As organisations move beyond the initial wave of AI experimentation, they’re asking different questions. How can it be used strategically to create lasting value? And what capabilities do leaders need to navigate the opportunities and challenges ahead?

There's a mounting pressure on senior leaders to think fast when it comes to AI adoption. However, many are already finding that the costly AI tools that they invested in are not offering the return of investment they imagined. And unregulated AI usage has already led to significant reputational damage for leading global corporations.

In a recent panel discussion (see below for full recording), our Portfolio Development Lead Emily Tannert Patterson spoke to Per Ola Kristensson, Professor of Interactive Systems Engineering, University of Cambridge, about the essential steps to leading successful AI adoption. From understanding your goals, to mapping your processes, and giving your teams the flexibility and training they need to use it with confidence.

Per Ola leads The AI Leader: From Strategy to Transformation – an in-depth programme based on the latest University of Cambridge research and industry insights – for leaders who want move beyond experimentation and drive strategic, long-lasting AI adoption that gets meaningful results.

Emily: Now, almost four years on from the infamous date, 30 November 2022, when ChatGPT became available to launch to the world, how are organisations progressing with their AI adoption journey?

Per Ola: In the first years, I saw a lot of optimism but also fear and top-down management. But just giving people this versatile, complex tool – without the right frameworks in place – tends not to work. You need to be able to explain to people why it’s useful for them and give them the training they need to use it effectively.

A big risk is complacency. This isn’t about people being deliberately complacent. It’s usually subconscious and develops over time. Whenever a human feels that part of their work is being carried out by an agent, you get this subconscious complacency which can lead to mistakes and things slipping through the net.

Another issue is de-skilling. People forget how to do things or simply lose the ability to do things. A little bit like how being over reliant on a calculator means you lose the ability to do arithmetic in your head. The same thing is now happening with writing. There have been lots of studies about university students who have essentially lost the ability to produce technical writing on their own, which is important because writing is thinking.

As AI tools are increasingly relied on, they could reduce people’s understanding of what is happening in their environment, the data that’s being processed and the implications of decisions.

The good news is that there are systematic frameworks that can help map out your organisation’s workflows and capabilities, so you can see what you can and can’t automate. And answer questions like: what degree of automation do we want? What extra functions do we need to govern it, reduce the risk, and ensure efficiency? And what are the adverse automation effects we have to monitor?

Emily: We're now at the point of maturity where some organisations are starting to get it right. So what are you seeing from them that others could harness?

Per Ola: The way to get it right is to fully understand your workflows. You have to understand how people do their work. Only then can you think about how to reconfigure it – and what you can automate – so people can be more effective and efficient.

This isn’t just about AI. You also need to look at the human processes – who inputs the data, takes responsibility for outcomes, and so on. This sounds like a lot of work, but it's not. All these things can be done in matters of hours or perhaps days, depending on the complexity. It gives you this map of opportunities and the clarity to make decisions about AI automation.

Emily: I can’t help but think that there's a lot to take in. What would a senior leader need to know?

Per Ola: As a senior leader, you have to be educated about AI. You don't have to know how the technology works, and I don't think that's even useful, because it changes every day, and there will be lots of new models coming out.

What’s important to understand are the frameworks. How to think about AI, how to map out your processes and make decisions about automation. How to get clarity on the risks that come with automation, and how to communicate all this in a way that everyone in the organisation can understand. This is the knowledge that is vital for good decision-making.

The nature of decision-making means there is always a cost to the decision. If there's no cost, it's not a decision. So you have to know enough to understand that cost. You probably understand your own business very well. You now also need to know how to think systematically about AI deployment.

Emily: What are the biggest misconceptions you still find that senior leaders have about AI, and particularly how these might impact their strategic decision-making?

Per Ola: It's phasing out a little bit now, but the idea that it's basically a tool, like a spell-checker, that you can just deploy. You give everybody something like Copilot, and they will just figure it out, and somehow magically we will be 20% or 30% more productive. That signals that the leaders involved don't really understand their own business.

The first thing you need to do is to understand how you're making money. What are your processes? Who reports to who? Where is the information flowing? Do we want to keep that workflow, or do we want to change it?

But regardless of decision, we're going to have to worry about at least two things. The first is change management, and depending on your organisation, this kind of change can be difficult to manage. It requires a strategic and systematic approach.

The other thing is the same as what happened when we started introducing IT and later the internet into organisations. And that is the phenomenon of technological appropriation – where people are figuring out how to use technology to further their own goals.

If people can't figure out how to appropriate something, they're just not going to use it. This is why it's so important to design enough flexibility into workflows, and provide enough training, support and incentives, while alleviating any fears people may have.

Emily: Do you have tips on how to encourage people to appropriate AI for their roles and encourage them to use it?

Per Ola: The most important part is training. Aspire to make everybody a champion of the technology, not a detractor. And you can do that by fully explaining the functions of the technology.

The most important mantra for appropriation is support, not control.

There will always be cases where people need do things that you didn't think about when you designed a process. If you created a very strict process, when cases like this come up people will deviate from it.

A classic example is procurement. If you have a strictly designed procurement system, what happens when you have to procure something you didn’t think about before? Rather than using your process, people will pick up the phone or use pen and paper. Then this supposed low-risk system becomes higher risk.

So one of the best things to do is to have this mantra in mind. Don’t overly control people. Train them so they understand the possibilities and support them to do things in a variety of ways.

Watch the full webinar

Want to lead AI transformation that gets meaningful results?

The AI Leader: From Strategy to Transformation gives you everything you need to drive successful AI adoption – from analysing opportunities to building frameworks and managing risks.

You’ll learn from Per Ola and other renowned University of Cambridge AI experts and develop a practical AI strategy unique to your challenges and goals.

See more of our AI insights, courses and bespoke learning opportunities(Opens in a new window).