Leading AI Adoption: Lessons From Building an AI Taskforce

Why AI isn’t about replacing people. It’s about empowering them.

If you’d asked me a few years ago whether I’d end up leading an AI Taskforce, I probably would’ve laughed.

At the time, I was already managing several technology projects when an opportunity came up to help drive AI adoption across CQUniversity. I went into it thinking that the biggest challenge would probably be understanding the technology and working out where we could use it.

What I found was quite different.

The bigger challenge was helping people understand where AI actually fit into their work, giving them the confidence to experiment with it safely, and making sure they understood that the goal wasn’t to replace what they did, but to help them do it better.

There was a lot of excitement around AI, but there was also a fair amount of uncertainty. People had questions about what these tools meant for their jobs, what information they could safely put into them, and whether the outputs could actually be trusted.

Looking back, that experience changed the way I think about AI implementation.

Today, I use a lot of the same lessons across software delivery, product management and my own day-to-day work. The tools have changed incredibly quickly, but the fundamentals haven’t really changed much.

For me, successful AI adoption is as much about people as it is about technology. You can give an organisation access to some of the best AI tools available, but if people don’t understand how to use them, don’t trust them, or can’t see how they fit into their work, adoption is going to be difficult.

Start with the problem, not the technology

One of the things we started doing in our AI workshops was asking people a really simple question:

“What’s the most frustrating part of your day?”

I liked this question because it stopped us trying to find somewhere to use AI and got us talking about the actual problem.

Instead of asking people where they wanted to use AI, we asked them to talk about the parts of their job that were repetitive, manual or simply taking more time than they should.

People talked about writing meeting minutes, digging through emails, copying information between systems, trying to find documentation and repeating the same administrative tasks every day.

From there, we had something solid to work with.

We could map out the current process, understand how often it happened, who it affected and roughly how much time it was taking. Then we could start looking at whether AI was actually the right tool to improve it.

Not every idea needed to become a project either.

If we were only going to save one person a few minutes once a month, there were probably better things for us to focus on. But if the same problem was affecting an entire team, happening regularly across the organisation, or improving it would have a direct impact on our students, then it became something worth looking at more closely.

We also used those workshops to bring people from different areas of the organisation into the same conversation. That gave us a better understanding of where similar problems were appearing across different teams and helped us prioritise ideas based on their broader impact.

We also reported those findings back to our CTO so he could see what we were discovering and which opportunities we thought were worth taking further. That helped us prioritise the ideas and make sure what we were exploring lined up with the broader direction of the organisation.

That was probably the biggest takeaway for me from those workshops. We weren’t trying to find places to force AI into the business. We were trying to understand where people were losing time, getting frustrated or doing repetitive work, and then work out whether AI could help.

AI implementation is really change management

One of the things that stood out to me while leading the AI Taskforce was that the technical side was often not the hardest part. The bigger challenge was helping people feel comfortable using AI and understanding where it could fit into their work.

There were genuine concerns across the organisation about what AI might mean for people’s roles, what information they could safely put into these tools, and where that information might end up. For some people, that uncertainty was enough to stop them experimenting with AI altogether.

Because of this, our workshops weren’t just about showing people what tools like Microsoft Copilot, ChatGPT and Claude could do. We also spent a lot of time discussing cybersecurity, responsible use and what information should and shouldn’t be shared with different AI platforms.

We worked with our internal cybersecurity teams to better understand how these tools handled data and used that information to build guidance for staff. If someone was working with internal or sensitive information, for example, we could explain why an approved tool like Microsoft Copilot might be more appropriate than putting that same information into a public AI platform.

Another important part was making sure people understood that AI-generated information still needed to be checked. AI could help draft something, summarise information or give you a place to start, but the person using it was still responsible for making sure the output was correct.

What I found was that giving people clear boundaries actually made them more willing to experiment. Once they knew what they could safely do, what they shouldn’t do, and where they could ask questions, a lot of that initial hesitation started to disappear.

For me, this was when I started to realise that successfully introducing AI was going to involve a lot more than giving people access to the technology. It was also about education, trust and bringing people along for the journey.

Build champions, not just training sessions

Another thing that worked really well was identifying people across the organisation who were genuinely interested in AI.

As we ran the workshops, there were always a few people who were keen to experiment, ask questions and start thinking about how the tools could help their own teams.

Rather than trying to have the AI Taskforce own every idea or answer every question, we started encouraging those people to become champions within their areas.

They could help their colleagues, share examples of what was working, and bring questions back to us when they weren’t sure about something.

This also made the adoption feel a lot less top-down.

People weren’t only hearing about AI from a central taskforce. They were seeing people they worked with using the tools and finding practical ways to make their jobs easier.

Over time, we could see that interest growing. Some of our early workshops only had around five people involved, while later sessions were bringing in 20, 30 or sometimes around 40 people.

We also set up a Teams community where staff could ask questions, share ideas and get support. That grew from roughly 100 people to around 300 while I was involved.

For me, that was probably one of the clearest signs that the approach was working. People weren’t just attending workshops anymore. They were starting to experiment themselves, bring us ideas and help each other work out where AI could be useful.

