Why AI? How leaders can think more strategically about AI adoption

Oscar Fernandez
Director and head of asset management tech transformation and architecture, Swiss Re

Belinda Finch
CIO and executive vice president, IFS

Phil Le-Brun
Director of enterprise strategy, AWS

Clare Walsh
Director of education, the Institute of Analytics

It can be easy to rush into getting new technology without stepping back and thinking, ‘Why do I want this?’ The most important consideration is the business problem you’re trying to solve and understanding different use cases. For example, at IFS we’ve got a cross-functional, executive-led team on our internal productivity project. We’re trialling several use cases and working out what AI can actually help with. 

It’s like the old adage of ‘If all you have is a hammer, everything looks like a nail’. You need to give that senior cross-functional team the business outcome you want. Then trust your people to get on with it and let them experiment. There may be multiple ways of solving a problem. I truly believe AI, machine learning and generative AI will play a major part in businesses going forward. But it doesn’t mean they’re the only answer. Give people the freedom to explore and experiment quickly and cost-effectively, and then use the cloud to scale those ideas. 

We’re in a phenomenal situation today with tech no longer being the barrier. Now, it’s down to people’s imagination and getting to the root cause of what they’re trying to fix.

You have to have a strong ‘why’. We see people with lots of ideas that struggle to articulate their value. This is usually a sign that they didn’t start with a problem, they started with the technology.

You should clearly formulate what your strategy and ambitions are when you’re using GenAI. Building on the hammer-and-nail example, you might have identified a big problem to solve but that doesn’t necessarily mean it’s a GenAI problem. You need to understand what you’re fixing and then whether GenAI is the answer. There’s also an educational element to that. It’s key to explain to people when they should expect GenAI to be a solution and when not.

We get the best answers when asking, ‘What frustrates you?’ And that’s usually the boring repetitive stuff. That’s often what we can quickly and easily replace. It’s important to move away from all-or-nothing approaches, e.g. you have to use AI or you have to have a human system.

We’re finding the majority of chief data officers say that getting their data foundation in place is essential. But companies don’t need to do it alone. Most companies are reliant on electricity, but they don’t build their own power plants; they rely on others to do that. It’s the same with the cloud. It can help lower the cost of experimentation, give you access to the widest range of tools possible and help you scale a good idea.

You may have all your data in the right place – and that’s important – but if the quality isn’t there it messes everything up. So data quality and the governance around the organisation of data are key. 

It’s an interesting challenge because it comes down to people, organisation and culture. A data and AI strategy can easily be made the CIO’s problem, even though it involves business data. 

A lot of C-suite education needs to happen to familiarise them with their roles in this. Generative AI has opened up a broader conversation about data, AI and machine learning because it’s become a boardroom topic. And if you’re at board level, you’re expected to understand people management and finance. Why wouldn’t you be expected to understand technology and data from a strategic perspective?

In some ways, it’s amazing if you’re an early adopter because you could catapult the business ahead. But equally, you could put your money on the wrong horse and have to redo everything. I think boards have to be really careful in evaluating exactly what they’re doing with AI, rather than thinking it’s the answer to everything.

Even if you’re picking the right use case, it’s important to build your solution in a way that allows you to pivot and switch as market offerings change. Incorporating this agility as part of your use case design process will provide you mechanisms to protect your investments.

Large language models are going to continue to develop. Approach this by thinking about how you can ensure you pull in the right model at the right time and can also switch to a different model in the future. That’s our approach because we know this market is not going to slow down.

Finding portfolio balance and identifying low-hanging fruit. The latter are use cases where you can quickly demonstrate value and use that to build momentum. In areas such as personal productivity, there are no-regret moves you could start leveraging now. Thinking more broadly, there are likely benefits across all of your value chain if you can spot the right pain points and how to apply AI solutions.

One of the questions I’m hearing more and more is: ‘If I make someone more productive, what does that mean?’ If you free up someone’s time, where’s the free time going? If you double the development capacity of a developer, are they just developing twice as much stuff no one’s using? In that last example, we think about development productivity, not developer productivity. That means improvements in delivering value to customers faster, often through tweaks to processes accompanied by technology.

The people who are still saying ‘Can I just ignore it?’ or ‘Will it blow over?’ should consider the evolution of mobile phones. None of us remember how we learned to use our mobile phones. We just continually bought a slightly better update. And now here we are, 10 or 15 years on and completely fluent in how to use them. People who didn’t go on that journey have been left behind.