From level 2 to level 3: the most difficult leap in AI maturity

The transition from being individually productive to becoming mature at the team and process level is where 80% of SME organizations get stuck. Not because of a lack of technology or budget, but because of three organizational obstacles that are consistently underestimated. This article explains why that leap is so difficult, what structurally needs to change, and which three concrete interventions make the biggest difference.

Why level 2 is so persistent

In the previous two blogs, we described level 2 as the phase where everyone uses AI but no one is directing the organization. We also introduced a self-assessment to help determine whether you are in the middle of it.

What we have not answered yet is this: why do so many organizations remain stuck at level 2, even when the director knows there is more potential?

The answer is not technology. The tools exist, the licenses are paid for, the training has been completed. The answer lies in how organizations change, and how difficult it is to move from individual ways of working to shared ways of working without it feeling like bureaucracy.

Level 2 is comfortable. People work in their own way, with their own tools, and achieve individual results. There is no friction, no discussion, no barrier. Level 3 breaks that pattern. It requires people to share their best approaches, someone to take ownership of quality, and agreements that everyone respects even when they are used to working differently.

That is not a technical change. It is a social and organizational change. And those are always harder than rolling out a new tool.

What structurally changes in the transition to level 3

The essence of level 3 is not that AI is used better. The essence is that AI usage becomes predictable. Repeatable. Independent of who happens to be available that day.

That requires three structural shifts.

Shift 1: From personal knowledge to shared knowledge

At level 2, the best AI knowledge is locked inside individuals. The employee with the best prompts, the fastest workflow, or the most consistent output may share things informally, but nothing is secured. If that person leaves, is absent, or changes roles, the knowledge disappears.

At level 3, that knowledge belongs to the organization. There are shared prompt libraries, documented workflows for recurring tasks, and moments, formal or informal, where teams learn from what works. That does not require an extensive knowledge management system. Sometimes it is simply a shared Teams folder, a monthly 20-minute lunch session, or a simple Notion page.

It is not about the system. It is about the intention that knowledge becomes shared property.

Shift 2: From no direction to lightweight governance

“Governance” is a word that often creates resistance among SME directors. It sounds like IT policies, compliance documents, and slow decision-making. At level 3, it is much simpler: three to five clear agreements that everyone understands and respects.

Which tools are the standard within our organization? Which data should never enter public AI models? Who is responsible for AI quality in each team? How do we verify AI-generated output before it is shared externally?

These are not bureaucratic questions. They are basic agreements that prevent the three risks from week 18 — fragmentation, the productivity trap, and knowledge leakage — from becoming real problems.

Gartner’s research on AI adoption shows that organizations with lightweight, pragmatic governance consistently outperform organizations with no governance, but also those with governance that is too heavy. The sweet spot is simplicity and clarity, not completeness.

Shift 3: From individual success to collective learning

At level 2, AI success is rewarded individually and ignored collectively. John gets praised for his fast proposals, but nobody asks how he does it. Petra found a smart way to summarize client reports, but only her direct colleague knows about it.

At level 3, there is a mechanism, however simple, that converts individual success into collective knowledge. A weekly AI tip in the team chat. A moment during the monthly meeting where someone shares what worked. A short internal session where a department demonstrates what they have learned.

Harvard Business Review describes this as the most underused lever in AI adoption: transforming scattered individual learning experiences into shared organizational intelligence. Not through major programs, but through small, consistent habits.

The three obstacles preventing level 3

Knowing what needs to change is one thing. Understanding why it does not happen naturally is another. Three obstacles consistently appear.

Obstacle 1: No one feels responsible for AI quality

In most organizations, AI is everyone’s responsibility and therefore no one’s responsibility. There is no person ensuring that successful approaches are documented. No person noticing when quality declines. No person onboarding new employees into the team’s AI workflows.

Without explicit ownership, this function simply does not exist. And without that function, level 3 does not grow.

The solution is not to create a new position or hire an AI manager. The solution is to give one existing employee per team an explicit responsibility: you are responsible for how our team uses AI. That costs two hours per month and creates continuous organizational learning.

Obstacle 2: There is no sense of urgency

Level 2 works. Not optimally, but well enough for today. There is no burning urgency to move forward unless the director deliberately introduces the long-term perspective.

