Free resource · Word grid

AI Maturity Grid: Assess your organization's readiness in 90 minutes

Six dimensions, four levels, observable descriptors and evidence to request, for a management workshop that leads to an action plan.

📄 Word, 26 pages🧭 6 dimensions, 4 levels⏱️ 90-minute workshop🆓 Free

Download the gridSee the six dimensions

“Where do we really stand with AI?” The question comes up in every executive committee meeting, and it almost always receives the same answer: impressions. The sales director thinks things are progressing, the IT department reports personal accounts everywhere, HR hasn't trained anyone, and the legal department is waiting for a charter. Everyone is right within their own area of ​​expertise, and no one knows what the organization should do first.

A maturity grid addresses this issue differently than simply adding up opinions. It breaks down adoption into six observable dimensions, proposes four levels per dimension with verifiable descriptors, and reveals what an average always masks: imbalance. This page explains how to read it, how to conduct the session that uses it, and provides the complete grid in Word format.

1. Why a grid, and not a score?

A maturity score out of 100 is reassuring but doesn't commit you to anything. It produces a statement ("we're at 47") that no one knows what to do with the next day. A six-dimensional profile forces you to look at where the organization is strong, where it's fragile, and what drives its pace.

Most of the organizations we work with don't have a problem of average performance. They have a problem of balance: advanced practices in a profession with no rules, or an impeccable charter that no one follows because no one has been trained. Average performance masks these discrepancies, and it is precisely in these discrepancies that the two key areas of focus lie.

The grid is based on three principles, and it is their strict application that makes the finding credible.

We note what is observable. Each box describes observable facts: the percentage of employees using AI each week, whether or not a charter has been published, and whether or not an agent is piloting the program. Neither intentions, nor ongoing projects, nor the quality of the discourse matter. An organization that is "about to launch" a program is simply demonstrating what it has already accomplished.

The level reached is the one in which all the descriptors are true. To be declared at level 3 in a given dimension, all the descriptors for level 3 must be met, not just the most favorable one. If even one is missing, the organization is at level 2, but with a solid foundation already in place for the next steps.

The smallest dimension takes control. An organization progresses at the pace of its core competency. Level 4 practices with Level 1 governance do not constitute a Level 4 organization; they represent an exposed organization. The first project to decide on is almost always the one that raises the core competency.

2. The six dimensions of AI maturity

Six dimensions are sufficient to describe the adoption of AI in an organization. They are independent of each other in the sense that one can be advanced in one and a beginner in another, and this is precisely what makes their cross-referencing useful.

ESG/SRI	What it measures	Key question
A. Adoption and skills	The proportion of employees who use AI each week, how they are trained, the ongoing support, the position of middle management.	How many people practice, and who helps them progress?
B. Business uses and impact	Adapting practices to the realities of each profession, prioritizing cases, measuring impact, and the number of cases used routinely.	Does AI produce identifiable value, profession by profession?
C. Agents and processes	Maturity regarding assistants and agents, mapping of delegable work, redesign of end-to-end processes.	Do we delegate complete work, and under what conditions?
D. Data and tools	The security and suitability of the tools provided, the preparation of data and knowledge, the ability to integrate into the information system.	Do our technical foundations allow us to scale up?
E. Governance and risks	The charter of use, compliance and security, monitoring of results, the framework applicable to autonomous agents, visibility on unregulated uses.	Do we know what is permitted, verified and tracked?
F. Strategy and management	The role of AI at the management level, program governance and responsibilities, and the evolution of the operational model.	Does the direction carry a trajectory, and does it steer it?

In the toolkit, each dimension occupies a page: the descriptors for the four levels, the list of evidence to request before rating (the measured, not estimated, usage rate; the date of the last charter update; the minutes of the last arbitration hearing), and a verification question to ask when the group is unsure between two levels. For governance, for example: "If an agent sent incorrect information to a client tomorrow, would we know within an hour what they did, why, and who was responsible for verifying it?"

3. The four levels

The same four levels are used for each dimension and for the overall profile. They correspond to the four stages of the adoption trajectory: embark, master, delegate, transform.

