Your business drawn as an org chart, with functions as boxes, a coordinator in the middle, and you at the top as the accountable seat. The output is a picture, not a system. Seeing the company you are already running is what stops you from automating a broken process.
Use it before you build or buy a single agent. Use it when you have a pile of tools and no structure, or when something you set up made a mess and you cannot say which part failed.
Do not use it to plan a hiring roadmap. This is a map of functions, not of people you intend to employ.
A rough sense of where your week actually goes. Not a job description, not an ideal-state plan. The messy version is better, because the boxes have to match what the business does now rather than what you meant it to do.
You are helping me draw my business as an org chart of AI agents.
I may be the only person in it. That is fine and expected.
Work through this in four steps. Do not skip ahead, and do not
automate or recommend tools at any point.
STEP 1. Ask me what my business actually does in a normal week.
From my answer, propose 6 to 8 FUNCTIONS as boxes. Use the real
work, not generic department names. Show me the list and let me
correct it before continuing.
STEP 2. For every box, apply this test and show your reasoning:
"Can this be finished in a few clear steps with a definite end?"
If a box fails, split it until each piece passes. Tell me plainly
which boxes failed and why. Do not soften this.
STEP 3. Name the coordinating seat above the boxes. Describe in
two sentences what it would gather, how it would sort by what
needs my decision, and what it would put in front of me.
STEP 4. Mark exactly ONE box as the first to fill, using two
criteria: it eats the most of my week, and it needs the least of
my judgment. Say why the runner-up lost.
Finish with the chart laid out as indented text, me at the top,
then the coordinator, then the sized boxes. Then one line naming
the first seat. Do not add an implementation plan.
Run it once, ten to fifteen minutes. Keep the output where you can see it, because the next two prompts both refer back to it.
Redraw it when the business changes shape, not on a schedule. A chart that describes last quarter is worse than no chart, because it looks authoritative while pointing at seats that no longer exist.
Six to eight boxes you recognise as your actual work, at least one of which got split because it was too big. A coordinator described in plain language. One seat marked first, with a stated reason the runner-up lost. If nothing got split, the sizing step was not applied honestly.
Naming boxes after departments a bigger company would have. Letting a box like "marketing" survive the sizing test intact, which it never should. Marking a first seat that needs your judgment, because those are the ones that feel most valuable and fail most often. Treating the chart as a plan to hire.
The sizing step is the whole prompt. Most people can list their functions in two minutes. Almost nobody splits them, and unsplit boxes are exactly where agents break, because a long connected job is the condition the research says they degrade under.
Source: Build to Thrive, The AI Blueprint, week of August 17, 2026.