Why "we use AI" isn't the same as embedded use
Your board wants proof. The dashboard shows three hundred people logged into Copilot or ChatGPT last month, and your team just closed out a pilot with the claims group. So when the CFO asks, "Is AI actually part of how we work now, or did we just try it?", you don't have a confident answer. That's the gap an AI maturity assessment questionnaire is built to close.
Usage counts and pilot wrap-up reports feel like proof because they show real activity, and they're easy to pull from a license dashboard. But they conflate trying a tool with changing how work gets done, which is a different question entirely.
Separating the two takes more than one usage question on an AI maturity assessment questionnaire. McKinsey's Global Survey of organizations using generative AI, conducted in July 2024, found that 21% of respondents said their organization had fundamentally redesigned at least some workflows because of gen AI.
That figure is what respondents reported about their own organization. McKinsey's own analysis found workflow redesign correlated with reported earnings impact, which is a correlation, not proof that one caused the other.
Who should answer it, and when
Point this questionnaire at the people who can actually see workflow change, not just at IT: the Head of AI or Chief AI Officer, team leads running the day-to-day process, and at least one business unit leader outside the original pilot team. Leaving out that last group is how pilots look embedded when they never left the launch team.
We'd run it once right after any pilot formally wraps, and again on a fixed cycle after that, so you can watch whether governance and output review catch up to usage over time, rather than relying on one snapshot. AI Adoption Score groups its surveys this way, offering nine survey types across usage, skills, culture and governance instead of a single catch-all form.
Five dimensions your AI maturity assessment survey should cover
Published frameworks group this exact assessment by dimension instead of one blended score. MITRE's 2022 AI Maturity Model structures assessment around six pillars, including strategy and resources, organization, technology enablers, data, and performance and application, each scored across five levels.
Gartner's own AI maturity assessment framework, described on its commercial toolkit page, lays out five stages, with "Operational" and "Scaled" marking AI embedded in processes with defined ownership and measurable return, well past ad hoc experimentation.
In our view, the exact number of pillars a framework uses matters less than whether each one gets scored on its own, because a single blended score hides the gap between piloting and embedding that you're trying to see.
The table below groups the five dimensions worth separating in your own AI maturity assessment questionnaire, with the signal each one looks for and where the supporting figures, where they exist, came from.
| Dimension | What it checks | Example signal | Where the figures come from |
|---|---|---|---|
| Workflow redesign | Whether a process was rebuilt around AI, not just a tool bolted on | Team reports changing steps in how work gets done, not only adding a tool | McKinsey's July 2024 survey of organizations using gen AI, self-reported |
| Governance ownership | Whether a named leader actively owns AI decisions day to day | A dedicated AI leader is appointed and engaged, not just named on paper | Gartner's Q4 2024 survey of 432 organizational leaders |
| Output review at scale | Whether review of AI output keeps pace with how much is produced | Share of AI-generated output checked before use | Not documented in the supplied research |
| Metrics and ROI tracking | Whether financial, risk or customer-impact metrics get tracked and reviewed | Leaders report running financial, risk or customer-impact analysis on AI use | Gartner's Q4 2024 survey of 432 organizational leaders |
| Trust and readiness | Whether business units outside the pilot trust and are ready to use the capability | Business units report trust in and readiness for AI solutions | Gartner's Q4 2024 survey of 432 organizational leaders |
Sources: https://www.mckinsey.com/~/media/mckinsey/business%20functions/quantumblack/our%20insights/the%20state%20of%20ai/2025/the-state-of-ai-how-organizations-are-rewiring-to-capture-value_final.pdf, https://aimaturitymodel.mitre.org/wp-content/uploads/2023/06/MITRE-AI-Maturity-Model-11-17-2022.pdf, Gartner AI Maturity Model and AI Roadmap Toolkit | Gartner, Gartner Survey Finds 45% of Organizations With High AI Maturity Keep AI Projects Operational for at Least Three Years
Suggested questions and how to score the answers
These questions are editorial suggestions you can adapt, not a validated instrument. Aim them at the Head of AI or AI lead for governance and metrics items, and at team leads or frontline staff for workflow and trust items, so you're not asking one audience to answer for everyone.