Level	Last Name	In one sentence	Trajectory time
1	Explorateur	The potential is open, the uses are isolated, nothing is yet organized.	Embark
2	Pioneer	Initiatives exist, led by volunteers; they are not yet connected.	Control
3	Accelerator	A real dynamic has been launched, structured on part of the perimeter; it needs to be industrialized.	To delegate
4	Transformer	The foundations are solid and well-managed; the challenge now lies in reinventing the processes, offerings, and model.	Transform

The rating system adds up the six levels, resulting in a total between 6 and 24: Explorer (6-9), Pioneer (10-14), Accelerator (15-19), and Transformer (20-24). However, the total is the least interesting of the three readings. The other two are the floor plan dimension, which designates the initial project, and the difference between the highest and lowest-rated dimensions: beyond two levels, the organization is unbalanced, and this imbalance must be addressed before expanding anything.

4. The 90-minute workshop

A completed grid, filled out individually, provides an opinion. When completed collaboratively, with evidence presented, it yields a conclusion. This is the difference between just another document and a decision made by the board of directors.

Six to eight people are sufficient, and their composition is not negotiable: the sponsor on the management committee, without whom the session produces a diagnosis that no one will arbitrate; the HR director or the training manager, who knows the real state of skills; a representative from the IT department, who knows what is actually deployed by difference with what is announced; one or two business managers, who bring the real use cases or their absence; the legal or compliance officer; and a facilitator separate from the sponsor.

Each participant individually rates the organization one week in advance on the sheet in Appendix A. The facilitator collects the ratings the day before and prepares the discrepancy table. This is often the most useful information from the session: when the IT department rates the tools 1 and the business unit rates them 3, the ensuing discussion is worth more than the final rating.

Temps	Sequence	What is happening
0 to 10 min	Feasibility study	The sponsor explains why this assessment is necessary now and what they will do with it. The presenter states the three rules and displays the discrepancy chart.
10 to 60 min	The six dimensions	Eight minutes per dimension: the gap is announced, those who noted the highest and lowest explain themselves with supporting evidence, the group checks the descriptors one by one, the sponsor decides if the disagreement persists.
60 to 70 min	Profile reading	Total, overall profile, floor dimension, support point, radar. The group compares with typical profiles.
70 to 85 min	Two construction sites	The first raises the floor. The second is the nearest level crossing. For each: a person in charge, a deadline, an observable sign of success.
85 to 90 min	Fencing	The sponsor is reformulating both decisions. The review date is set at 90 days and the next rating at six months later.

Three rules govern the session. The organization is assessed as it is, not as desired: a budget that has been approved or a tool that has been ordered does not count. A grade must be justified by evidence, such as a figure, a document, or a dated example. And in case of persistent disagreement, the lower grade prevails, unless evidence presented during the session proves otherwise: an organization that overestimates itself makes poor decisions, while an organization that underestimates itself by one notch loses little.

5. Six typical profiles, and what they reveal

Throughout our work with clients, the same radar charts keep recurring. Each one represents a specific risk and requires a precise interpretation. Recognizing your own chart can save an entire meeting.

Profile	What this reveals	The risk	Reading
The archipelago	High-performing units without a common direction: each profession cobbled together its own solution.	Tool dispersion, duplication, inability to scale.	Connect before expanding: a sponsor, a program manager, and the library of existing uses as the first deliverable.
The fortress	The framework is written, the charter is distributed, and almost no one practices it.	AI is being developed outside of these platforms, on personal accounts, completely uncontrolled.	Open by training: a pilot population trained and equipped in 90 days, with the charter taught rather than displayed.
The flash in the pan	Many users, little identifiable value; uses remain generic.	A loss of momentum after six months, due to a lack of results to show.	Specialize by profession and measure: three documented cases in two professions, one indicator each.
The laboratory	Advanced agents on fragile foundations: poorly governed data or an absent framework for autonomy.	The incident that discredits the entire program.	Consolidate before connecting: autonomy framework, logs, incident procedure, reliability of priority sources.
The declaration	The ambition is there and budgeted, but execution has not yet begun.	Loss of credibility for management, who made an announcement without delivering.	A pilot project visible in 90 days: a population, a profession, a use case, a figure.
The balance	The organization progresses in a consistent manner, regardless of its level.	Slowness: without a marked point of support, no lever pulls the others.	Choose a lever: focus the 90 days on the dimension closest to the next level.

