The AI Audit — AI in Crayon

~ score 0 — 21 · pick what's actually happening

The audit.

Pick the score that matches what's actually happening today — not what you wish was happening. Your running total updates in the corner.

01 / 7 Does the AI actually know your business?

0 We open a blank chat every time. No setup. 1 We have some saved prompts but use them inconsistently. 2 We’ve built a custom GPT or project that knows our business, and the team uses it consistently. 3 Same as 2, and we update it every few months as the business changes.

02 / 7 Do you give the AI the customer or job details before asking for help?

0 We just ask the question. No customer history, no job notes, nothing. 1 Sometimes we paste in details. Inconsistent. 2 Standard practice — customer file, job notes, or thread always go in before we ask. 3 Same as 2, and we have templates or shortcuts that make it automatic.

03 / 7 Have you told the AI what not to do?

0 Nope. AI can produce anything and we’ll probably use it. 1 Informal team understanding (“don’t trust the prices”) but nothing written down. 2 Written rules — what AI can do on its own vs. what needs your eyes. 3 Same as 2, and the rules are baked into the system so the AI enforces them itself.

04 / 7 Has the AI seen examples of work you’d be proud of?

0 We’ve never shown it any examples. 1 Sometimes we copy-paste an old email, but nothing in the system setup. 2 Three to five real samples are loaded into the system. AI knows what “good” looks like. 3 Same as 2, and we refresh the samples when the business evolves.

05 / 7 Do you tell the AI what “good” looks like, or just give it steps to follow?

0 We always give it step-by-step instructions. 1 Mix of both, no consistency. 2 We describe what we want done and let the AI figure out how to get there. 3 Same as 2, and we know when to use steps (regulated work) vs. outcomes (everything else).

06 / 7 Do you actually track whether the AI is helping?

0 We don’t measure anything. “It feels like it’s helping.” 1 Someone notices the wins informally. 2 We track two or three real numbers (hours saved, items drafted, complaints reduced) and check monthly. 3 Same as 2, and the numbers actually change what we do.

07 / 7 When the AI gets something wrong, do you fix the system — or just rewrite it?

0 We rewrite the output and forget about it. 1 Sometimes we update the prompt when something breaks badly. 2 We have a process — misses get logged, the system gets updated regularly. 3 Same as 2, and the AI is visibly better today than three months ago because of it.

~ your score, decoded

What the number means.

Four bands. Each has a specific next move. As you score, the band you're landing in highlights automatically.

0 — 7

You bought AI. You didn't deploy it.

You’re paying for a tool that isn’t really set up. That’s most business owners. Good news — nothing to undo. Bad news — a few weeks of focused work would change everything.

Next move ~ Pick the question you scored lowest on. Spend a week closing that one gap.

8 — 14

You started. The gaps are obvious.

You’re getting some value but nowhere near what’s available. Most owners who think they’re “doing AI” are actually here. Usually a couple of ingredients are solid and the rest are empty.

Next move ~ Find your two lowest. Fix the higher-cost one this month. Second next month.

15 — 19

You're past deployment.

The tool is working. The team trusts it. You’re refining, not building from scratch. Most businesses never get here.

Next move ~ Push on speed. Run more cycles, polish less per cycle. Leverage is now in throughput.

20 — 21

AI is part of how the business runs.

The system is high-functioning. The tool gets sharper every month because the feedback loop is live. Hard to imagine running the business without it.

Next move ~ Document what you’ve built. You’re a model — for your team, your peers, your clients.

The pattern most owners miss.

The two questions almost every operator scores lowest on are #4 (Examples) and #7 (Feedback Loop). Both are compounding investments — they don't feel urgent today, but they're the difference between an AI tool that stays frozen forever and one that keeps getting sharper every month.

~ send your read to reu

Want an honest read on your score?

Drop your email and hit send. Your score and answers go straight to Reu, you'll get a confirmation copy in your inbox, and he'll come back with a short read on your two lowest-scoring ingredients and what moving one of them would actually change for your business. Within 48 hours. Personal read. No pitch.

Your email Anything you want to flag? (optional)

Your results go straight to Reu at [email protected].

Is your AI actually working?