AI Operations Audit Template: Free 20-Question Self-Assessment

Most agencies know their AI ops are messy. They do not know how messy. This is the 20-question self-audit the AGL team runs with new clients. Score it honestly. The number will tell you where you are on the sprawl curve.

How to use this audit

Answer each question with a 0, 1, or 2. Zero means no, or never, or not really. One means partial, or sometimes, or in progress. Two means yes, consistently, with an owner. Maximum score is 40. Higher is better.

Do not answer for how you wish the shop ran. Answer for how it actually ran last week.

The 20 questions

Section 1: Context ownership

1. Every active client has a single documented source of truth for brand voice, offers, and current campaigns.
2. That source of truth was updated in the last 14 days.
3. When a new hire starts on an account, they can get to production output in under 5 business days using the source of truth alone.
4. Client meeting notes are captured and connected to the source of truth within 24 hours.
5. There is one named owner per client account for context updates.

Section 2: Tool stack sanity

6. You can list every AI tool your team uses from memory in under 60 seconds.
7. Every AI tool in the stack has one named human owner.
8. Every AI tool has a documented job. Not "help with content." A specific job.
9. In the last 90 days, at least one tool was removed from the stack for redundancy.
10. Total AI tool count is 8 or fewer.

Section 3: Workflow discipline

11. The top 5 recurring workflows on each client account are documented as input, process, output.
12. Two operators running the same workflow produce output that looks consistent to a client reviewer.
13. Workflow updates are logged in a place the whole team can see.
14. AI-produced output is reviewed by a human before it reaches a client.
15. When a workflow breaks, the fix updates the source of truth, not just the individual chat.

Section 4: Measurement and governance

16. You know, within 20 percent, how many hours per week your team spends on AI context setup.
17. You track a consistency signal on client output, not just a volume signal.
18. There is a recurring meeting, at least monthly, to review AI ops health.
19. New tools go through a documented intake before joining the stack.
20. You can name the last 3 things you removed, changed, or improved in AI ops in the last quarter.

Scoring rubric

Add your total. Compare to the bands below.

32 to 40: governed. Your AI ops are ahead of most agencies. The gains are compounding. Focus is on the last 20 percent of consistency and on scaling the model to new accounts. You are not the audience for a rebuild. You may still benefit from an outside pressure test.

24 to 31: functional but leaking. You have real systems in place. They work most days. The leaks show up in busy weeks, in new hires, and in inconsistent client output. This is the band where a targeted install of an AI project manager pays back in 60 to 90 days.

16 to 23: sprawl in progress. You have tools and workflows, but the ownership is blurry. Context lives in operators' heads. Quality is a function of who happened to run the task. This is the danger zone. Every new client makes it worse until you install a system of record.

8 to 15: full sprawl. Your AI stack is a liability. Operators are working around the tools. Client complaints about consistency have started. New hires take 60-plus days to ramp. The fix is not another tool. It is a rebuild starting with context ownership.

0 to 7: pre-system. You are early. This is not shameful. Skip the sprawl by installing the operating model before you install more tools. Start with one client, one context spine, one workflow. See the 14-day install guide.

If you scored under 24 and you have added more than 2 AI tools in the last quarter, the score will keep dropping until you install an owner. New tools without an owner accelerate sprawl.

Which questions to fix first

The audit is deliberately weighted. Questions 1 through 5, on context ownership, are the highest-leverage. If you score below 6 in that section alone, fix that before touching the rest.

The 3 fastest wins in section 1: pick one client, write a single-page source of truth for the account, and assign an owner with a weekly 30-minute update ritual. That alone will move most agencies from a 12 to an 18 in 30 days.

The audit is not about scoring perfectly. It is about seeing the shape of the leak. Every agency has a shape. Fix the biggest leak first, not the easiest one.

What the audit does not measure

This audit is a snapshot. It does not measure your AI ROI, your team morale, or your client retention. Those are outcomes. The audit measures the operating conditions that produce those outcomes. Fix the conditions and the outcomes follow within a quarter.

Go deeper with the Sam diagnostic

The 20-question audit is a coarse read. For a personalized score with recommendations by account, run the free 3-minute diagnostic at /sam. Sam gives you a shop-specific number, the top 3 leaks, and where to start.

For the strategic context, read the canonical page on AI operations sprawl.

The bottom line

You cannot fix what you have not scored. Twenty questions, 10 minutes, honest answers. The number tells you whether you need a tune-up or a rebuild. Either way, the next step is the same: pick one client, install context ownership, and measure the recovery.

Talk to an expert today from Chicago, IL: book a strategy call.