An AI enablement maturity model gives an organization a structured way to assess where it currently stands across the dimensions that determine AI capability. Most published maturity models are designed for enterprise; mid-market needs a focused subset. This is the five-dimension, five-level model we use across mid-market engagements to anchor the planning conversation. It complements our AI enablement roadmap pillar — the roadmap describes the destination; this describes where you are.
Why have a maturity model at all
Three reasons mid-market companies benefit from a structured maturity assessment:
- It separates aspiration from reality. Leadership often believes the company is further along on AI than it actually is, because individual employees use AI tools and that gets generalized to "we are AI-enabled." The assessment surfaces the gap.
- It produces a defensible plan. Without the assessment, planning conversations cycle around opinions about what should be next. With it, the plan flows from the dimensions that score lowest.
- It enables progress measurement. Re-running the assessment quarterly shows whether the organization is moving forward, stagnating, or backsliding.
The published maturity models we have seen — Gartner's, Microsoft's, various consultancy versions — are designed for enterprise organizations. They have 8-12 dimensions and 5-7 maturity levels, with rubrics that assume staffed AI organizations. Applied at mid-market, they produce assessments that score everyone "early-stage" without useful differentiation.
What follows is the focused five-dimension, five-level version. It produces meaningful differentiation at mid-market scale.
The five dimensions
| Dimension | What it measures |
|---|---|
| Strategy | Whether AI ambition is named, scoped, and resourced |
| People | Whether the organization has the skills to deploy AI capabilities |
| Data | Whether AI applications can access the data they need to be useful |
| Platform | Whether the technical foundation supports building and operating AI |
| Governance | Whether AI use is bounded by clear policies and processes |
These five collectively determine an organization's ability to deploy production AI capabilities. Weakness in any one of the five constrains the others — strong strategy with weak data produces ambition without execution; strong platform with weak governance produces velocity that creates risk.
The five levels
| Level | Name | Description |
|---|---|---|
| 1 | Aware | The organization knows AI matters; specific actions are individual |
| 2 | Active | Specific AI initiatives are underway; coordination is informal |
| 3 | Coordinated | AI work has structure; an owner exists; the foundation is being built |
| 4 | Capable | Multiple AI applications in production with shared infrastructure and governance |
| 5 | Optimizing | Continuous improvement of AI capability; AI is core to how the organization operates |
The levels are not strict ratings — an organization can be Coordinated on People and Active on Platform. The composite picture matters more than a single number.
The assessment rubric
For each of the five dimensions, the criteria for each level:
Strategy
| Level | Criteria |
|---|---|
| 1 — Aware | Leadership talks about AI. No named ambition or specific initiatives. |
| 2 — Active | One or more AI use cases identified and being explored. Resourcing is ad hoc. |
| 3 — Coordinated | Named AI ambition (3-5 priority use cases). Executive sponsor identified. Budget allocated. |
| 4 — Capable | AI roadmap aligned with broader product/business roadmap. Multiple production initiatives. Investment level adequate. |
| 5 — Optimizing | AI strategy treated as enterprise-strategic. Continuous re-evaluation of opportunity surface. AI ambition shapes hiring, partnerships, M&A. |
The strategy dimension is the easiest to overrate. Many mid-market companies rate themselves "Coordinated" because leadership has stated AI is important. The criterion is whether ambition has been concretely scoped and resourced, not whether it's been declared.
People
| Level | Criteria |
|---|---|
| 1 — Aware | Some employees use AI tools individually. No coordinated capability. |
| 2 — Active | A small number of engineers have built AI prototypes. Skills are concentrated in 1-3 individuals. |
| 3 — Coordinated | Named AI lead or coordinator. Multiple engineers actively building AI applications. Skill development is intentional. |
| 4 — Capable | An AI capability function exists (Guild, Hub-and-Spoke, or Dedicated team per our CoE structure piece). Skills distributed across multiple teams. |
| 5 — Optimizing | AI skills are a baseline expectation across engineering. Specialized AI roles exist. Learning is continuous and supported. |
People is the dimension that most often constrains progress. The right skills must exist before the platform and applications can be built well.
