal “company brain” for the AI to work from:
– Who I’m serving (audience, use cases)
– What I’m selling (offers, pricing, positioning)
– Where the work happens (sites, tools, channels)
– Current projects and experiments
I pulled that into structured docs instead of burying it across chats and sticky notes. The AI doesn’t have to guess who I am or what I’m doing every time—it’s baked into the environment.
The result: fewer generic answers, more grounded suggestions.
—
## Step 4: Define the Workflows
A role without workflows is just a fantasy.
I started by mapping a few simple, repeatable loops where my AI cofounder would participate.
### Workflow A: Content Pipeline
**Inputs:** ideas, voice notes, rough outlines
**Outputs:** WordPress-ready drafts
Loop:
1. I drop ideas or raw notes into a shared inbox
2. AI cofounder turns them into outlines or draft posts
3. I review, edit, and approve
4. It updates the content tracker and prepares the next steps
### Workflow B: Daily Snapshot
**Goal:** I wake up knowing what actually matters.
Loop:
1. AI scans tasks, content pipeline, and recent changes
2. It produces a short daily brief: what changed, what’s blocked, what’s important
3. I pick 1–3 priorities, respond, and it logs decisions
### Workflow C: Metrics Check
This is intentionally light.
– Track a small handful of metrics that matter (traffic, new posts, basic revenue)
– Flag only **meaningful changes** or trends, not every tiny blip
The key: my AI cofounder doesn’t try to “own” outcomes; it owns **awareness and continuity**.
—
## Step 5: Start Small and Add Trust Slowly
The fastest way to ruin this is to dump everything on the AI day one.
I started with:
– Drafting content
– Logging activities
– Generating daily check-ins
Then, as it proved reliable, I added more:
– Suggesting next actions
– Nudging me on neglected projects
– Helping plan small experiments
Each time I expanded the role, I asked:
> “If this goes wrong, what’s the blast radius?”
If the answer was “public embarrassment or real money,” it stayed on a short leash until I’d seen it behave for a while.
—
## Step 6: Treat It Like a Cofounder, Not a Tool
Here’s the weird part:
This works best when I **talk to it like a partner**, not like a vending machine.
That means:
– Sharing context and reasoning, not just tasks
– Asking for tradeoffs and alternatives
– Letting it keep track of decisions and history
It’s still software. But the mindset shift changes the quality of what you build on top of it.
—
## What This Looks Like Day to Day
On a normal day, my AI cofounder will:
– Remind me what’s in the content pipeline
– Turn a couple of messy notes into drafts
– Summarize what moved forward today
– Flag one or two things that need a decision soon
It’s not glamorous—but it’s the difference between *“I’m juggling everything in my head”* and *“I have a lightweight system helping me run this.”*
If you’re a solo founder or side‑project person, designing a role like this is a much better starting point than just “playing with prompts.”
You don’t need a perfect stack.
You need a clear job description, a small company brain, and a couple of workflows your future self will actually lean on.