AI Maturity Model: 5 Stages & the People Layer

Most organisations shopping for an ai maturity model already have one, or three. They score strategy, data infrastructure, technology stack, and governance. They produce a tidy slide for the board. And they leave the hardest variable (whether people can actually use AI) as a single line item labelled "talent" or "culture." That gap is where transformation stalls.

This post lays out a five-stage maturity framework, explains why the people layer keeps getting skipped, and then fills it in with real workforce data from 2,212 assessed professionals. If you are building the case for where your organisation stands on AI, this is the structure and the evidence you need.

What Is an AI Maturity Model?

An ai maturity model is a staged framework that describes how an organisation progresses from ad-hoc AI experimentation to systematic, enterprise-wide AI capability. It gives leadership a common vocabulary, a diagnostic baseline, and a roadmap for investment.

Most models share a similar architecture. They define three to six stages, each characterised by observable behaviours across several dimensions, typically strategy, data, technology, process, and governance. An organisation assesses itself (or is assessed) against each dimension, receives a stage rating, and identifies the gaps between its current state and its target state.

The value is not the label. It is the gap analysis: knowing which dimensions are lagging, which are ahead, and where investment will have the highest return.

How Maturity Models Differ from Readiness Assessments

A readiness assessment asks: "Can we start?" A maturity model asks: "How far along are we, and what does the next stage require?" Readiness is a snapshot before a programme begins. Maturity is a longitudinal measure that tracks progress over quarters and years. Both are useful; they answer different questions. If you need a readiness baseline first, an AI readiness assessment is the right starting point.

Common Dimensions in AI Maturity Frameworks

Regardless of the source, most frameworks evaluate some version of these dimensions:

• Strategy & Vision, Is AI tied to business outcomes, or is it a side project?
• Data & Infrastructure, Are data pipelines production-grade, governed, and accessible?
• Technology & Tooling, Are models deployed, monitored, and maintained?
• Governance & Ethics, Are there policies for risk, bias, privacy, and compliance?
• People & Skills, Do employees know how to work with AI?

The first four get detailed rubrics. The fifth gets a checkbox.

What Are the Five Stages of AI Maturity?

Most ai maturity frameworks converge on five stages, moving from no formal AI activity to AI as a core operating capability. The labels vary; the progression is consistent.

Stage 1: Ad Hoc

AI use is individual and uncoordinated. A few employees experiment with chatbots or code assistants. There is no strategy, no budget line, no governance. Leadership may not know AI is being used at all. Data exists in silos. The organisation is reactive.

Stage 2: Exploring

Leadership acknowledges AI as relevant. Pilot projects appear, usually in IT, data science, or marketing. A small budget is allocated. Data quality conversations begin. There is no enterprise-wide policy, and pilots are disconnected from core business processes. Most organisations that say they are "doing AI" are here.

Stage 3: Formalising

AI moves from pilots to programmes. A centre of excellence or AI steering committee forms. Data governance policies are written and (partially) enforced. Use cases are prioritised by business impact. Training programmes appear, though they are often generic. AI governance frameworks start to take shape. The organisation can point to measurable ROI from at least one AI initiative.

Stage 4: Scaling

AI is embedded in multiple business functions. MLOps and model monitoring are in place. Data infrastructure supports real-time or near-real-time workloads. Governance is operationalised, not just documented. Cross-functional teams include AI-literate members. The organisation can deploy new AI capabilities in weeks, not quarters.

Stage 5: Transforming

AI is a core operating capability, not a technology layer. Business models, products, and decision-making processes are redesigned around AI. The workforce treats AI tools as standard infrastructure, the way they treat email or spreadsheets. Continuous learning systems keep skills current as models and tools evolve. Governance is adaptive, updating as regulations and capabilities change.

Why Do Most AI Maturity Models Skip the People Layer?

Most ai maturity frameworks treat people as a derivative of the other dimensions, the assumption being that if you build the infrastructure and write the policies, the workforce will follow. This assumption is wrong, and the data proves it.

According to Gallup (May 2026), only 25% of US employees say their organisation has communicated a clear plan for integrating AI. The strategy exists on a slide deck. It has not reached the people who need to execute it.

The gap is even sharper on training. BCG's AI at Work 2025 report, surveying over 10,600 workers across 11 countries, found that only 36% of employees feel adequately trained in AI use. And Microsoft and LinkedIn's 2024 Work Trend Index reported that only 39% of people who use AI at work have had AI training from their company.

