Most AI maturity models measure the wrong thing.This one measures what AI does to the people.

Q1. When your people use AI recommendations in their work, do they understand what automation bias is and how it affects their judgment?

No. AI is treated as a tool like any other. Some individuals are aware, but there is no organisational response. Training covers this. People can articulate the risk. Processes are designed to counter it, for example people form their own judgment before seeing AI advice. We measure human-AI decision quality and iterate workflows based on findings.

Q2. Has anyone in your organisation studied what happens to decision quality over the course of a day spent working alongside AI systems?

No. We have anecdotal awareness that fatigue affects judgment. We have discussed attention depletion and cognitive load in the context of AI. We have built breaks, task rotation, or attention restoration into AI-assisted workflows. We monitor cognitive load across work sessions and adjust accordingly.

Q3. If you removed the AI system tomorrow, would your people's independent judgment have been degraded by prolonged reliance on it?

We have never considered this. We suspect it might be an issue but have no evidence. We are aware of the research on AI bias inheritance and have discussed it. We run periodic exercises where people make decisions independently of AI to maintain skill. We measure independent decision quality over time and can demonstrate it has not degraded.

Q4. In your AI-assisted processes, is it clear where the AI recommendation ends and the human decision begins?

No. It is opaque. AI-assisted decisions are labelled, but the human role is not defined. Decision rights are assigned. It is clear who can override and when. Overrides happen in practice, not just in policy. Override accuracy is measured. The architecture evolves based on outcomes.

Q5. Has a human in your organisation ever actually overridden an AI recommendation, and was that override respected?

We do not track this. It may have happened but we have no visibility. Overrides are possible in theory. Yes. We have evidence of overrides, and the overriding person was supported. We analyse override patterns to identify both unwarranted deference and unwarranted resistance.

Q6. Could you demonstrate to a regulator that your "human in the loop" is a cognitive reality rather than a compliance checkbox?

No. We could show the policy exists. We could show that decision rights are assigned and people understand them. We could show evidence of humans exercising judgment, including disagreeing with AI. We could show outcome data proving that human oversight adds measurable value to decision quality.

Q7. If you asked your board, your middle managers, and your frontline AI users the same question about your AI governance readiness, would they give the same answer?

We have never compared these perspectives. Probably not, but we have not checked. We know there is a gap and we have discussed it. We actively seek the frontline view and surface discrepancies to leadership. Governance assessments routinely include all three layers and the differences are reported without defensiveness.

Q8. When was the last time someone in your organisation said "we do not know" about an AI risk and that was treated as an acceptable answer?

Never. Uncertainty is not tolerated. It happens informally but would not appear in a board paper. It happens and is accepted without consequence. It is actively encouraged. Near-misses are reported and analysed. Structural mechanisms exist to protect people who surface uncertainty or dissent.

Q9. Does your AI risk register reflect what is actually happening, or what you wish was happening?

We do not have one. It exists but probably reflects aspiration rather than reality. It is honest. We update it when practice diverges from policy. We test it against operational reality and correct discrepancies. Independent audit confirms alignment between register and practice.

Q10. Do your people believe the organisational narrative about AI ("it will augment, not replace"), or do they privately think it is nonsense?

We have not asked. We suspect scepticism but have no data. We have surfaced the gap between narrative and lived experience. Workers affected by AI are involved in designing how it is deployed in their area. We measure human flourishing alongside AI performance. Deployment that damages people is treated as failure.

Q11. Is there a difference between how your organisation manages the change (new tools, new processes) and how it supports the transition (identity, purpose, mastery)?

We do not distinguish between these. We manage the change. Transition support is not something we have considered. We recognise that transition is different from change and have begun addressing it. People have access to genuine transition support: reskilling with time to learn, safe spaces, managers trained to recognise transition stress. Autonomy, competence, and relatedness are explicitly considered in every AI deployment decision.

Q12. Would someone in a routine cognitive role in your organisation feel psychologically safe enough to say "I am worried AI will make my job redundant"?

No. That would be career-limiting. They might say it to a trusted colleague but not to management. They could say it without consequence. Management would actively respond with support and involvement. This kind of honest expression is expected, valued, and structurally protected.

Q13. If an AI-assisted process in your organisation produced a harmful outcome tomorrow, could you name the single person who would be accountable?

No. Responsibility is distributed across teams and committees. We could name someone on paper, but they might not agree or have the authority. The named person has the authority, budget, and access to govern the system. That person has already intervened under pressure, and the organisation supported them. Accountability is structural. Incentives, reporting lines, and escalation paths make it the path of least resistance.

Q14. When an AI system fails in your organisation, do you investigate for systemic causes or blame an individual?

We blame someone. We investigate but tend to find the nearest person to hold responsible. We understand the principle of systemic investigation but do not always follow it. We investigate systemically. Honest error reporting is encouraged. Just culture principles are embedded. Ownership, escalation, override, and audit responsibilities are all clearly assigned and tested.

Q15. Do your people who report AI governance concerns get protected or punished?

They would be punished, formally or informally. It depends who they are and what they report. They are not punished, but nothing happens. They are protected and the concern is investigated. Reporting is structurally incentivised. The organisation actively looks for what it is getting wrong.