How AI changes IT strategies

When asked how the advent of AI in business has changed the CIO role, Loake draws an analogy with the late 1990s, when the web and ecommerce were the main talking points in board meetings. “If you go back to the dawn of the internet, AI is very much like that,” he says. “Everybody has to have an AI strategy. It’s like having an internet strategy.”

For Loake, while many internet strategies were not worth the paper they were written on, having a well-articulated AI strategy is like a North Star, which broadly stipulates an AI-enabled businesses. “We believe that AI is a generational technology which will underpin many, many things,” he says. 

For Hiscox, this means having a multi-provider environment that offers access to different tools, models and capabilities. “We can blend these together as the different parts of the organisation advances,” says Loake.

The principles that underpin the AI strategy are based on flexibility. “You’re going to have to place some bets and be flexible along the way,” he says. “Your operating model, your architecture, your team must also be set up and optimised for agility and the ability to pivot when new things come along.”

One of the limits of off-the-shelf AI systems is that they are commoditised, which means every business has access to the same functionality. This can make it harder to argue the case that the AI system is delivering a business differentiator.

Drawing on email as an analogy, Loake notes that a commoditised AI system is similar to email, an application every business uses. Considering email as a business differentiator, he says: “What’s my customer journey? Where do I want to put in email to offer a very tactile or empathetic moment in the customer journey? For us, the claims process is the moment of truth within an assurance journey, and it’s super important that we’re there for our customers offering a highly empathetic service. You’re probably not going to do that over email.”

But the business has considered how the claims process can be supported by an AI like M365 Copilot, and this use case differs compared with how other insurers – who may be more geared towards high volumes and low cost – would use AI. “They might use it in a completely different way to support their value proposition,” adds Loake. 

Takeaways from the pilot 

From the 300-user, six-month pilot, he says a number of distinct user traits were identified.

There were the people who were interested in giving M365 Copilot a go to solve a problem, and if it didn’t work, they would try something different. There was a lot of usage from this cohort of users, and, as Loake notes: “We saw that people who used it regularly also got way more value.” 

Some people were curious about the technology and were very interested in new ways of working, and would “keep chipping away” at M365 Copilot. Loake says these people tended to self-organise as AI champions. “We set up a Microsoft Teams channel where people could talk about Microsoft Copilot,” he says. 

The iterative process of trying the Microsoft AI on different problems and the regularity of usage drove up the value they got from M365 Copilot.  

The initial pilot showed that it could benefit people in certain types of job roles. “The tool is particularly good when people have to start things from a blank sheet of paper or when they need to consolidate large amounts of information,” says Loake.

He adds that it also proved useful for people who need to have lots of meetings and do not wish to attend all of them. Here, meeting summaries and action points from meetings are part of the functionality available in M365 Copilot. 

Hiscox found that some people had job roles where they did not attend many meetings, or they were not required to produce discussion documents or did not spend a lot of time analysing other people’s information. In those job roles, M365 Copilot was less useful. “There were some people who fell in the, ‘I used it; it didn’t quite work for me’ camp, and, ‘It’s rubbish’, who said they were not going to use it again,” says Loake.

An assessment of the pilot showed that 15% of the people who took part were saving an hour extra per day on their workload; 20% achieved half an hour of savings; a fifth were seeing 10 to 15 minutes; while a quarter of the people who piloted M365 Copilot reported savings of a few minutes of work time. 

The prospect that some people could achieve significant time savings led to Hiscox embarking on a wider roll-out of 1,000 users. Loake carried over the policy where users had to apply for a licence, as this improved the chances of them actually using the tool.

He found that the M365 Copilot champions became more important as more people began using the AI software, and this led to an adoption programme. “We put more structure around the M365 Copilot champions and ran an adoption programme, where we had drop-in sessions,” says Loake.

Hiscox put in place M365 Copilot champions across each department. As part of their job roles, they were tasked with helping other people to use the AI tool. Groups such as those in claims processing organised workshops. 

The firm has now rolled out M365 Copilot to over 3,000 employees across 14 countries. “Once, you’ve democratised tools like M365 Copilot and give them to everybody, some people will use them loads and get loads of value, but some will get limited value,” says Loake.

But for him, having an AI tool available to everyone in the business means people can “feel like they’re in the game”, which he points out is very important. “You do want your users to feel like they’ve got the consumer-type tools in their hands, and it’s important for the executives to recognise what’s happening in their business,” says Loake.

A more focused plan  

From a CIO strategy perspective, he urges IT leaders to make some big bets, but not “go all with AI”.

Having 1,000 proof of concept applications that never make productions is not the best approach. For Loake, it’s far better to focus on five, which can be rigorously tested and pushed out to the business. “You can build these really tangible use cases,” he says.

This, adds Loake, enables the business and IT chiefs to stay in the game while AI models evolve. Then, as the new models become stable and released, they can be deployed for new, tangible use cases using a platform that has the flexibility to adapt.