The Measurement Problem

The reason people get skipped is not indifference, it is measurement difficulty. Strategy can be documented. Data pipelines can be audited. Model performance can be benchmarked. But AI fluency (the ability of a person to communicate with AI tools, evaluate their output, and integrate them into real work) is harder to quantify.

Traditional approaches fall short:

• Self-assessment surveys measure confidence, not competence. People consistently overestimate their own AI skills.
• Multiple-choice tests measure recall, not application. Knowing what a transformer is does not mean you can use one effectively.
• Tool usage metrics (logins, queries, sessions) measure activity, not quality. High usage can coexist with poor practice.

Without a reliable measurement layer, the people dimension stays vague. And vague dimensions do not get funded.

What Gets Measured Gets Managed

The consequence is predictable. Organisations invest in data platforms, model infrastructure, and policy documents, the dimensions they can measure and report on. The people layer gets a line in the L&D budget for a generic "AI awareness" workshop, and leadership assumes the gap is closed. It is not. You can read more about how organisations are approaching this in AI Fluency Measurement Tools [2026].

What Does People-Layer Maturity Look Like When You Measure It?

When you measure the people layer with evidence-linked assessment rather than self-report, the distribution is sobering. Across 2,212 professionals assessed by AISA, the workforce spreads across five AI fluency tiers, and the concentration is heavily weighted toward the lower stages.

The Workforce Distribution

69.3% of assessed professionals fall in the Emerging or Developing tiers. These are people who may use AI tools daily but lack the skills to use them well, to prompt effectively, evaluate output critically, understand model limitations, or apply AI to complex workflows.

Only 7.5% reach Advanced or Expert. This is the population that can lead AI transformation, mentor others, and design AI-native processes.

Where the Gaps Are Deepest

AISA measures five dimensions of AI fluency. The dimension averages across all 2,212 assessments reveal where the workforce is weakest:

Technical Understanding (knowing how models work, what they can and cannot do, and how to select the right tool) is the weakest dimension at 38.9. Safety & Responsibility (covering data privacy, bias awareness, and responsible deployment) sits at 41.1. These are the dimensions that governance frameworks assume the workforce already has.

The implication for maturity models is direct: an organisation can have a Stage 4 data infrastructure and Stage 2 people. The infrastructure does not compensate for the skills gap. It amplifies it, more powerful tools in the hands of undertrained users create more risk, not less.

Mapping Fluency Tiers to Maturity Stages

The five AISA tiers map naturally onto the five maturity stages at the people level:

When 69.3% of your workforce is in the bottom two tiers, your people layer is at Stage 1 or 2, regardless of where your technology and governance sit. That mismatch is the gap most maturity assessments miss.

How Do You Find Out Where Your Workforce Sits?

The first step is measurement that produces evidence, not opinion. You need an assessment that tests what people can actually do with AI, not what they think they can do, and not whether they can pass a multiple-choice quiz about transformer architecture.

AISA's AI readiness assessment measures AI fluency across five dimensions (Prompting & Communication, Critical Thinking, Technical Understanding, Workflow & Application, Safety & Responsibility) using a conversational format. A separate AI evaluator scores each response against a published rubric, requiring evidence in the candidate's own words. Every score is linked to specific statements the candidate made, no black-box ratings.

What a Workforce Baseline Looks Like

A team AI assessment produces a distribution map: how many people sit in each tier, which dimensions are strongest and weakest, and where the gaps are by role, function, or business unit. This is the data you need for board reporting, not "we ran a training programme" but "here is where we are, here is where we need to be, and here is the gap."

For a comparison of available tools, see AI Skills Assessment for Teams: 7 Tools [2026].

From Baseline to Gap Analysis

Once you have the distribution, the gap analysis writes itself:

1. Identify the target tier for each role or function. Not everyone needs to be Expert. A marketing team may need Proficient-level prompting and critical thinking. An engineering team may need Advanced-level task decomposition and technical understanding.
2. Map the current distribution against the target. The delta is your investment case.
3. Prioritise by business impact. Which functions are deploying AI tools today? Where is the risk highest if skills are low? Where is the ROI highest if skills improve?
4. Set a timeline and re-measure. Maturity is longitudinal. A single assessment is a snapshot. Quarterly or semi-annual re-assessment tracks whether investment is working.

How Do You Move the Organisation Up a Stage?