Section 6 of the kit complements this reading with the table of action items: for each of the six dimensions and each of the three possible transitions, the action item that triggers the threshold crossing and the observable signal that proves it has been crossed. Moving from level 2 to 3 on adoption, for example, means building a structured pathway with practical application for a target population and naming a network of relays; the signal is a weekly usage rate of 30% and a list of relays by name.

6. What the kit contains

The grid is a 26-page, editable Word document designed to be printed and annotated during the session. It contains, in order: instructions for use and three ways to use it; the three scoring principles, the six dimensions, and the four levels; the grid itself, one page per dimension, with observable descriptors, evidence to request, and the verification question; the scoring sheet with the profile calculation, floor and support point identification, and a blank radar chart to draw; the timed schedule for the 90-minute workshop, its three rules, and its five common pitfalls; the six typical profiles and the rule for choosing the two construction sites; and the table of construction sites by dimension and transition. It also includes six appendices: the individual scoring sheet to be completed before the session, the workshop report template, the six-monthly monitoring grid for four scoring methods, the correlation with our online diagnostic tool, frequently asked questions, and resources for further exploration.

The kit, part by part: what each part contains and what it is used for. The templates to fill in are in the Word document.

Part 1. Why a grid, and not a score?

This section lays out the three notation principles (we note what is observable, the level reached is that in which all descriptors are true, the smallest dimension controls), the six dimensions and the four levels: Explorer, Pioneer, Accelerator and Transformer.

Part 2. The grid

One page per dimension: the observable descriptors for each level, all of which must be true, the evidence to request before scoring, and a verification question if the group hesitates between two levels. Each level builds upon the achievements of the previous levels.

A. Adoption and skills

This dimension measures the actual proportion of employees using generative AI in their work, their skills development, ongoing engagement, and the position of middle management. At level 1, less than 10% of employees use it weekly and no training is provided; at level 4, more than 60% use it weekly, training paths are tailored to each level and role, and managers set expectations for their teams.

B. Business uses and impact

This dimension measures the adaptation of uses to the realities of each profession, the prioritization of use cases, the measurement of their impact, and the number of cases actually used routinely. At level 1, uses remain generic (writing, summarizing, translation) and no use cases are routine; at level 4, a living library of use cases per profession is maintained, and more than ten use cases are routine in several professions, with monitoring of time saved, quality, and return on investment.

C. Agents and processes

This dimension measures the maturity of AI assistants and agents, the existence of a map of delegable work, and the redesign of end-to-end processes. At level 1, the subject remains unclear, and AI is added to the existing system; at level 3, agents are being piloted in real-world business cases; at level 4, they are deployed, monitored, and connected to the organization's tools, and human-agent workflows are operated and measured.

D. Data and tools

This dimension measures the availability of secure and appropriate AI tools, the preparation of data and knowledge to support assistants and agents, and the ability to integrate with the information system. At level 1, personal accounts predominate; at level 4, business-specific tools are deployed at scale, data is governed, and a target architecture is in place. Verification question: Does a new employee have access to a secure AI tool on their first day?

E. Governance and risks

This dimension measures the usage charter, compliance, security, results monitoring, the framework applicable to autonomous agents, and visibility into the use of public-facing tools. At level 1, no usage rules are formalized; at level 3, a charter is written and disseminated; at level 4, the charter is dynamic and an operational framework covers the GDPR, the AI Act, cybersecurity, traceability, and human oversight.

F. Strategy and management

This dimension measures the role of AI at the management level, the existence of program governance with identified managers, and the evolution of the operating model. At level 1, the topic has little presence at the management level; at level 3, an ambition is defined with a budget; at level 4, the vision is driven by management with a managed roadmap. Verification question: "What AI arbitration decision has management made in the last three months, and where is it documented?"

Part 3. The Grading Sheet

Step 1: Note each dimension

For each dimension, the selected level from 1 to 4, the main proof, and the first missing descriptor for the next level are entered. This missing descriptor is the nearest checkbox: this is often where the most profitable project lies.

