# AI maturity: 12 readiness questions before you scale AI

> Most AI maturity models score the whole company. These 12 readiness questions, asked workflow by workflow, show what is ready to scale, who owns it and how to measure it.

Language: en · Category: governance · Author: Nicolas Corod (Founder and chief executive officer) · Published: 2026-11-12

## What does AI maturity actually measure?

AI maturity describes how far a company has moved from scattered AI experiments to AI that changes real work, with clear owners, limits and results. A maturity score tells you where the organisation stands on average. It rarely tells you which workflow is ready to change next, and that is the decision leadership has to make.

"Are we ready for AI?" is the wrong question. Many companies asking it already use AI, often without saying so. The sharper question is narrower: which piece of work are we ready to change, who owns it, and how will we know it worked?

The gap between using AI and getting value from it is well documented. In October 2024, [BCG reported that 74% of companies struggle to achieve and scale value from AI](https://www.bcg.com/press/24october2024-ai-adoption-in-2024-74-of-companies-struggle-to-achieve-and-scale-value). Buying more licences does not close that gap. Answering a few precise questions, workflow by workflow, usually does.

## What is the difference between AI readiness and AI maturity?

AI readiness asks whether a company can start: data, skills, budget and sponsorship. AI maturity asks how well AI already runs once it is in use: governance, scale and measured results. Both are usually scored for the whole organisation, which hides the fact that one team can be far ahead of another.

The frameworks that rank today for "ai maturity", from the Gartner AI maturity model to consultancy ladders with five or six stages, share the same structure. They score pillars such as data, technology, talent, governance and strategy, then place the company on a curve. That is useful for a board slide. It is less useful on Monday morning, when a team has to decide what to change first.

| Lens | Question it answers | Unit scored | Typical output |
| --- | --- | --- | --- |
| AI readiness | Can we start? | The whole company | A gap list: data, skills, budget |
| AI maturity | How well does AI run today? | The whole company | A stage on a five or six step curve |
| Workflow readiness | Which work can we change now, and how will we know? | One workflow | An owner, a baseline, limits and a first version |

The third line is the one most maturity frameworks skip. It is also the one that produces a result you can show.

## Which 12 questions show whether you are ready to scale AI?

Twelve questions, asked about one workflow at a time, show more than a company-wide score. They fall into four groups: the work itself, the people who own it, the limits the AI works within, and the value you expect. A workflow that answers all twelve is ready to scale; one that cannot answer the first three is not ready to start.

### The work: where will AI change something?

1. **Which workflow, precisely, are you changing?** "Improve sales with AI" is a wish. "Prepare a sales proposal from a client brief" is a workflow: it has an input, steps and a result someone can check.
2. **What does it cost today?** Time, errors, delays or quality: pick the measure and write down today's figure before anything is built. Without a baseline, every later result is an opinion.
3. **Which steps stay human, and why?** Some steps carry judgement, client relationships or learning that juniors need. Deciding this upfront is part of the design, not a concession.

### The people: who owns the change?

4. **Who knows this workflow best, and will they own the AI role?** The expert who designs a role should stay its owner: they review requests from colleagues and decide what changes. If nobody owns it, it drifts. [Making tacit know-how explicit](/en/blog/knowledge-vs-know-how/) is often the most valuable part of the project.
5. **How is AI already used here, unofficially?** In July 2026, a [Qualinsight study cited by the Swiss business magazine PME](https://www.pme.ch/strategie/ia-au-travail-72-des-romands-lutilisent-souvent-en-secret/rl801nw) found that 72% of working people in French-speaking Switzerland already use AI at work, often in secret. In the same article, Qualinsight founder Esther Sève notes that the gain becomes visible and measurable once the company gives its agreement. These existing uses are a starting point, not a fault. The [shadow AI statistics for 2026](/en/blog/shadow-ai-statistics-2026/) show the same pattern across Europe.
6. **What have you told the team?** Silence is already a message. Say where you are going, what is decided and what is still being tested, even before every answer exists.

