There’s a tendency to talk about AI implementation as though it’s a technology project.
Choose the right platform, get the licences, train people, write a policy, turn it on and then, somehow, we act surprised when people don’t use it.
The technology is rarely the hardest part of organisational change. The harder part is asking people to change something they’ve become comfortable doing, particularly when they’re not entirely sure what the change means for their job, their value or their future.
That’s where change management matters, and where stakeholder management can make the difference between AI becoming something people reluctantly tolerate and something they genuinely see as useful.
I’ve spent much of my working life around organisational change, and one thing I’ve learnt is that people don’t usually resist change simply because they’re difficult or conservative. They resist change when they don’t understand it, when they don’t feel part of it, or when they suspect that a decision has already been made somewhere above them and their job is simply to accept it.
For some people, it’s exciting. It removes the boring parts of a job, makes research faster, helps with writing, analysis and administration, and gives people access to capabilities they didn’t have before. For others, it’s unsettling, and understandably so.
If a machine can produce a first draft in seconds, what does that mean for the person who used to spend half a day producing the first draft? If AI can analyse a document, summarise a meeting or create a presentation, where does that leave the person whose role has traditionally involved doing those things?
Those are reasonable questions, and telling people to “embrace AI” doesn’t answer them.
This is why I think the conversation about AI in organisations needs to move beyond technology adoption and into something much more human, because the real challenge isn’t simply teaching people how to use a new tool, it’s helping them understand how that tool might change the way they work and giving them enough confidence and agency to work that out for themselves.
Change management has been around for decades, but AI gives us a particularly interesting test of whether we actually understand it.
The old model of organisational change often looked something like this: leadership makes a decision, develops a plan, communicates the plan, trains employees and then measures adoption. There’s nothing inherently wrong with that sequence, but AI is moving too quickly for organisations to rely entirely on it, because the technology itself is changing while people are still working out what it means.
That makes stakeholder management more important, not less.
Stakeholder management is sometimes treated as a communications exercise. Identify the stakeholders, develop some messages, send some emails, hold a town hall and tick the box.
But good stakeholder management isn’t really about getting information out. It’s about understanding what people need to know, what they’re worried about, what they might contribute and where they might influence others, which means listening has to be just as important as communicating.
That distinction matters enormously when introducing AI.
Take a finance team, for example. The executive team might see AI as an opportunity to increase productivity, the finance team might see an opportunity to automate repetitive reporting, someone in compliance might immediately be thinking about privacy, data security and accountability, while someone in IT might be wondering what this means for existing systems. And an employee who has spent ten years building expertise in a particular process might quietly be wondering whether the organisation still needs their expertise.
They’re all looking at the same change but they’re just standing in different places. The mistake is assuming that one message will work for everyone.
A good change programme starts by understanding those different perspectives before trying to persuade people of the benefits, because if you begin with the assumption that everyone simply needs to be convinced, you can easily mistake legitimate concerns for resistance.
Who are the people who will be directly affected? Who has influence over others? Who is likely to be enthusiastic? Who is sceptical? Who has legitimate concerns? Who understands the organisation well enough to see problems that senior leaders might miss?
And perhaps most importantly, who do people actually listen to? That last question is often overlooked because we tend to confuse position with influence.
The person with the biggest title isn’t necessarily the person with the greatest influence. In most organisations there are people who have accumulated what you might call informal authority – the experienced employee everyone asks when something doesn’t make sense, the person who knows how the organisation really works rather than how the organisational chart says it works.
If those people become advocates for change, they can help move an organisation considerably faster. If they become opponents, they can slow it down just as effectively.
That doesn’t mean you need to turn every influential person into an AI evangelist. In fact, that can be counterproductive because people can tell when they’re being recruited to sell something.
Instead, bring them into the conversation early. Ask them what they think, what could go wrong, where AI could genuinely make their work easier. But primarily ask what shouldn’t be automated.
And then listen to the answers. There’s a subtle but important difference between consultation and communication.
Communication says, “Here’s what we’re doing.”
Consultation says, “Here’s what we’re considering. What are we missing?”
The second approach gives people some ownership of the change, and that ownership matters because successful change isn’t something an organisation does to its people. Eventually, it has to become something people do themselves.
This is particularly important with AI because there’s another layer of uncertainty underneath the technology.
Trust.
People need to trust the technology, but they also need to trust the organisation using it.
They need to know what happens to the information they put into an AI system, they need to understand when they can rely on an output and when they need to check it, and they need to know who is accountable when AI gets something wrong.
They also need some honesty about the future. Pretending AI won’t change jobs is no more helpful than telling everyone that their jobs are about to disappear. Neither is particularly useful, and both approaches undermine trust because most people can see that something is changing even if they don’t yet know exactly what that change will look like.
The more honest conversation is that AI will change many jobs, but exactly how it changes them will depend on the organisation, the role and the choices people make along the way.
That’s where change management becomes less about managing resistance and more about creating the conditions for adaptation.
Give people permission to experiment, give them practical examples rather than abstract promises, create spaces where someone can say, “I tried this and it didn’t work,” and let people share what they’ve discovered. Recognise that some employees will move quickly while others will need more time, and don’t confuse training with change.
Training someone to use an AI tool doesn’t necessarily mean they understand why they should use it, where it fits into their work or when they shouldn’t use it.
The most useful AI training I’ve seen is often much less about buttons and features and much more about real work that we need to bring as human beings.
That last part is becoming increasingly important, because the more capable AI becomes, the more valuable human judgement becomes.
In other words, AI can increasingly help us do the work, but it can’t take away our responsibility for deciding what work is worth doing in the first place.
This is why I’m increasingly wary of the language of “AI transformation” when it’s used without any reference to people.
An organisation doesn’t transform because it has bought AI. It transforms when people change the way they work, and people change the way they work when they understand the reason for the change, have some influence over how it happens and believe the organisation is being honest with them.
That’s a communications challenge as much as it is a technology challenge.
It’s also why stakeholder management shouldn’t be something bolted onto an AI implementation after the technology has been selected. It should be there from the beginning, before the procurement, before the launch and before the training calendar, because stakeholders can tell you things that a technology assessment can’t.
They can tell you where the friction will be, which processes are likely to benefit and which fears are genuine and which ones are based on misunderstanding.
They can tell you who needs to be in the room and sometimes they can tell you that the organisation is trying to solve the wrong problem. That may be the most valuable contribution of all.
We often talk about change as something that needs to be managed. Perhaps we should think about it as something that needs to be understood.
AI is going to keep arriving in our workplaces whether organisations are ready for it or not. The question is whether we choose to drag people behind it or bring them with us.
The organisations that get this right won’t necessarily be the ones with the most sophisticated AI systems. They may simply be the ones that understand that technology changes quickly, while trust takes time.
And that when people are given a voice in change, they’re much more likely to help shape where it goes.
That’s the human part of AI we shouldn’t automate away.
Follow Adrian Drayton on his Substack – The Long Conversation – where he writes about faith, culture and media and the questions shaping our common life.
