Why an AI maturity assessment means four different things
Your CEO asks a simple question: "Where do we actually stand on AI?" You pull up an AI maturity assessment, Gartner's or another vendor's, and hand back a stage, a score, a chart. The problem starts right there.
Common practice treats any AI maturity assessment as interchangeable: one clean number to put on a board slide. That gets something right, a single score is easy to report upward.
Where it falls short is this: Gartner, AWS, IBM and PwC each built their own framework, with different pillars, different scoring logic and different inputs. A "level 3" from one model tells you nothing about a "level 3" from another.
Before you run any of them, know what each one actually scores, and what it quietly leaves for you to check on your own.
Gartner's AI maturity assessment: seven pillars, five stages
An AI maturity assessment from Gartner scores organizations across seven pillars: strategy, data, governance, engineering, operating model, culture and AI product or value, according to Gartner's own description of the tool (F1).
The model sorts organizations into one of five maturity stages. The lowest, which Gartner calls Foundational, describes ad hoc experimentation with limited coordination (F2). We think a seven-pillar score is a reasonable way to brief a board, but on its own it says nothing about whether your sales team is actually using the tools you bought them.
Sources: Gartner AI Maturity Model and AI Roadmap Toolkit | Gartner
AI maturity assessment: AWS, IBM and PwC differ
An AI maturity assessment from AWS used to mean the Cloud Adoption Framework for AI, or CAF-AI, which organizes capability across six perspectives: business, people, governance, platform, security and operations. AWS marks that whitepaper for historical reference only, so check which document you're actually using (F3).
AWS's newer generative AI maturity model replaces it in practice. It defines Business as a pillar covering strategic alignment and measurable impact on business goals (F4), and sets four focus areas for prioritizing investment, starting with innovation and feasibility (F5).
- CAF-AI whitepaper (2024): six perspectives, marked for historical reference only.
- Generative AI maturity model: six pillars and four focus areas, AWS's current framework.
An AI maturity assessment from IBM today points back to work IBM says it introduced in 2020: seven dimensions for enterprise AI applications, since aligned to a GenAI adoption model (F6). One dimension, AI governance, IBM defines as the ability to monitor and manage AI activities within an organization (F7).
We think IBM's own framing leans on big claims for what the model does. A member of IBM's AI Applications team wrote on the company's Community blog that this framework has become a standardized methodology, describing it as a measure of how mature AI is in any enterprise application.
“This framework, now a standardized methodology in our organization, is a measure of how mature AI is in any enterprise application.”
— Tanvi Kakodkar, IBM AI Applications team member · Source: AI maturity framework for enterprise applications
An AI maturity assessment from PwC can mean one of two different engagements. The US AI Readiness Assessment scores 12 enterprise domains using senior-leader interviews and PwC's own benchmarking compendium (F8), mapping results to one of five levels from AI-Vulnerable to AI-Driven Leader (F9). PwC Canada separately describes a six-stage model focused on redesigning work, a distinct framework from the US readiness assessment (F10).
We're skeptical of comparing scores across vendors at all: even PwC's two models don't share a scale, so a result from one engagement and a stage number from the other aren't the same currency.
Sources: Foundational AI capabilities, Aspects of generative AI maturity, IBM Architecture Patterns: GenAI, Data & Hybrid Cloud, PwC AI Readiness Assessment for business leaders, Move from AI strategy to results | PwC Canada
Side-by-side: what each model scores and assumes
Laid next to each other, the four frameworks ask different questions for different audiences. The table below applies the same lens to all four: what each scores, what it outputs, and what it leaves unmeasured.
