Key Steps for Change Management with AI: A Practical Guide
The key steps for change management are: diagnose organizational readiness, build a leadership coalition, communicate meaning before tools, train in successive waves, then embed new behaviors into processes. In an AI project, each step is amplified. Resistance is stronger. Trust issues are more visible. And mistakes cost more.
The Problem Nobody Says Out Loud
You’ve approved an AI project. The budget is there. The vendor is selected. And yet, six months later, the tool is underused, teams are working around the system, and your HR director reports that “people aren’t playing along” — that’s their wording, not mine.
This isn’t a technology problem. It’s a change management problem.
In Morocco, Jamila Boussaâ signals in Medias24 that AI adoption in companies remains uneven, even as momentum builds. What I observe with my clients confirms this: the organizations that move forward aren’t those with the best tool. They’re the ones that prepared their teams before deploying.
Here’s how to do it.
Step 1: Diagnose Before You Decide
Before talking about AI, ask your managers one simple question: “What takes up most of your time each week that shouldn’t?”
The answers give you your real use cases. Not the ones your vendor sold you. The ones your teams actually live with.
Then assess three dimensions: your organization’s data culture, the AI literacy level of your managers, and the quality of your existing AI governance. If all three are weak, start small. One process. One department. One visible win.
What I observe with my clients: projects that fail are almost always those where the diagnostic was skipped in favor of an impressive demo.
Step 2: Build the Coalition Before You Launch
You need a coalition of leaders who believe in the project before it begins. This is the founding principle of the major change management models, Kotter and ADKAR alike. It still holds in 2026.
In an AI context, this coalition has one specific requirement: it must include profiles who aren’t naturally tech-enthusiastic. A converted skeptical HR director is worth ten times more than an enthusiastic CTO. Because the skeptical HR director is who your teams find credible.
Identify two or three influential people in each affected department. Not the most enthusiastic. The most respected within their teams. Involve them in the design phase, not just the deployment phase.
This is exactly what I cover in my 2-to-3-week AI Governance Sprint, which starts from this human mapping before any technical decision. Learn more about my services.
Step 3: Communicate Meaning, Not Features
Your teams don’t want to know what the tool does. They want to know what it changes for them.
Is my job at risk? Will I have to learn something difficult? Will my value in the organization decrease?
These questions aren’t asked in meetings. They circulate in hallways. If you don’t address them explicitly, they become rumors.
Change communication in an AI project must answer three things: why now, what isn’t changing, and how the organization supports those who need more time.
As I explained in my analysis on which jobs will survive AI, the fear isn’t irrational. It’s legitimate. Your role as a leader is to name it, not minimize it.
Step 4: Train in Waves, Not in Mass Sessions
The classic reflex: organize one big training session for everyone on the same day. Result: everyone saw the slides, nobody changed their habits.
AI skill-building works differently. It requires repetition, practice on real cases, and time to integrate.
Structure training in three levels. First, AI literacy for all managers: understanding what AI can and cannot do, without going into technical details. Then, tool proficiency for the operational teams directly concerned. Finally, deep expertise for internal champions who will carry the project over time.
To calibrate that third level, my article on the 5 most used AI tools in business in 2026 gives you a useful baseline.
In Morocco, companies are facing a real shortage of AI experts, as SNRTnews signals. This means you won’t easily find external profiles to train your teams. Invest in two or three internal champions trained in depth. It’s more sustainable than a one-off external engagement.
Step 5: Embed in Processes, Not in Intentions
Change isn’t real until it’s written into daily processes.
Concretely: if your HR team uses an AI tool to pre-screen candidates, this isn’t a pilot project anymore. It’s the new recruitment process — as an example of embedding, not a universal prescription. It must be documented, evaluated, and integrated into the team’s performance indicators.
If the tool remains optional, it will be abandoned the moment daily pressure returns. And it always returns.
Embedding also requires AI governance: who validates decisions made with AI assistance, who is responsible and accountable when errors occur, what guardrails are in place. These aren’t technical questions. They’re managerial ones.
Pitfalls to Avoid
First pitfall: confusing deployment with adoption. An installed tool is not a used tool.
Second pitfall: letting ungoverned AI develop in parallel. When your teams use personal AI tools without a defined framework, you lose control of data and decision consistency.
Third pitfall: measuring only post-training satisfaction scores. What matters is behavioral change three months later. Not the immediate feedback score.
What You Can Expect
Well-executed change management on an AI project doesn’t guarantee technological success. It guarantees that if the tool is good, it will be used. And that if adjustments are needed, your teams will be positioned to flag them rather than silently work around them.
That’s the difference between a project that lives and one that sits in an annual review report.
If you’re a CHRO or CEO and want to structure your approach before launching your next AI project, request a free diagnostic. We look at where you stand before talking about tools.
FAQ
What’s the difference between classic change management and AI project change management?
The structure is the same: diagnosis, coalition, communication, training, embedding. What changes is the intensity of resistance and the speed at which roles evolve. In an AI project, teams see their tasks transform within weeks. The pace of change management must match that speed.
Do you need an external consultant to manage change?
Not necessarily. What you need is someone internally who has the legitimacy and time to manage the human project in parallel with the technical one. If that person doesn’t exist, targeted external support on the diagnostic and design phase is worth more than a long, generic engagement.
How do you measure the success of AI change management?
Three concrete indicators: actual tool usage rate at 90 days, number of processes effectively modified, and the ability of teams to identify new use cases on their own. If that last indicator is positive, the change is embedded.
Are Kotter and ADKAR models still relevant in 2026?
Yes. Not because they’re perfect, but because they ask the right questions in the right order. Kotter forces you to build the coalition before communicating. ADKAR forces you to distinguish awareness from desire from capability. These distinctions remain useful regardless of the technological context.