Using AI is easy. Running on it is measurable.
AI First Matrix measures how far a company has moved from using AI tools to running on AI: 11 dimensions in three layers, every answer graded by the evidence behind it, and a plan that starts with what holds you back.
Specimen reading · worked example from the method
A manufacturer with 1,200 employees. Plenty of AI use; little of it built into the work.
27
Level 2 · Assisted
A plain average would say 31. Uneven profiles cost points.
A
Foundations
Can we?
40
Strategy & leadership
55
Data & context
40
Platform
45
People & fluency
35
Governance & risk
30
B
Adoption
Are we?
23
Usage
60
Workflow redesign
25
Agentic operations
10
Offering & market
20
C
Value
Does it pay?
17
Value discipline
20
Realised impact
15
What makes a reading trustworthy
Most AI maturity scores are one executive's opinion. Here the score depends on what can be shown, and it tells you how far to trust it.
- Evidence is graded
- Every answer carries a grade from A to D. What one person says counts for less than what the data shows, and the score says how sure it is.
- Three questions, kept apart
- Can we? Are we? Does it pay? One blended number hides which of them is the problem.
- The weakest link decides
- A strong area cannot make up for a weak one, so the plan starts where a fix is worth most.
- Built for agents
- Whether software agents can safely carry out delegated work is measured in every layer, not added at the end.
Evidence grades
Each indicator is scored 0 to 4 against written anchors. The grade of the evidence caps how high the score can go — which is why a self-assessment gives a range, not a number.
| Grade | What counts | Score up to |
|---|---|---|
| A | Telemetry or assessor-verified artifact | 4 of 4 |
| B | Artifact provided, unverified; or two sources agree | 4 of 4 |
| C | Multi-rater survey consensus | 3 of 4 |
| D | Single informant claim | 2 of 4 |
Where you need to be
The top of the scale is not the goal for everyone. How exposed your business model is to AI sets the level you need — the specimen, with high exposure, is two levels short.
L1
Exploring
L2
Assisted
L3
Embedded
L4
Agentic
L5
AI-First
Rows: business-model exposure to AI · columns: maturity level