Step 2: Calculate the overall profile

The six levels add up to a total of between 6 and 24. From 6 to 9, the Explorer profile calls for securing the first uses; from 10 to 14, the Pioneer must link its initiatives in a coherent program; from 15 to 19, the Accelerator must industrialize; from 20 to 24, the Transformer reinvents processes, offers and operating model.

Step 3: Identify the floor and the support point

The floor dimension has the lowest score; if both are tied, the dimension that most exposes the organization is chosen: Governance and Risks, then Data and Tools, then Strategy and Management. The support point is the highest-scoring dimension, which is used as leverage. A difference of 2 or more between the two indicates an imbalance that must be addressed before scaling anything up.

Step 4: Draw the radar

Each level is plotted on its axis, then the six points are connected: a small, regular hexagon, an unbalanced star, or a large, regular shape. The drawing is kept for comparison in the following semester.

Part 4. The 90-minute workshop

Who to bring together

Six to eight people: the sponsor (a member of management), the training manager or HR director, a representative from the IT department, one or two business managers, the legal or compliance officer, and a facilitator separate from the sponsor. Beyond that, the session devolves into a presentation; below that, there's a lack of evidence.

Before the session

Each participant receives the grid and individual scoring sheet one week in advance, completes it independently in 20 minutes, and provides evidence for their scope. The facilitator collects the sheets the day before and prepares the variance table: for each dimension, the lowest, highest, and median score.

The usual pitfalls

The kit describes five pitfalls and their countermeasures: the pioneer's rating, where the most advanced profession rates the organization at its own level; the prescriber's rating, hence the rule of rating the sponsor last; the debate over words, settled by the verification question; the shopping list, limited to two projects; the flat radar, because level 2 also has to be proven.

Part 5. Read the profile

The six common profiles

Each profile is identified by the shape of the radar chart: the archipelago (Uses of 3 or more, Strategy of 2 or less), the fortress (Governance of 3 or more, Adoption of 2 or less), the flash in the pan (Adoption of 3 or more, Uses of 2 or less), the laboratory (Agents of 3 or more, Data or Governance of 2 or less), the declaration (Strategy of 3 or more, everything else of 2 or less), and the balanced profile (difference of 1 or less between all dimensions). For each profile, the toolkit provides the risk assessment and how to interpret it.

Choose the two construction sites

The first project raises the floor dimension by one level, no more. The second is the closest transition: the dimension where only one descriptor is missing to change levels. A well-formulated project can be summed up in one line: a verb, a scope, a responsible party, a deadline, an observable signal.

Part 6. Moving on to the next level

For each dimension, there are three transitions, with the construction site marking the crossing of the threshold and the observable signal proving that it has been crossed. The grid does not reward jumps of two levels.

From 1 to 2: Boarding

Raise awareness widely and recruit volunteers (signal: 10% of employees use AI every week), identify three quick value cases in a volunteer profession, provide a secure tool to a pilot population, disseminate minimum safety guidelines and appoint a sponsor who presents a statement of intent to management.

From 2 to 3: Mastering

Build a structured path with practical application and name a relay network (signal: 30% weekly usage), document the use cases with an indicator for each, launch an agent pilot with a named supervisor, draft the charter with the legal, IT and business departments, and formalize the ambition with a budget and bodies.

From 3 to 4: Delegate and transform

Break down the paths by level and by profession (signal: 60% weekly usage), manage a portfolio of more than ten cases routinely, operate measured human-agent workflows, equip a new arrival from the first day, handle a simulated incident in less than an hour and obtain from management a written and followed arbitration decision per quarter.

Annexes

Six annexes: the individual rating sheet, the workshop report template, the semester monitoring grid, the correspondence with the online diagnosis, the frequently asked questions and the resources to go further.

Appendix B. Workshop report template

A page sent to participants within 48 hours, which serves as a reference for the 90-day review: rating retained with its proof and the gap between participants, overall profile, floor, support point, then the two projects with responsible party, deadline, signal of success and first step within 15 days.

Appendix C. Semester Monitoring Grid

The same assessment is repeated every six months by the same group and is read dimension by dimension. Three questions are asked for each assessment: Has the floor level risen by one level? Has the gap between the support point and the floor level decreased? Has a dimension decreased? A level is generally gained in two to four semesters on a given dimension.

