All four platforms available
Talk to usSee a governed action

Readiness check

Twelve questions about how your agents change records.

The readiness check is a free, 12-question self-assessment for teams whose AI agents change records. It asks who may change what, which changes need a named approver, what is recorded for each action and how refusals are handled, then gives a plain result and the next step. It takes a few minutes and needs no sign-up.

It scores your practices, not our product. You can score well without elan1. No sign-up.

About five minutes · 12 questions in four groups · answers stay in this browser tab

Progress 0 of 12 answered

01Agents and access
Q1Inventory. Do you know which agents can change a record in a system of record? Question 1 of 12.
Q2Credentials. What credential does an agent write with? Question 2 of 12.
Q3Rules outside the model. Are the rules checked by code with no model in it? Question 3 of 12.
02People and limits
Q4Named approver. Above a threshold, does a named person other than the requester approve? Question 4 of 12.
Q5Limits. Is it written down what an agent may do alone, and what never? Question 5 of 12.
Q6Refusals. When a check fails, what happens? Question 6 of 12.
03Records and evidence
Q7One record. Can you show who proposed a change, which rule version applied, who approved and when? Question 7 of 12.
Q8Tamper evidence. Would you know if an old log entry were edited or deleted? Question 8 of 12.
Q9Hand-over. Could an outside auditor get the evidence for one change without system access? Question 9 of 12.
04Failure, residency and cost
Q10Failure direction. If your checking service goes down, do agent writes stop? Question 10 of 12.
Q11Residency and erasure. Do you know where agent logs live and how you would erase from them? Question 11 of 12.
Q12Cost. Can you say what each agent change costs, and whether refusals are charged? Question 12 of 12.

01Method

How the result is worked out.

Shown in full, so you can check it — or run it on paper.

12 questions, unweighted, out of 24. Answers stay in your browser unless you email the result.

  • Each answer scores 0, 1 or 2, in the order shown: the first answer 0, the last 2.
  • No question is weighted. Twelve questions, so the most you can score is 24.
  • The result is one of three bands. No percentile, no comparison with anyone else.
  • Nothing is sent anywhere. Copying or emailing the result happens on your device, from your own apps.

This check scores practices, not products. It is not an audit and not legal advice.

The three bands
ScoreBandWhat it says
0–8Start with an inventory.Agents may already be changing records you have not listed. Write down which ones, with what credential, and who owns each. Then pick one process where a person already signs off.
9–16The parts exist; the path is not single.You have checks and logs, but in different places, and some run after the fact. Next: rules outside the model, a named approver before the change, one record per change.
17–24Ready to put one process on a governed path.Your practices are in place. Can you now show them to an outsider, one change at a time, without giving access to your systems?

02Questions

About the AI agent readiness assessment.

// NEXT

See what one governed change looks like.

The walkthrough follows one write from intent to proof record, including the branch where it is refused.

See a governed actionRead the specimens