| Framework | What it scores | Output | What it leaves unmeasured |
|---|---|---|---|
| Gartner AI Maturity Assessment | Strategy, data, governance, engineering, operating model, culture, AI product or value | One of five stages, starting with Foundational | Individual employee usage and skills |
| AWS CAF-AI / generative AI maturity model | Business, people, governance, platform, security, operations, plus four investment focus areas | Capability maturity by perspective | Who, inside each perspective, is actually using AI day to day |
| IBM AI maturity framework | Seven enterprise dimensions including governance, aligned to a GenAI adoption model | Dimension-level maturity | Frontline usage and skill gaps |
| PwC AI Readiness Assessment / six-stage model | 12 enterprise domains, or six work-reinvention stages | Level from AI-Vulnerable to AI-Driven Leader, or stage 1 to 6 | Individual employee behavior |
Don't pick a framework because it's the most recognizable name on the board slide. Pick the one whose lens matches your reporting line: Gartner and PwC compress everything into one executive score, while AWS's models organize capability across separate perspectives such as business, people and governance, which can help route ownership of each gap (F3, F4).
Sources: Gartner AI Maturity Model and AI Roadmap Toolkit | Gartner, Foundational AI capabilities, Aspects of generative AI maturity, IBM Architecture Patterns: GenAI, Data & Hybrid Cloud, PwC AI Readiness Assessment for business leaders, Move from AI strategy to results | PwC Canada
Where all four leave a gap
Run one of these and you'll know where your organization stands on strategy, governance and platform readiness. That's a fair win. What you won't get: none of the four frameworks reviewed here publishes a dimension that scores what individual employees actually do with AI day to day.
That's not a flaw in Gartner's, AWS's, IBM's or PwC's work, it's simply outside the job each one was built for. Treating any of them as a workforce diagnostic, rather than a strategy-and-platform diagnostic, is where leaders get the read wrong.
Measure the model, then measure your people
An AI maturity assessment from Gartner, or any of the other three, is still a strategy-and-platform diagnostic. Pair whichever one you run with a separate, recurring measurement of employee-level usage, skills and self-reported barriers to close the gap described above.
AI Adoption Score maps every person, team and department to one of five maturity levels across skills, usage, culture, impact, vision and governance. Its impact dimension reflects what employees report feeling about AI, not a measured productivity outcome, so treat that reading as perception, not proof.
Re-run whichever maturity model you choose on a fixed cadence and watch for movement between stages. A single score is a snapshot, not a verdict.
Frequently asked questions
Can I compare a Gartner maturity score with an IBM one?
Not directly. Gartner scores seven pillars into one of five named stages, while IBM's framework covers seven enterprise dimensions aligned to a separate GenAI adoption model. The scoring logic and stage definitions differ enough that a numeric comparison between them isn't meaningful.
Does AWS's maturity framework still apply, given it's marked historical?
The original CAF-AI whitepaper is marked for historical reference only, so don't treat it as current guidance. AWS's generative AI maturity model, published through its prescriptive-guidance documentation, is the active version, though that page does not state its own publication date.
Which PwC model should I use, the readiness assessment or the six-stage model?
That depends on your question. PwC's AI Readiness Assessment scores 12 enterprise domains for an overall readiness level, while PwC Canada's six-stage model focuses specifically on redesigning work, so pick whichever matches whether you're asking about readiness or about job redesign.
How often should we re-run an AI maturity model?
As a working rule, treat any single score as a snapshot rather than a verdict, and re-run the same framework on a fixed cadence, such as every couple of quarters, so you can track movement between stages instead of reacting to one number. This is editorial guidance, not a documented standard.
Do any of these frameworks measure employee skills directly?
No. As the comparison above shows, each framework scores organizational strategy, governance and platform readiness rather than individual skill or usage data, so treat that gap as the reason to add a separate, recurring employee-level measurement alongside whichever model you choose.
Sources: Gartner AI Maturity Model and AI Roadmap Toolkit | Gartner, IBM Architecture Patterns: GenAI, Data & Hybrid Cloud, Foundational AI capabilities, Aspects of generative AI maturity, PwC AI Readiness Assessment for business leaders, Move from AI strategy to results | PwC Canada
