May you live in interesting times, goes the Chinese curse – one that the tech sector would be wise to remember. Two years after the release of ChatGPT, the AI boom that buoyed the share prices of Microsoft, Google and Nvidia and launched a thousand startups now seems to be running out of steam. The blind embrace of all things generative AI by CIOs of companies large and small appears to have given way to an overriding feeling of scepticism about the once-promising technology. Yes, generative AI applications can help resolve customer service inquiries, generate simple summaries of complex legal documents and help defend an organisation against data breaches – but how that translates into concrete ROIs remains to be seen.
Steven Webb is confident that the answer will not only arrive soon but prove that many investments in generative AI applications were almost certainly worth the money. As chief technology officer for the UK branch of Capgemini, Webb works regularly with a diverse range of organisations to implement their own generative AI solutions. While he concedes that attitudes about AI are shifting from hype to something more nuanced, Webb argues that there’s still a great deal of road to run with such applications – especially when CIOs embrace a coherent data strategy from the outset.
Tech Monitor: Recent reports suggest that CIOs’ interest in generative AI is declining. Do you also see that among Capgemini’s clients?
Steven Webb: I think I’d start by saying, look: generative AI is an emerging technology. And, I think we’re seeing a very similar pattern emerge to that of all new tech, which is that there’s a natural hype cycle and there’s a level of experimentation that goes with that. Now I think we’re beginning to move beyond pure hype.
But I don’t see a decline or a drop-off or whatever term you would choose to use at all. I think there’s just that resetting of expectations. I think if I look at the customers that we work with, you can see that many of them are actually increasing their investment in generative AI. Specifically, about 31%of organisations that we recently polled said that they’ve increased their generative AI capability across all functions of their business.
That was at 3% in 2023.
To me, that just starts to show how much the technology is proliferating across all departments in all organisations. Where I find it gets really interesting is when companies move beyond that initial proof of concept and begin to register impacts at a more strategic level.
And that’s where a lot of organisations are beginning to realise that for generative AI to really work for them, they need a smart data strategy. We saw that 35% of organisations now believe that they’re going to have to rethink their business models to properly harness this technology. That was around 15% in 2023.
I also think we’ve seen organisations that have absolutely seen both productivity gains and operational gains. In that poll we conducted, 8.1% of respondents said that they’d seen increased productivity, while 6.3% had seen operational increases, too.
TM: What principles should your average CIO abide by, then, to ensure a successful generative AI deployment in their organisation?
SW: First, they need to think about guardrails. For example, do you have a good understanding of the ethical dimensions of this deployment? Are you able to guarantee that the generative AI application you’re about to use can’t be abused by staff? And how does that relate back to the overall governance principles within your organisation?
Obviously you then need a data strategy, which begins with a comprehensive understanding of the size and provenance of the enterprise data that the application is using to make its decisions. Do you, as your company’s CIO, know how to utilise that data in the right way?
And in those organisations moving beyond POC, many are at the stage where they’re evaluating how well a given generative AI platform can be tailored to very niche applications. That might mean choosing a language model that’s a lot smaller than the LLMs they were tinkering with at the experimental stage of their journey – one more capable of breaking down a single problem really well instead of an alternative that can do many things to a fairly average standard.
The environmental sustainability and the cybersecurity, too, of these platforms are becoming big concerns – IP leakage, for example, is a big worry. These are all things that we’re seeing CIOs mulling over.
TM: And what kinds of generative AI deployments is Capgemini helping to realise right now?
SW: We’re delivering projects in all shapes and sizes. That goes from working on deployments designed to deliver productivity improvements to rolling out chatbots in customer service centres and, at the other end of the spectrum, helping to solve really complex problems in medical biology.
One project we’re particularly proud of is our collaboration with a global consumer goods company, where we helped them design a model capable of taking individual ingredients and suggesting new and interesting recipes. Another involves Heathrow Airport, and its ongoing work to improve the customer service journey using AI.
TM: In previous interviews, you’ve expressed your personal enthusiasm for the so-called ‘metaverse,’ a concept that’s dropped out of mainstream tech discourse of late. What’s the world been missing about the potential of the metaverse in business?
SW: Yes, specifically in business – much more than what we’ve seen on the consumer side of things. In the former, the reality of what we can achieve through digital twins and obviously what we can do within that business metaverse construct, I think, is hugely powerful.
We’re actually doing a lot of work at Capgemini in this space, now. After all, the utility of a digital twin of, say, an airport or a production line, and precisely mapping the flow of people or goods through that physical space, is really obvious. For me, that starts to get even more interesting when you layer AI over that, and use that technology you automatically analyse or summarise or re-test the results produced by that digital twin.
Read more: As AI transforms cybersecurity, Cisco’s Martin Lee has only one piece of advice for IT managers: expect the unexpected