Data
| Level | Criteria |
|---|---|
| 1 — Aware | Data is fragmented across systems. AI applications would have difficulty accessing it. |
| 2 — Active | Some data flows have been built for specific AI use cases. Pattern is per-use-case. |
| 3 — Coordinated | Common data access patterns established. The AI applications that need data can get it without per-application integration work. |
| 4 — Capable | Data is well-structured for AI use cases. Quality, governance, and lineage are tracked. New AI applications can be built on existing data infrastructure. |
| 5 — Optimizing | Data infrastructure evolves with AI capability. Feedback loops between AI applications and data quality improve both. |
Data maturity is a long lead-time investment. Companies cannot rapidly move from level 1 to level 3 — the work involves structural changes to systems-of-record and integration patterns that take months to years.
Platform
| Level | Criteria |
|---|---|
| 1 — Aware | No common AI infrastructure. Each project starts from scratch. |
| 2 — Active | Some shared components emerging (an LLM provider relationship, a vector DB choice, etc.). Pattern is informal. |
| 3 — Coordinated | Common platform components exist: LLM provider, observability, evaluation infrastructure, retrieval infrastructure if relevant. |
| 4 — Capable | Platform supports rapid building and deploying of new AI applications. Operations are mature. Cost monitoring and management are in place. |
| 5 — Optimizing | Platform evolves continuously. New AI capabilities (agents, multi-modal, etc.) are absorbed without re-architecture. Cost-quality optimization is ongoing. |
The platform dimension is what platform engineering teams build. Investment here pays back across all AI applications.
Governance
| Level | Criteria |
|---|---|
| 1 — Aware | No AI-specific policies. AI use is implicit. |
| 2 — Active | Acceptable use policy drafted. Vendor evaluation happens but is per-team. |
| 3 — Coordinated | Documented policies in place: AUP, vendor evaluation, incident response, risk register. Owner identified per our governance framework. |
| 4 — Capable | Governance is integrated into engineering practice. New use cases run through documented review. Quarterly review cadence. |
| 5 — Optimizing | Governance evolves with AI capability. New risks identified proactively. Audit and compliance posture is mature. |
Governance maturity often lags platform maturity at mid-market. The pressure to ship features outweighs the pressure to govern them, until something goes wrong. Catching up after an incident is more expensive than building governance proactively.
Composite scoring
Each dimension scored 1-5 produces a composite of 5-25.
| Composite | Stage | Typical mid-market posture |
|---|---|---|
| 5-9 | Pre-program | AI work happens individually; no coordination |
| 10-14 | Building | Coordination beginning; first applications taking shape |
| 15-19 | Operating | Mature mid-market AI capability |
| 20-25 | Leading | Enterprise-grade capability; rare at mid-market |
A typical mid-market company starting structured AI work falls in the 8-12 range. After 12 months of effort following the enablement roadmap, 14-18 is achievable. Above 20 starts requiring enterprise-scale investment that most mid-market companies do not justify.
The composite hides important variation. Two companies at composite 15 can look very different — one strong on Strategy/People/Platform but weak on Data/Governance, another the inverse. The composite is the headline; the per-dimension picture drives the planning.
How to run the assessment
A 90-minute working session with the right participants:
Participants (3-6 people):
- The executive sponsor of AI work (CTO, VP Eng, or whoever owns it)
- The named AI lead or platform engineering lead
- A product representative
- Optionally: data lead, security lead, governance lead if those roles exist
Pre-work: each participant scores the five dimensions independently before the session, with brief notes on why.
Session structure:
- 15 min: align on the assessment criteria
- 45 min: discuss each dimension; share scores; understand where they differ; converge on a shared score
- 20 min: identify the highest-priority dimension to invest in next quarter
- 10 min: schedule the next assessment (typically quarterly)
The differences between participants' scores are often more valuable than the scores themselves. They surface what each person knows that others don't.
The output: a one-page document with the dimension scores, the composite, the highest-priority next-quarter focus, and the date of the next reassessment.
Common patterns we see
Cross-engagement, the dimension distributions cluster:
Strategy ahead of execution. Many mid-market companies score 3 or 4 on Strategy and 1 or 2 on Platform/Data/Governance. Leadership has named ambition; the ground-level capability hasn't caught up. The corrective action is to slow the strategy expansion until execution catches up; expanding the ambition without the capability produces wishful planning.
Platform ahead of governance. Engineering-led organizations often build platform capability before governance. The dimension imbalance shows up as risk accumulation that gets paid for later. The corrective action is to invest in governance before expanding platform further.
People ahead of everything. Some companies have strong AI talent (often hired during the 2023-2024 enthusiasm) but the talent is under-utilized because Strategy and Data haven't caught up. The talent leaves; the organization regresses. The corrective action is to align ambition and data with the people you have.