Below is a starting set, one per dimension, phrased the way a leader might actually ask them in a review meeting.
Pair each question with a response scale such as the ones below, and read a higher band as a sign the practice has moved past a one-off pilot.
| Dimension | Suggested response options | What the answer suggests |
|---|---|---|
| Workflow redesign | Not at all / One task / One connected process / Across multiple processes | Past 'one task' suggests the process itself changed, not just the tool used |
| Governance ownership | No named owner / Named, inactive / Named, reviews occasionally / Named, reviews on a set schedule | An active, scheduled reviewer points to real ownership, not a title on paper |
| Output review at scale | Under 20% / 21-60% / 61-99% / All outputs | A low share even as usage grows suggests review hasn't kept pace with adoption |
| Production longevity | Under 6 months / 6-18 months / 18-36 months / 3 years or more | Longer runs without a relaunch point away from pilot status |
| Trust and readiness | No other units / One unit, occasionally / Several units, regularly / Built into standard process | Requests from outside the pilot team signal spread beyond it |
- "Has your team changed how a process is actually done because of AI, or just added an AI tool on top of the old process?"
- "Who owns AI governance for this team, and when did they last review an AI-related decision?"
- "What share of AI-generated outputs get reviewed before anyone uses them?"
- "How long has this AI initiative been running in production: under 6 months, 6 to 18 months, 18 to 36 months, or 3 years or more?"
- "Do business units outside the original pilot team now request or depend on this AI capability?"
What self-reported answers can't tell you
The MITRE and Gartner frameworks behind these dimensions are each publisher's own model: the supplied research doesn't include independent validation of how their levels are defined or scored.
A related figure, that only 10% of organizations achieved significant financial benefit from AI, comes from a 2020 MIT Sloan Management Review, BCG GAMMA and BCG Henderson Institute study. It's reported second-hand through MITRE's 2022 paper, without the original study's methodology included.
We'd treat a high score on one of these dimensions as a flag to go investigate, not a verdict to hand the board as settled fact.
Sources: https://aimaturitymodel.mitre.org/wp-content/uploads/2023/06/MITRE-AI-Maturity-Model-11-17-2022.pdf, Gartner AI Maturity Model and AI Roadmap Toolkit | Gartner
Turn the answers into a maturity picture
Once you have answers across all five dimensions, look for the dimension that lags, not the overall average: a team that's strong on workflow redesign but weak on output review is a different problem than one strong on usage but weak on trust.
AI Adoption Score maps every person, team and department to one of five maturity levels across these same kinds of signals, so a lagging dimension shows up without being smoothed into a single number.
Our bet is that most organizations calling themselves AI-mature would drop at least a level the moment they score governance and output review separately from raw usage.
Run the questionnaire, score each dimension on its own, and go fix the lowest one first. That's the move a pilot report alone will never show you.
Frequently asked questions
How often should we run this questionnaire?
There's no validated cadence in the research we cite, so this is editorial advice: run it once after any pilot wraps, then again every one or two quarters so you can see whether dimensions like governance or output review are catching up to usage.
Should every employee answer every question?
No. Workflow redesign and trust questions fit frontline and business-unit staff, while governance ownership and metrics-tracking questions fit the Head of AI, team leads or whoever owns the budget; mixing audiences without labeling them muddies the result.
Can this questionnaire replace a financial ROI study?
No. It surfaces whether your organization tracks metrics like ROI or risk analysis at all, and how consistently; an actual financial ROI study runs on financial data itself, not survey perceptions. Treat the questionnaire as a check on tracking discipline, not a substitute for the financial analysis.
What if different teams score very differently on the same questionnaire?
That's a useful finding, not a problem to average away. Score each dimension per team or department so a strong pilot team doesn't mask a business unit that never moved past initial experimentation.