### The limits: what is the AI allowed to do?

7. **Which data and tools may the AI reach, and with whose permissions?** The safe default is simple: the AI acts with the permissions of the person asking, and nothing more. In skilder, a role groups a team's skills, its approved tools and its access rights, and every skill runs within the permissions of the user who calls it.
8. **Can you reconstruct what happened afterwards?** If a client, an auditor or a manager asks why a result looks the way it does, someone must be able to answer. An [AI audit trail](/en/blog/ai-audit-trail-execution-logs/) records who asked, which skill ran and which tools it touched.
9. **Which rules apply to you?** For a European company, or a Swiss one whose AI output is used in the EU, the [EU AI Act](https://eur-lex.europa.eu/eli/reg/2024/1689/oj) has required providers and deployers to ensure a sufficient level of AI literacy among their staff since 2 February 2025 (Article 4). Data protection law (the General Data Protection Regulation in the EU, the Federal Act on Data Protection in Switzerland) applies to whatever personal data the workflow touches. An [enterprise AI governance framework](/en/blog/enterprise-ai-governance/) sets these limits once, so each new workflow does not reopen the debate.

### The value: how will you know it worked?

10. **Which indicator did you choose before building?** Time saved is one indicator among several. Depending on the workflow, value also shows in errors avoided, deadlines kept or client satisfaction. Usage counts say little: [valuing what AI skills deliver](/en/blog/measuring-agent-skills-value/) means comparing against the baseline from question 2.
11. **What result would make you stop or change course?** Set the threshold before the first version goes live. A three-week test that shows nothing is a useful result, as long as people can say so without fear.
12. **Who decides what happens to the capacity you free up?** Freed capacity is the human time left once supervision, exceptions and checks are deducted. Leadership decides where it goes: more clients, better quality, training. Leaving that decision open is how gains quietly disappear.

## Why do maturity scores mislead companies that want to scale?

Maturity scores mislead because they average across the organisation and favour foundations over results. A company can score "emerging" overall while one team already runs a reliable AI workflow, and "data first" advice delays the very projects that would show which data matters.

Take a fictitious 200-person logistics company, used here as an illustration only. Its finance team already closes the month with an AI-assisted reconciliation that two controllers review. Its sales team has not yet agreed on a proposal template. A company-wide survey averages the two and lands on "stage 2". Neither team learns anything from that number. Finance needs limits and an audit trail to scale; sales needs an owner and a baseline before anything is built. The twelve questions separate those two situations because they are asked about the work, not about the organisation.

Most maturity frameworks assume a sequence: clean the data, build the platform, train everyone, then deploy. Each step is reasonable. Together they can push the first visible result out by many months, and motivation fades before anything changes in daily work.

The evidence points to a different order. Dalith Steiger-Gablinger, co-founder and managing partner of Swisscognitive, put it plainly in [PME in February 2026](https://www.pme.ch/strategie/pourquoi-lia-ne-booste-pas-encore-la-productivite-suisse/q9wp7q7) (quoted in French, translated here): "AI cannot repair broken processes, it only makes them visible." That is an argument for starting with one well-understood workflow, not for waiting until every process is fixed.

Durability matters too. A [Gartner survey published in June 2025](https://www.gartner.com/en/newsroom/press-releases/2025-06-30-gartner-survey-finds-forty-five-percent-of-organizations-with-high-artificial-intelligence-maturity-keep-artificial-intelligence-projects-operational-for-at-least-three-years) found that 45% of organisations with high AI maturity keep their AI projects operational for at least three years. Keeping a project alive that long takes an owner, a budget someone defends and a result someone can show, which is exactly what questions 4, 10 and 11 above test.

## Does a company need to be ready before it starts with AI?

No. A company does not need to be ready to start; it becomes ready by moving forward. Start where AI can create value and where a team wants to change how it works. The results give a concrete reason to tackle the less attractive work: data quality, systems and documentation.

This does not mean skipping the foundations. Data still needs preparing, but for a specific workflow, not in the abstract. Limits still need setting, in the first weeks, before the first version goes live. What changes is the order: one workflow, one owner, one baseline, then the next one.

In practice, the first loop looks like this:

1. Pick one workflow that answers questions 1 to 3 clearly.
2. Name its owner and set its limits (questions 4, 7 and 9).
3. Record the baseline and the stop threshold (questions 2 and 11).
4. Ship a first version, measure it against the baseline, then decide: improve, extend to another team or stop.

Each loop raises the company's maturity where it counts, in real work. If you want to test these twelve questions on one of your own workflows, the [free 30-minute exchange with an expert](https://koalendar.com/e/meet-with-nicolas-9?utm_source=skilder.ai&utm_medium=blog) is built for that: together, you identify the first workflow to change and how to measure it.

**Takeaway:** Stop asking whether the company is ready for AI, and start asking which workflow is ready to change, who owns it and how you will know it worked.

[Book a free 30-minute exchange with an expert](https://koalendar.com/e/meet-with-nicolas-9?utm_source=skilder.ai&utm_medium=blog) to apply the twelve questions to your first workflow.