Appendix D. Correspondence with the online diagnosis

The online diagnostic NextStart.AI asks 22 questions on the same six pillars and gives a score out of 100. Each question offers four answers which correspond to the four levels of the grid, and a dimension noted 3 in the grid corresponds to a pillar between 60 and 80 out of 100.

7. Frequently Asked Questions

Should the entire organization or just one entity be rated?

The scope corresponds to the decision-making authority of the sponsor present. A business unit can be evaluated independently if it decides on its own tools, training, and rules. Otherwise, the organization that provides these three elements is evaluated. The selected scope is recorded in the report; otherwise, comparisons to the following semester are meaningless.

How does this differ from your online maturity assessment?

Le online diagnosis The first method involves asking 22 questions in six minutes and providing a score out of 100 with a radar chart: this is an individual snapshot, useful for preparing the session. The second method provides a collective, proven assessment in 90 minutes. Both are based on the same six pillars, and Appendix D of the kit provides the exact correlation between the grid levels and the diagnostic scores.

How much time should I allow?

Twenty minutes for an individual speed-reading assessment, 90 minutes for the management workshop which produces a shared assessment and two workstreams, and 45 minutes for the follow-up assessment every six months. Preparation requires a week's lead time, allowing each participant time to complete their individual assessments and gather their evidence.

Does the grid apply to an SME?

Yes, with the same descriptors. The evidence is easier to gather, and the workshop can be held with four people. Level 4 on the Strategy and Management dimension is even more accessible than in a large group, as management is closer to the field.

At what rate are we progressing?

In the organizations we support, progress is generally achieved in two to four semesters on a given dimension, more quickly in governance (primarily through writing and dissemination), and more slowly in adoption and staffing (through work on practices). Progressing one level per year on two dimensions is a good pace.

What role does the AI Act play in the rating?

It appears in the Governance and Risk dimension, at level 4, within the expected operational framework. The obligation to control AI, as stipulated in Article 4 of the European regulation and now a best-efforts obligation under the omnibus regulation (EU) 2026/1744, applies to any organization deploying AI systems for its employees; the training program and the charter are the two areas where this is implemented concretely. The framework is a guidance tool and does not constitute legal advice.

8. Going further

The grid produces an observation and two action items; it doesn't manage them for you. NextStart.AI intervenes in the two natural next steps: the diagnosis Choose your use cases, when the assessment needs to be further investigated department by department with the various business units, the IT department, HR and the legal department, and the AI program frameworkWhen construction projects become a trajectory to be budgeted. Depending on the dimension that the grid designates as the floor, the formats differ: acculturation of the management committee Regarding strategy, Prompt Party et AI ambassadors for adoption, Agent Party (1 hour or half a day) for agents, charter of practices for governance.

Read also : AI Adoption Roadmap Kit in 12 Months, AI Charter in Business: The 2026 Model (Downloadable), AI Ambassador Network: The 2026 Toolkit, Managing a team in the face of AI: the 2026 toolkit et AI systems governance: the 2026 toolkit.

The entire series of free kits. To drive adoption: AI charter template, AI Ambassadors Kit, Manager's toolkit for dealing with AI, Adoption Roadmap Kit, AI Measurement and ROI Kit et AI Governance KitFor the AI ​​project sponsor: AI Act kit, GDPR kit, Security and data kit, AI and Social Dialogue Kit, frugal AI kit, AI Sovereignty Kit et responsible AI kitBy profession: AI HR kit, AI marketing kit, AI finance kit, AI kit for lawyers, AI kit for salespeople, AI toolkit for project managers, AI innovator kit et communicator's AI kit.

9. sources

The grid is based on the NextStart Method. The regulatory references cited on this page refer to the following sources.

• European Union, Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Article 4, AI mastery).
• European Union, Regulation (EU) 2026/1744 (omnibus IA), adopted on 8 July and published on 24 July 2026amending in particular Article 4 of Regulation (EU) 2024/1689, in force since 27 July 2026.
• European Commission, AI Literacy – Questions & AnswersUpdated July 27, 2026.