Data behind everything. Almost universal. Data work has long lead times; investment in it usually lags. The companies that invest in data early (even before specific AI use cases are clear) have an advantage they don't fully see until they want to deploy AI capabilities.
Governance jump after an incident. Governance progress is often event-driven. A near-miss or a real incident causes a sudden investment in policies and processes. The pattern is human — proactive governance is harder to fund than reactive — but the cost of waiting is real.
Trajectory over time
A typical 12-month trajectory for a mid-market company that engages seriously with AI enablement:
| Quarter | Strategy | People | Data | Platform | Governance | Composite |
|---|---|---|---|---|---|---|
| Q0 (start) | 2 | 2 | 1 | 1 | 1 | 7 |
| Q1 | 3 | 2 | 2 | 2 | 2 | 11 |
| Q2 | 3 | 3 | 2 | 3 | 2 | 13 |
| Q3 | 3 | 3 | 3 | 3 | 3 | 15 |
| Q4 | 4 | 3 | 3 | 3 | 3 | 16 |
This represents disciplined progress. Real organizations sometimes move faster on individual dimensions and slower on others. Backsliding is also possible — Strategy can regress if leadership turns over, Platform can regress if key engineers leave, Governance can regress if the dedicated owner gets reassigned.
The maturity model is most valuable when the assessment is rerun on schedule and the trajectory tracked. A static assessment is a snapshot; the trajectory is the diagnostic for whether the practice is healthy.
Where this model breaks
Honest limitations:
- It's coarse. Five levels per dimension with prose criteria leaves room for interpretation. Different assessors can arrive at different scores. The discussion that produces the converged score is more valuable than the precise number.
- It assumes structured organizational AI work. Companies where AI happens in completely uncoordinated individual contributions don't fit the framework's progression. The Aware level is the floor; below that, the framework doesn't help.
- It does not address industry-specific maturity. Regulated industries (healthcare, finance) have additional dimensions (compliance, audit, ethics) that the model treats lightly.
- The "Optimizing" level is rare at mid-market. Most mid-market companies will never reach Level 5 across all dimensions because the investment level required exceeds what justifies. This is fine; Level 4 ("Capable") is the appropriate target for most mid-market organizations.
Using the model
The model is most useful as a planning anchor:
- Run the assessment with the right participants.
- Identify the dimension that scored lowest.
- Apply the enablement roadmap to that dimension specifically — what stage of the roadmap addresses the gap?
- Run the next quarter's work focused on that dimension.
- Reassess at quarter end.
Most mid-market companies move one composite point per quarter when the practice is healthy. Faster progress is possible (with significant investment); slower progress suggests the practice is stalled and needs diagnosis.
FAQ
Q: What's the difference between this and Gartner's AI maturity model? Gartner's model has more dimensions and is designed for enterprise. Ours is a focused subset for mid-market organizations that don't have the staffing or budget to pursue all the enterprise dimensions. We use Gartner's model for genuine enterprise engagements; the mid-market specialization is the version above.
Q: How often should we run the assessment? Quarterly is standard. Less frequently produces less actionable results; more frequently is overhead without proportional benefit. Six months is a long time for AI capability — much can change in that window.
Q: Who should own the assessment? The named AI lead or AI capability owner. They run the session, capture the results, and reference them in planning conversations. The executive sponsor reviews each quarterly result.
Q: What if leadership disagrees with the assessment? The disagreement is informative. Often leadership underestimates execution gaps because they see strategy clearly. The assessment session is the place to surface and resolve the disagreement; the conversation is more valuable than the score.
Q: Can we use the assessment to compare ourselves to peers? Carefully. Companies don't share these assessments publicly, so peer comparison is mostly anecdotal. Within an industry, the specifics of regulatory environment and competitive dynamics matter more than absolute maturity. We recommend focusing on your own trajectory rather than peer comparison.
Q: At what point does mid-market maturity transition into enterprise maturity? Roughly when the company crosses 500 employees and AI applications cross 10-15 in production. At that point, the dimensions and rubrics from enterprise frameworks (Gartner, Microsoft, etc.) start fitting better than the mid-market specialization.
*For the broader framework this fits into, see our pillar on the AI enablement roadmap for mid-market. For the structures that own AI capability, see our AI Center of Excellence structure piece. For the governance practices that mature alongside, see our LLM governance framework.*
