AI-Native Government

Can your agency answer for its AI?

A five-minute readiness assessment for public-sector leaders. Six questions place your agency on the AI-Native ladder and show you the one thing to fix first.

Start the assessment

Every governed decision can answer four questions

Whathappened
When ithappened
Who setthe rule
Why itwas valid

The stakes

When an AI-influenced decision is challenged — in an appeal, an audit, a records request, or the press — can you show what happened, why, and who was responsible? Most agencies cannot, yet. This assessment shows how ready yours is, and exactly where it is exposed. Because in the public sector, you are not AI-native until you are governed.

What you get

Your result, in five minutes

Your rung on the AI-Native ladder: AI-Assisted, AI-Enabled, or AI-Native.

A six-dimension profile showing where you are governed and where you are exposed.

Your single weakest link — the place to start, whatever your overall rung.

A plain-language guide to the four questions every governed AI workflow can answer.

How it works

Three steps

01
Answer six questions about how your agency uses AI. About five minutes.
02
Enter your email to unlock your rung and full profile.
03
Get your result, the guide, and the AI Native newsletter for public-sector leaders.

Who it is for

The whole AI project team

For the people inside state and local agencies who carry the risk: technology, policy, operations, and procurement leaders on the same AI project. Score it alone, or with your team. The gaps between your answers are often the most useful part.

TechnologyPolicyOperationsProcurement

Start the assessment

The six questions

Score your agency

Six questions place your agency on one of three rungs — AI-Assisted, AI-Enabled, or AI-Native — and show the single dimension where you are most exposed.

Q1 What happened, the record
When AI takes or shapes an action, is there a record of what it did, captured as it happens, not reconstructed later?
No AI activity to record here yetAI helps individuals but leaves no real trail; we would reconstruct from logs, emails, and memorySome logs exist in vendor tools, but they are partial and not readily reviewableA contemporaneous record captures what each AI action did, in context, in a system we control
Q2 When it happened, the timeline
Can you establish exactly when an AI action occurred, with a timestamp that cannot be altered afterward?
No AI events to time here yetTiming is approximate or absent; we could not defend a precise sequenceTimestamps exist in vendor systems but could be edited or are not authoritativeEvery AI action carries a tamper-evident timestamp we can stand behind, across systems
Q3 Who set the rule, the accountable owner
For any AI action, can you name the human or role who authorized the rule behind it, with permissions bound to that authority?
No AI rules in force here yetAI is used informally, with no defined authority and no named ownerA written policy exists, but permissions are not enforced per role and ownership is unclearEvery AI action maps to an accountable owner and authorized role, enforced in the system
Q4 Why it was valid, the condition and the proof
Before an AI action executes, is a defined condition checked and the proof of it preserved?
No AI actions to validate here yetAI acts or advises with no defined precondition and no preserved proofSome conditions are checked, but proof is incomplete and corrections overwrite the originalEvery consequential AI action is gated on a defined condition, with proof and correction history preserved
Q5 Human oversight, meaningful approval
Where a human must approve an AI action, is the review real, or a rubber stamp?
No AI outputs entering decisions here yetAI outputs are accepted with little or no review, or review is nominalHumans are formally in the loop but often approve without enough context, and escalation is unclearApproval is bound to the exact action, structured around what the human must see, with escalation and a record
Q6 Portability and vendor accountability, whose proof is it
Do you own your trust layer, or does it live inside a vendor platform you cannot fully query, audit, or take with you?
No AI procured or deployed here yetAI arrives ad hoc, with no governance, audit, or data-ownership requirementsProcurement recognizes AI risk, but contracts do not require audit access or portable proofEvery AI acquisition requires transparency, audit access, and that proof and records live in systems we own

About youWhich best describes your role?

Which best describes your organization?

Your nameName of your organization or agencyWork emailWe will send your result and the AI Native newsletter for public-sector leaders. Unsubscribe anytime.

Not ready for the full assessment?

Get the newsletter instead. AI Native is a weekly read for public-sector leaders putting AI into real work without losing control or public trust.

Subscribe to the newsletter

Your result, in five minutes

Three steps

The whole AI project team

Q1 What happened, the record

Q2 When it happened, the timeline

Q3 Who set the rule, the accountable owner

Q4 Why it was valid, the condition and the proof

Q5 Human oversight, meaningful approval

Q6 Portability and vendor accountability, whose proof is it

Not ready for the full assessment?