AI Change Management Best Practices for Stalled Programs

Ground truth before prescription

The first practice is the easiest to skip: start with an honest read of where you are today.

Most programs begin with a vision deck. A roadmap. A set of assumptions about what people need to learn and how quickly they will adopt. The strategy is built before anyone asks whether the organization is ready, whether trust exists, or whether the workforce believes the augmentation story.

Ground truth means admitting what you do not know. It means mapping where adoption breaks before you design the fix. It means treating resistance as data, not defiance.

In practice, this looks like asking three questions before you build the program:

• Where is adoption stalling right now, and who is routing work around the system?
• What do people believe about AI and their future, and is that belief grounded in what leadership has said and done?
• Where is trust already broken, and what past initiatives taught people that this one will not stick?

You cannot design a transformation program without answering those questions first. If you skip this step, you are prescribing in the dark. The best-case outcome is wasted budget. The worst case is you accelerate the erosion of trust.

The AI Profit Readiness Assessment is built for this. It gives you a first read in about two minutes: eight questions on how your team uses AI, and the first move to make.

People before Process before Platform

The second practice: sequence the transformation in the order that works.

It is common to do this backward. They choose the platform first. Then they design the process around what the platform can do. Then they tell people to adopt it and wonder why nothing changes.

The right sequence is People before Process before Platform.

People first means clarity on identity. Who do they need to become to do this work well? Not what they need to do - who they need to be.

A workforce that believes AI will replace them will not adopt a tool designed to augment them. The identity story has to be rebuilt first, or the behavior change will not follow.

Process second means designing workflows around how people actually work, not how the platform wants them to work. This is where you map the messy middle - the distance between the ideal state in the vendor demo and the reality of a Tuesday morning in operations. Process design is where you decide what gets automated, what gets augmented, and what stays human because trust or judgment or creativity cannot be delegated to an AI model.

Platform last means choosing tools that fit the people and the process, not the other way around. The platform should be the easy part. If you have clarity on identity and you have designed process around real work, the technology choice becomes obvious.

Teams get this backward because platform selection feels like progress. It is tangible. It has a vendor and a contract and a go-live date. Identity work is harder to measure, and process design forces you to admit that the current state is messier than anyone wants to say out loud.

But if you skip People and Process, Platform will not save you. You will have expensive software and flat adoption, and the explanation you give the board will sound like every other failed transformation: the people were not ready.

The real explanation is simpler. You never made them ready. You assumed the platform would do that work for you.

Resistance is data, not defiance

The third practice: treat pushback as a source of insight, not an obstacle to overcome.

When a manager routes work around the AI tool, that is not stubbornness. That is a signal. When a high performer asks in a town hall whether their job is safe, that is not fear-mongering. That is a trust gap you need to name and repair.

Most programs treat resistance as something to manage. They add more training. They escalate to leadership.

They mandate adoption and measure compliance. None of that changes the underlying belief driving the behavior.

Intelligent resistance is not random. It shows up in predictable places:

• Middle managers who have been through several transformation programs in a few years and watched each one fade.
• High performers who see AI as a threat to the expertise that made them valuable.
• Frontline staff who were told the last system would make their jobs easier, and it made them worse.

These people are not the problem. They are showing you where the program will break if you do not redesign it.

The best practice is to listen before you prescribe. Map the resistance. Ask what past programs taught people about how this company handles change. Ask what they believe about AI and whether leadership has earned the credibility to tell them it will be different this time.

Then design the program to address what you find. If trust is broken, rebuild it before you ask for adoption. If the identity story is wrong, rewrite it. If middle management does not believe, do not mandate compliance - give them a reason to model the behavior.

Resistance is the data you need to make transformation real.

The messy middle is where transformation happens

The fourth practice: design for the space between announcement and adoption.

A program usually has a clean start and a clear target state. The messy middle, the long stretch where behavior changes, gets treated as execution when it needs design.

This is where programs die.

The messy middle is where people test whether leadership actually believes the augmentation story. It is where middle managers decide whether to model the new behavior or route work around it. It is where the workforce learns whether this transformation is real or performative.

You cannot script this phase. You can design for it.

Designing for the messy middle means:

• Building feedback loops that surface what is breaking in real time, not in quarterly reviews.
• Giving middle managers the language and the autonomy to adapt the program to their team's reality.
• Celebrating early adopters not for compliance, but for the results they are driving.
• Admitting publicly when something is not working and redesigning it, instead of pretending the plan was right all along.

The messy middle is not chaos. It is the transformation doing its real work. If you try to control it, you kill it. If you design for it, you give it room to succeed.

Why most best practices fail in practice

The problem with best practices is that most of them are borrowed from organizations that are not yours. They worked somewhere else, under different conditions, with a different culture and a different starting point.

You cannot import someone else's transformation design and expect it to work. You can learn the principles. You have to build the program yourself.

The principles that hold regardless of where you start:

• Start with ground truth, not prescription.
• Sequence People before Process before Platform.
• Treat resistance as data.
• Design for the messy middle, not just the start and the end state.
• Rebuild trust before you ask for adoption.

Those principles are not controversial. They are common sense. The reason most programs ignore them is that they require admitting what you do not know, slowing down to go faster, and designing transformation as a human problem first and a technology problem second.

The shortcut is tempting: the vendor deck, the clean timeline and the promise that if you follow the playbook, adoption will follow. Treat transformation as a design problem instead, starting with an honest read of where your people are.

What to do instead

If your AI program is stalled, the fix is not more training or a mandate from the top. The fix is stepping back and asking whether you designed for the reality you are in, or the reality you wished you had.

Start with ground truth. Get a clear read on where adoption is breaking and why. Not what you think is happening - what is actually happening. The AI Profit Readiness Assessment will give you that read in about two minutes, and it will show you where the gaps are before you invest in the wrong fix.

If the read shows gaps that need hands-on work to close, the next step is the AI Profit Sprint. It is a 90-day engagement for leaders who need more than a diagnosis. It walks you through how to sequence People, Process, and Platform in the right order, how to design for the messy middle, and how to rebuild trust before you ask people to adopt.

Ready to get a clear read on where your AI program is stalling?

A program can have the right strategy and still stall because the design skipped ground truth.

If you are accountable for an AI transformation that is not moving, start with a discovery call. We will walk through where adoption is breaking, what your people believe about AI and their future, and whether the program you have is designed for the organization you are in.

Book a discovery call here. It is a conversation about what is in the way and what to do about it.