Moving up a maturity stage requires coordinated progress across all dimensions, strategy, data, technology, governance, and people. The people layer is often the binding constraint. You cannot scale AI if the workforce cannot use it. You cannot govern AI if employees do not understand what needs governing.

Close the Awareness Gap First

With only 25% of employees aware of their organisation's AI plan (per Gallup), the first intervention is communication, not training. People need to know what the organisation is doing with AI, why it matters, and what is expected of them. This is a leadership task, not an L&D task.

Build Skills by Dimension, Not by Tool

Tool-specific training ("how to use ChatGPT") becomes obsolete every time a new model ships, and new models ship weekly. The AISA data shows that the weakest dimensions are Technical Understanding (38.9) and Safety & Responsibility (41.1). These are durable skills that transfer across tools and models.

Structure learning around dimensions:

• Prompting & Communication, How to give AI clear, structured instructions and iterate on output
• Critical Thinking, How to evaluate AI output, detect hallucinations, and verify claims
• Technical Understanding, How models work, what they can and cannot do, and how to select the right one
• Workflow & Application, How to integrate AI into real work processes
• Safety & Responsibility, Data privacy, bias, and responsible deployment

Measure, Invest, Re-Measure

The maturity model is not a one-time exercise. It is a management system. The cycle is:

1. Assess, Baseline the workforce with evidence-linked measurement
2. Analyse, Identify gaps by dimension, role, and function
3. Invest, Deploy targeted learning, not generic workshops
4. Re-assess, Measure again after 90–180 days
5. Report, Show the board movement in tier distribution, not just training hours

AISA's results hold up on retake: 98% of people who retake the assessment land in the same or an adjacent tier. That stability means you can trust the delta between assessments as real movement, not measurement noise.

For a structured approach to building the competency framework that sits behind this cycle, see AI Competency Framework for Teams.

Connect People Data to the Rest of the Model

The people layer is not a standalone metric. It connects to every other dimension:

• Strategy without skilled people is a document. Skilled people without strategy is chaos.
• Data governance requires employees who understand data classification and privacy constraints, Safety & Responsibility in AISA's framework.
• Technology scaling requires users who can evaluate model output and integrate tools into workflows, Critical Thinking and Workflow & Application.
• Governance requires people who understand what AI can and cannot do, Technical Understanding.

When you present the maturity model to the board, the people layer should sit alongside the other dimensions with the same rigour: measured, scored, tracked over time. An AI readiness assessment gives you the starting point. The AI Fluency Index gives you external benchmarks to contextualise your organisation's position.

---

Related reading: AI Fluency Measurement Tools [2026], A comparison of enterprise tools for measuring AI capability across teams.

Related reading: How Often Do People Use AI at Work? [2026], Usage frequency data and what it does (and doesn't) tell you about skill.

Related reading: Measure AI Literacy in Employees [2026], Practical approaches to baselining AI literacy at the organisational level.

Frequently Asked Questions

What are the stages of AI maturity?

Most AI maturity frameworks define five stages: Ad Hoc (no formal AI activity), Exploring (pilots and experimentation), Formalising (structured programmes with governance), Scaling (AI embedded across business functions), and Transforming (AI as a core operating capability). Each stage is characterised by observable behaviours across strategy, data, technology, governance, and people dimensions.

How do you assess AI maturity in an organisation?

You assess AI maturity by evaluating each dimension (strategy, data, technology, governance, and people) against defined stage criteria. For the people layer specifically, evidence-linked assessment tools like AISA measure what employees can actually do with AI across dimensions such as prompting, critical thinking, and technical understanding. The combination of infrastructure audits and workforce skill measurement gives a complete picture.

What is the difference between AI maturity and AI readiness?

AI readiness is a pre-programme snapshot: "Can we start deploying AI?" It evaluates prerequisites like data availability, executive sponsorship, and baseline skills. AI maturity is a longitudinal measure: "How far along are we, and what does the next stage require?" Readiness is assessed once before a programme begins. Maturity is tracked over time to measure progress and guide ongoing investment.

How do you measure the people layer of AI maturity?

Self-assessment surveys and tool usage metrics are common but unreliable, they measure confidence and activity, not competence. Evidence-linked assessment, where a separate evaluator scores what a person demonstrates they can do, provides a more accurate baseline. AISA's assessment across 2,212 professionals shows that 69.3% of the workforce falls in the Emerging or Developing tiers, with Technical Understanding (38.9 average) as the weakest dimension.