A few months ago, I consulted with two organisations implementing identical AI platforms. Same vendor, same features, same implementation timeline. One is now seeing significant productivity gains. The other has abandoned the project entirely.

The difference wasn’t technical. It was human.

Organisation A invested heavily in change management. They communicated early and often. They addressed concerns directly. They trained people thoroughly. They adjusted based on feedback.

Organisation B treated the implementation as a technical project. They focused on configuration and integration. They announced the change two weeks before go-live. They offered a single training webinar. They were surprised when adoption stalled.

This pattern repeats constantly. The technology works. The change management doesn’t.

What Change Management Actually Means

Change management has become one of those corporate phrases that sounds meaningful but gets applied to everything. Let me be specific about what I mean.

Change management is the systematic approach to moving people from a current state to a desired future state. The Prosci ADKAR model, widely referenced in change management literature, includes:

• Creating awareness of why change is happening
• Building desire to participate in the change
• Developing knowledge of how to change
• Providing ability to implement new skills
• Reinforcing to sustain the change

If any of these elements is missing, adoption fails. You can have perfect awareness but no desire. You can have desire but no knowledge. You can have knowledge but no practical ability. And even if all of those are present, without reinforcement, people revert to old behaviours.

Why AI Implementations Need More Change Management, Not Less

Some leaders assume that AI tools are intuitive enough that extensive change management isn’t necessary. “People use technology in their personal lives,” the thinking goes. “They’ll figure it out.”

This underestimates several factors:

Professional identity is at stake. Using a new social media app doesn’t threaten how you see yourself. Adopting AI tools that might reshape your role does. The psychological stakes are higher.

Habits are deeply ingrained. People have been doing their jobs a certain way for years. Those patterns don’t change because a new tool is available. They change through deliberate effort over time.

The capabilities are ambiguous. Most consumer technology has obvious use cases. AI tools are open-ended. People need guidance on what to use them for, not just how to use them.

Fear is a factor. Concerns about job displacement, skill obsolescence, and being left behind create resistance that technical solutions can’t address.

The Prosci ADKAR Model in Practice

The ADKAR model I referenced earlier provides a useful framework. Here’s how I apply it to AI implementations:

Awareness

People need to understand not just what’s changing, but why.

The “why” matters more than most organisations realise. “We’re implementing AI to improve productivity” doesn’t create awareness. It creates questions: Whose productivity? At what cost? Does this mean layoffs?

Better: “Our customer service team is struggling to respond within our target window. This creates frustration for customers and stress for the team. We’re implementing AI tools to help the team respond faster while maintaining quality. No one’s role is being eliminated.”

Build awareness through:

• Clear, honest communication from leadership
• Opportunities for questions and discussion
• Explanations of what this means for specific roles
• Transparency about what you know and don’t know

Desire

Awareness doesn’t automatically create desire. People can understand why change is happening and still not want to participate.

Desire comes from answering “what’s in it for me?” That answer varies by individual:

• Reduced tedious work
• Opportunities to develop new skills
• Better outcomes for customers
• Career advancement possibilities
• Keeping up with industry developments

Identify what motivates different groups and communicate accordingly. And be honest—if the primary benefit is organisational efficiency, don’t pretend it’s about employee wellbeing.

Knowledge

Once people want to change, they need to know how.

This is where training comes in, but knowledge goes beyond formal programs:

• Documentation and job aids
• Examples and templates
• Peer support and mentoring
• Access to expertise when stuck

Knowledge-building should be ongoing, not a one-time event. People need different information at different stages of their adoption journey.

Ability

Knowledge and ability aren’t the same thing. I can know how to ride a bicycle and still fall off when I try.

Ability develops through practice. Organisations need to create conditions for that practice:

• Time allocated for learning and experimentation
• Safe spaces to make mistakes
• Feedback on performance
• Opportunities to apply skills in real work

A common failure mode is expecting immediate proficiency after training. Ability takes time to develop.

Reinforcement

This is where most change initiatives fail. Initial enthusiasm fades. Old habits reassert themselves. Without reinforcement, adoption plateaus or declines.

Reinforcement mechanisms include:

• Recognition for successful adoption
• Ongoing communication about progress
• Updating processes to embed new ways of working
• Removing barriers that make old approaches easier
• Continued investment in capability development

Common Change Management Mistakes

Starting too late. Change management often begins when implementation is already underway. By then, rumours have filled the information vacuum and resistance has calcified.

Treating communication as a one-way broadcast. Sending emails and newsletters isn’t communication. People need dialogue—opportunities to ask questions, voice concerns, and get real answers.

Underestimating middle management. Senior leaders set direction. Frontline employees do the work. But middle managers determine whether change actually happens. Invest heavily in this group.

Assuming one approach fits everyone. Different people have different concerns, different motivations, and different learning needs. Segmented approaches work better than universal ones.

Declaring victory too early. Initial adoption numbers don’t mean the change has stuck. Track behaviour over months, not weeks.

The Resource Imbalance

Here’s a statistic that should concern everyone in technology implementation: organisations typically spend 10-20% of project budgets on change management. Research consistently shows that projects with robust change management are several times more likely to succeed.

We underinvest in the thing that determines success.

I understand why this happens. Change management feels soft. Technology feels tangible. It’s easier to justify spending on things you can touch.

But the ROI on change management consistently exceeds the ROI on additional technical features. Every dollar spent helping people adopt technology creates more value than another dollar spent on implementation.

Making the Case Internally

If you’re trying to secure resources for change management, frame it in business terms:

• “Our last three technology implementations fell short of projected benefits. The common factor was adoption. This time, we’re investing appropriately in the human side.”

• “The technology investment is $X. To realise the expected benefits, we need adoption of at least Y%. Based on similar initiatives, that level of adoption requires Z investment in change management.”

• “We can implement quickly with minimal change support, or thoroughly with comprehensive support. The first approach costs less upfront but historically delivers lower returns.”

The Ongoing Nature of Change

AI isn’t a one-time change. The technology evolves constantly. Use cases expand. New tools emerge. Capabilities improve.

This means change management isn’t a project phase that ends. It’s an ongoing capability the organisation needs to maintain.

Build sustainable change infrastructure:

• Internal change management expertise
• Communication channels that persist
• Learning systems that can scale
• Feedback mechanisms that capture ongoing experience

The organisations that do this well don’t just survive one AI implementation. They build the muscle to adapt continuously.

The Bottom Line

Technology implementation fails at the human level far more often than it fails at the technical level. Yet most organisations invest almost entirely in the technical side.

Change management isn’t an add-on. It’s not a nice-to-have. It’s the factor that determines whether your technology investment creates value or becomes expensive shelfware.

Invest accordingly.