Focus on practical wins

One of the quickest wins we found was also one of the simplest: meeting minutes.

Using Microsoft Copilot, we could generate a transcript, pull out the key points and create action items without someone having to spend extra time writing everything up afterwards.

We still reviewed the output before sharing it, but it gave us a really good starting point and meant people could move on to the next piece of work much faster.

We also kept improving the process rather than treating the first Copilot output as finished. If the meeting summary wasn’t pulling out the information people actually needed, we’d adjust the prompt and try again. Over time, the notes became much more useful, particularly for people who hadn’t been able to attend the meeting, and teams could start acting on the outcomes sooner.

I’ve taken a lot of the same thinking into my work since then. At 3D Walkabout, for example, we’ve been using Unity’s command-line tools alongside AI agents to automate parts of how we test our VR training applications.

Before, testing a VR interaction meant a developer or tester had to put on a headset, load into the environment and manually work through the interaction to make sure it behaved as expected. We’re now able to have AI agents run through those interactions as part of the development process and carry out end-to-end interaction checks inside Unity.

This gives us another way to catch issues earlier and reduces some of the repetitive testing developers would otherwise have to do manually. We still need people to test things like usability and how an interaction actually feels in VR, but by the time we get to that point we’ve already been able to validate more of the underlying behaviour.

They’re very different examples, but the thinking behind them is the same. Neither is about trying to completely automate someone’s role. They’re about finding repetitive parts of a process where AI can take some of the workload away.

I’ve found that once people can actually see that happening in their own work, it becomes much easier to start talking about where else AI could be useful.

Keep people in the loop

One principle I’ve carried through every AI initiative I’ve worked on is that there should still be someone accountable for the important decisions.

AI can help generate documentation, review code, summarise information, identify patterns and even recommend what to do next.

But when those decisions have a real impact on the business, I still think there needs to be an appropriate level of oversight from someone who understands the context and is responsible for the outcome.

I’ve carried that same thinking into software delivery as well.

For example, AI can review code, flag potential issues or help automate parts of a deployment process, but I still want someone to review what it’s recommending before significant changes make it into production.

I’m a big fan of automation, but I don’t think the goal should be to remove people from every step of the process. For me, it’s about removing the repetitive work while keeping the right people involved where their judgement and experience actually matter.

AI has changed how I work

All of this has also changed the way I work personally.

Today, AI is part of a lot of what I do, whether I’m working through a technical problem, planning a project, reviewing an idea or trying to make sense of something complex.

A lot of the time I’ll start with a fairly rough idea and use AI to help me pull it apart. I’ll question the assumptions, look at different approaches and start turning the idea into something I can actually build.

From there, I might use it to help structure an implementation plan, break the work down into GitHub issues, review an architecture decision or improve the documentation as I’m building. I’ll often keep using AI throughout the process as another set of eyes rather than only using it at the start.

I don’t really see that as AI thinking for me.

For me, it’s more like having another tool that enhances what I already do, working alongside my own experience to test ideas and get to a better outcome faster.

Sometimes I agree with what it gives me, and sometimes I don’t. But even when I disagree, the process of questioning it can still help me think through the problem more clearly.

That’s probably where I get the most value from AI in my own work. It doesn’t replace the experience or judgement I’ve built up over the years. It gives me another way to apply it.

A great example of AI done well

One example I really like is what IKEA did with its customer service teams.

IKEA introduced an AI assistant called Billie to help handle routine customer enquiries. From 2021 to 2023, Billie resolved around 47% of the enquiries it received, while IKEA reskilled 8,500 call centre staff in areas including remote interior design and digital retail sales.

Rather than only looking at AI as a way to automate customer service, they also looked at where their people could add more value.

Instead of spending as much time answering the same routine questions, those staff could use their knowledge of IKEA products and work more directly with customers on interior design.

I think this is a really good example of the opportunity AI creates.

AI will change jobs, and I don’t think we should pretend that it won’t. But that doesn’t mean the conversation always needs to be about removing people.

It can also be about looking at the work AI is taking away and asking what that now gives people the opportunity to do instead.

That’s much closer to how I think about AI adoption. Use the technology to take away some of the repetitive work, then look at what that allows people to spend more of their time doing instead.

You can read more about this example in IKEA’s article, AI and remote selling bring IKEA design expertise to the many.

Final thoughts

Looking back at my experience leading the AI Taskforce, and how I use AI in my work today, I think the biggest lesson for me is that successful AI adoption is about much more than the technology.

You need the right tools, security and governance in place, but you also need to bring people along with you.

People need to understand why you’re introducing AI, where it can genuinely help them, what the boundaries are and where their own experience and judgement still matter.

I’ve also found that the best opportunities don’t usually start with someone saying, “We need to use AI here.”

They start with someone explaining a problem they’re having, a process that’s taking too long, or a part of their job that feels unnecessarily repetitive.

That’s why I still like going back to the same question we used in those early workshops:

“What’s the most frustrating part of your day?”

Once you understand that, you can start working out whether AI can actually help.

For me, that’s what good AI adoption looks like. Not trying to put AI everywhere, but using it where it can make a real difference and giving people more time to focus on the work where their experience matters most.