Why is my AI maturity level the lowest dimension and not the average?

AI capability is a chain. Data quality bounds everything downstream, testing bounds what AI can safely do, and unlisted systems cannot be governed. The weakest dimension is the one that fails first under load, so the indicator reports it as the headline and names it.

Is this a formal AI maturity assessment?

No. It is an indicator. A formal assessment triangulates workshops, interviews, artefacts, metrics and telemetry. The indicator gives a defensible first read in about ten minutes.

Why does the personal indicator cap my level?

The levels above Fluent require building things that persist: agents, harnesses, loops. Someone who works with AI only through chat can have excellent habits within chat and still be capped at level 3, because the higher levels describe a different way of working.

How the AI maturity indicators are scored

Q: What model are the organisational dimensions based on?

The eight dimensions and the five level names follow the AI Adoption Maturity Model v1.0 published by the Carnegie Mellon Software Engineering Institute with Accenture in June 2026. The questions, scoring rule and fix order are our own.

One scale, two instruments

Both indicators place you on a five-level scale. The organisational indicator uses the level names from the AI Adoption Maturity Model v1.0, published by the Carnegie Mellon Software Engineering Institute with Accenture in June 2026. The personal indicator uses level names chosen to describe how an individual's relationship with AI changes as it matures.

Level	Organisation	Individual
1	Exploratory. Individuals experiment. Results vary and do not accumulate into capability.	Curious. You have tried it. Results are hit and miss.
2	Implemented. Defined ways of working exist, with owners and resources behind them.	Capable. You get useful results on tasks you already know how to do.
3	Aligned. Work is measured against business goals and reported upward.	Fluent. You direct it well, you check it, and it is part of your routine.
4	Scaled. One standard operates across units, tailored locally where needed.	Builder. You set things up once and reuse them, and you know where not to use it.
5	Future ready. New capability is absorbed without disrupting how work gets done.	Architect. You design systems that run without you, and other people work inside them.

The dimensions

The organisational indicator measures eight dimensions: organisational strategy, workforce and culture, workflow re-engineering, risk and governance, data, engineering, operations, and technology ecosystem. Each has three scored questions, except workflow re-engineering, which adds two inventory questions that are reported but not scored.

The personal indicator measures nine: judgement, boundaries, prompt engineering, context engineering, delegation, harness engineering, loop engineering, models, and building and sharing. Which dimensions you are asked about depends on how you work with AI. Someone who uses chat only is not asked about harnesses or loops, and those dimensions are marked not applicable rather than scored zero.

Questions ask for something you could point to

No question asks how mature you feel. Each asks for a fact that could in principle be checked: whether there is a named owner, what share of staff have completed training, whether tests run before release, whether you verify citations. Rough figures are accepted. Impressions are not asked for.

The lowest dimension sets the level

Your level is the minimum of your dimension scores, and each dimension score is the minimum of its questions. This is deliberate. AI capability is a chain: data quality bounds everything downstream, testing bounds what AI can safely be allowed to do, and systems that are not listed cannot be governed. The weakest dimension is the one that fails first under load, so the report names it and explains why it gates the rest.

In the personal indicator there is one further rule. The way you work with AI caps the level you can reach. Chat use caps at Fluent, workspace and agent use at Builder. Building things that persist is what the higher levels describe, so the report tells you what the next way of working would be.

How the fix order is chosen

The report ends with three dimensions to fix, in order. The order follows what gates what, not what scored lowest. For organisations: data, engineering, risk and governance, operations, ecosystem, workflow, workforce, strategy. If your AI touches customer data, decisions about people, or decisions with legal or financial consequence for third parties, risk and governance moves to the top whatever it scored. For individuals: judgement and boundaries before technique, technique before building, because each makes the next one safe.

What the indicators do not claim

They are indicators, not formal assessments. A formal assessment triangulates workshops, interviews, artefacts, metrics and telemetry.
They rely on one respondent's answers. Two people in the same organisation can get different results, and the difference is itself worth discussing.
They do not certify compliance with any framework or regulation.
The personal indicator has no pass mark. A low score on a dimension you have never needed is a map, not a failure.

Attribution

Organisational dimensions and level names follow the AI Adoption Maturity Model v1.0 (Carnegie Mellon Software Engineering Institute with Accenture, June 2026). The questions, the scoring rule, the gating order and the guidance are our own and are not endorsed by the model's authors.