How do I turn a recurring AI task into a repeatable system?
Short answer
Build an intake: a fixed master prompt plus a short list of scope questions you answer before the AI starts. Write the prompt in six lines (role, context, task, constraints, format, example), add about five questions that set the boundaries of the job, and save it under the task's name.
I caught myself about to type the same 600-word research request into AI for the third time this month. Different subject each run, the same instructions pulled from memory, slightly worse on every pass.
A blank prompt box is the most expensive habit in my business. Every recurring task lived there until I started writing intakes instead.
Where does the time really go?
Here is what a year of building AI systems for founders has taught me. The work that drains you is rarely the hard work. It is the repeatable work you keep re-explaining from scratch. Anything you ask AI to do more than once deserves an intake.
An intake has two parts: a fixed master prompt, and a short list of scope questions you answer before the AI touches anything.
What goes in the master prompt?
The master prompt follows one shape:
- ROLE: who the AI is for this job
- CONTEXT: what it is for, who reads the result
- TASK: the one specific ask
- CONSTRAINTS: limits, budget, what counts as in-scope
- FORMAT: what the output should look like
- EXAMPLE: one line of good output, so quality is not a guess
The example line does more work than it looks like. It shows the AI the standard instead of describing it, and it gives you something concrete to check the result against.
What are scope questions?
Scope questions set the boundaries of the job before it starts. For a competitive research pull, mine read like this:
- What counts as an active account?
- How deep do we go per profile?
- How many targets?
- What does "leading" actually mean here, raw follower count or what an audience responds to?
- Which findings are shareable, and which stay internal?
Answer those once, in writing, and the next run starts from a known line instead of a blank one.
Why does scope matter so much?
The bottleneck has moved. A year ago the real question was whether AI could do the task. Now it can. The new constraint is whether you told it the right boundaries before it started. Vague scope in, generic work out, every time. Your intake is where the judgment lives.
It is also what lets you hand the work off. Once the judgment is written down, the person running the intake no longer has to be you.
Do this today
Ten minutes, one recurring task.
- Pick one task you have handed to AI at least twice.
- Open a blank doc and write the fixed prompt using the six lines above.
- Add five questions that define the scope of that job.
- Save it as [task]-intake.
Next time the work shows up, you fill in answers instead of rebuilding the request.
Meredith's rule
Anything you ask AI to do twice deserves an intake.
Questions
What should a good AI prompt template include?
Six parts cover a recurring job well: the role the AI plays, the context and audience, the single task, the constraints, the output format, and one example line of good output. The example sets the quality bar so the result is not a guess.
How do I get AI to stop giving me generic research?
Define the scope before it starts. Write down what counts as in-scope, how deep to go, how many targets to cover, what "leading" means for this job, and which findings stay internal. Vague scope produces generic work. Clear boundaries produce output you can use.
Can someone on my team run my AI prompts for me?
Yes, once you turn them into intakes. Save the fixed master prompt and the scope questions in one document named for the task. Anyone can fill in the answers and run it, because the judgment now lives in the document instead of your head.
Related service
Agentic System Install
Want this worked out for your business?
Bring the real problem to a working session and leave with a plan.
Book a session