Organizations that stay at level 2 do not do so because they are lazy or uninterested. They do it because other priorities demand attention every day. AI maturity is important, but not urgent — until the moment a competitor who did continue growing wins a tender you can no longer keep up with.

This is a classic strategic problem: the silent cost of standing still remains invisible until it becomes unavoidable. A director who wants to reach level 3 must create that urgency by positioning it as a strategic priority, not as an IT project.

Obstacle 3: The first step is too large

Many organizations that consciously want to move toward level 3 begin with interventions that are too large. A broad AI policy rollout. Organization-wide training. A new knowledge-sharing platform. All projects that take months, create resistance, and are difficult to sustain.

The most effective approach is the opposite: start small, start fast, start concrete. Choose one team. Make three agreements. Appoint one owner. Evaluate after six weeks. Then scale.

MIT Sloan describes this as the “beachhead” approach to AI adoption: create one successful internal example, use it as proof and inspiration, and allow growth from there organically. Not through top-down rollout, but through bottom-up evidence.

The three interventions that make the biggest difference

Based on everything above, there are three concrete interventions that create the transition from level 2 to level 3 the fastest and most sustainably.

Intervention 1: Appoint one AI owner per team

Not an AI manager. Not a new position. Simply one existing employee per team who becomes explicitly responsible for three things: documenting successful approaches, onboarding new employees, and reporting monthly on what the team learns.

Choose someone who is already enthusiastic about AI, but do not choose the most technical person. Choose the person who communicates and collaborates best. AI ownership is a social role, not a technical one.

Intervention 2: Make three agreements, not thirty

Do not write an extensive AI policy. Make three agreements that everyone can remember and follow.

An example:

• We use [tool X] as the standard for [task type]
• Client data and internal financial information never enter public AI models
• AI-generated output is always reviewed by a human before external use

Three sentences. No document. No procedure. Just clarity on the basics.

Intervention 3: Create one weekly learning moment

Ten minutes during the team meeting. One person shares something that worked that week: a prompt, an approach, a mistake and how it was solved. Not as a formal presentation, but simply as part of the meeting.

This is the simplest and most powerful way to turn individual learning experiences into collective knowledge. After three months, more shared AI intelligence will have been built than through almost any training program.

Frequently asked questions about the transition to level 3

How long does the transition from level 2 to level 3 take?
For a team of 10 to 20 people: 8 to 12 weeks with consistent execution. For an organization of 50 to 200 people: 3 to 6 months, depending on the number of teams and the level of resistance.

Should the director personally lead this process?
Yes, but not as the executor. The director must position it as a strategic priority and create room for the owners to fulfill their role. Without leadership support, level 3 remains a hobby for enthusiasts.

What if there is resistance within the team?
Resistance during the transition to level 3 is almost always caused by uncertainty, not unwillingness. People fear losing their way of working, being judged on AI usage, or receiving extra work. Be explicit about what changes and what does not. Level 3 does not remove autonomy; it adds structure to existing autonomy.

Should all teams move at the same time?
No, and that is not advisable. Start with the team already using AI the most. Let that team create the first agreements, appoint the first owner, and establish the first learning moments. Use that team as internal proof. Other teams will follow naturally when they see results.

How do I know whether the transition succeeded?
Two concrete signals: a random new employee can work at the same quality level within two weeks, and if your best AI user leaves tomorrow, the knowledge does not disappear with them.

The honest expectation

The transition from level 2 to level 3 is the most difficult step in AI maturity. Not the most technically complex, but the most organizationally demanding.

It requires something from you as a director: the willingness to treat AI maturity as a strategic priority, not as an IT project. It requires something from your team: the willingness to share individual ways of working and accept shared responsibility.

Organizations that make this transition build something that cannot be caught up with through a few training days: collective AI intelligence. Knowledge that exists inside the organization, not inside individuals.

Next week we conclude the month with the end-state vision. What does a level 4 organization look like in practice? Which decisions are made differently? And what does that concretely mean for you as a director?

Do you want to make the transition from level 2 to level 3 with guidance? At LinkLeaders, we help directors and their teams through exactly this transition. No large-scale programs, just concrete steps that match the scale and pace of your organization.

Call: +31 (0)182 570927 | Email: contact@linkleaders.nl