Your Agent Is Live. Will Anyone Use It? 

Agent Born · Episode 03

Most conversations about AI agents seem to focus on the moment they are built. Common topics include whether the agent can retrieve the right information, make the right recommendation, take the right action, and perform reliably, securely and at scale.

I’m not saying that any of this doesn’t matter. Of course it does. What I am saying is that once the agent is live, a different question becomes more important:

Will anyone actually use it?

This is the point at which many AI projects and pilots start to lose momentum. Not because the technology is weak or the idea was wrong, but because some organisations treat the launch of an agent as the end goal when, in reality, it is the beginning of the change.

And this is where I think we need to be honest about AI adoption.

It’s still adoption. It’s right there in the name.

I’m going to use a phrase that used to annoy me, but now makes perfect sense, especially in this context:

“It’s the same, but different.”

The Same

In many ways, adopting AI is no different from adopting any off-the-shelf or well-designed bespoke business solution. Whether it’s a Microsoft Dynamics implementation, a new workflow, a new reporting model or a redesigned business process, the fundamentals remain the same.

People need to understand what is changing, why the change matters, how it fits into the way they work, how to trust it enough to use it and, eventually, how to build new habits around it.

The Difference

Traditional business systems are built around defined requirements, tested processes and a planned go-live. Once deployed, organisations often feel the work is complete. I know I’ve simplified things, but this process provides a sense of completion.

AI doesn’t behave like that.

An AI agent may have been created for one fixed process with one fixed outcome. However, the AI tool behind it, the one users can also interact with directly, sits across existing work, reveals new possibilities and changes how people approach tasks they already perform every day. It can draft, summarise, compare, challenge, retrieve, recommend and, in some cases, act.

That means adoption cannot simply be about teaching people which button to press.

It has to be about helping people rethink the work itself.

That is a much bigger shift.

When we worked with Hitachi Rail on Copilot adoption, this became very clear. The challenge was not simply to explain what Copilot or agents could do. The challenge was to help people connect those capabilities to real work: the repetitive task, the slow process, the document-heavy activity, the meeting follow-up.

Adoption starts when someone sees a genuine reason to use it.

But seeing the value once is not enough. At least, it’s not enough for me. Adoption happens when that first moment becomes a repeated behaviour.

One-off training rarely changes how people work. Yes, it may create awareness, and yes, it may create interest. It may even create enthusiasm. But habits are formed through repetition, reinforcement and relevance.

People need safe opportunities to try, fail, improve and try again. They need simple prompts they can use immediately. They need examples that feel relevant to their role. They need permission to experiment without feeling they have to become experts first.

Learning is not a side activity in AI adoption.

It is the adoption strategy.

AI itself can also become part of that learning journey.

Users can ask it to explain concepts, challenge their thinking, improve their prompts or compare different approaches. In other words, people do not only learn about AI. They learn with AI.

Use it. Learn from it. Improve. Share. Repeat.

In my experience, adoption accelerates when people learn from each other. A colleague sharing a useful prompt can be more powerful than a formal training slide. A champion explaining how they used an agent to save time in a real workflow can create more confidence than a generic demo. A community that shares wins, mistakes, questions and examples creates the momentum that keeps adoption alive.

This is especially important with agents.

An agent should not become a clever tool created somewhere else and handed to users at the end. Users need to understand what it is for, where it fits, what it can and cannot do, and how their feedback will shape its future. Otherwise, even a well-built agent risks becoming another underused tool.

People’s thinking will move from “Can AI help me with this task?” to “How could we redesign this way of working now that AI is available?”

That is the shift.

The organisations that succeed with agents and AI tools will not be the ones that treat adoption as a communications plan that starts after go-live. They will be the ones that build adoption into the way the agent is introduced, learned, tested, improved and shared.

Because a live agent is not the outcome.

A better way of working is.

Coming soon: Episode 04 | Governance & Long Term

In Episode 4, Andy Mooney will discuss why even the most intelligent AI agent is only as good as the data and knowledge behind it.