Arun Agrahri
How Much of Your Company Lives in 3 People's Heads?

AI 11 min read

How Much of Your Company Lives in 3 People's Heads?

The Locked-Context Audit

The models are smart enough. The bottleneck is the company-specific context locked in your senior people's heads. Score your company on 9 questions, then get the teardown: the closed loop, the 5 building blocks, the roles a system replaced, and what it ships, across a portfolio run with zero employees (cofounders + agents).

Score your company in 4 minutes

Score each question 0, 1, or 2, honestly. Higher means more of your company runs as a system instead of in someone's head. Add up all nine (max 18) and find your band below. Most operations-heavy companies land between 3 and 9.

Knowledge: is it in a system, or in heads?

If your most senior operator went dark for a month, how much of the business keeps running without them?(0 little / 1 some / 2 most of it)

Could a new hire get the WHY behind your key processes from a system, or only by asking a person?(0 only by asking a person / 1 mostly people / 2 from a system)

Is your company-specific judgment (pricing exceptions, edge cases, "we don't do it that way because…") written anywhere an AI could read?(0 it's in heads / 1 scattered / 2 yes, structured)

Workflow: clean and repeatable, or tribal?

Do your core workflows reliably produce the same output from the same input, without one specific person?(0 no / 1 sometimes / 2 yes)

How much of your team's repeatable work (triage, routing, the same reports, checking others' work) is documented or automated, not improvised?(0 under 20% / 1 20–60% / 2 over 60%)

When you add an AI tool, does it plug into clean, connected data, or into scattered SaaS no agent could query across?(0 scattered / 1 partial / 2 clean and connected)

Loop: do you measure and improve, or set-and-forget?

Do your automated processes measure their own output and feed it back to improve, or run the same way forever?(0 run the same forever / 1 some feedback / 2 a closed loop)

Is there a systematic check on the quality of what your systems produce (not a person eyeballing it occasionally)?(0 none / 1 ad hoc / 2 systematic)

When the AI gets something wrong, does the system get better next time on its own, or must a human catch it every time?(0 a human catches it every time / 1 sometimes / 2 it auto-improves)

Your score: — / 18

0–6 · Open Loop Most of your company lives in a few people's heads, and the bleed compounds every quarter. The teardown shows what closing it looks like.
7–12 · Half-Built You've automated; you haven't closed the loop. The missing feedback arrow is where the leverage is.
13–18 · Closing the Loop You're ahead of almost everyone. You may not need an audit yet, but the teardown shows the one layer most companies at your level still miss.

Guide Map

  1. 1 Read your score honestly
  2. 2 The bottleneck isn't the model
  3. 3 Here's the machine, running
  4. 4 It gets better while you sleep
  5. 5 The 5 building blocks
  6. 6 5 roles I never hired for
  7. 7 What the machine ships
  8. 8 What an open loop costs you

Also includes Roles I Never Hired For · You Can't Audit Your Own Blind Spots · Frequently Asked Questions

5,969

lines of self-audit

59

automated tests

15

cron jobs run the loop

0

employees