Google customers turn away from agents and towards CX at Cloud London
AI can achieve much more than nebulous efficiency gains; CIOs are using it for real customer value
The hype about agentic AI is fading and CIOs are starting to look at practical, customer-facing use cases for the technology.
Ten months ago, Google’s Cloud Summit in London was agents, agents everywhere. This year AI still dominated, but talk about agents was notably absent.
That’s not because the tech hasn’t developed since then, far from it. But AI agents have less to prove in 2025, having moved from proof of concept to production – at least, so Google would have us believe (although the company’s own staff aren’t entirely on board).
“Two years ago, we were in this purgatory of POCs,” Google Cloud’s president of customer experience, Hayete Gallot, told me. Now, things have “matured a lot.”
That may be why CIOs were keen to talk about other uses of AI at the summit, which revolved around not just operational efficiency, but customer value.
“I sit on a lot of panels and industry events, and all of the focus seemed to be on the bottom line, operational efficiency and productivity gains,” said Starling Bank CIO Harriet Rees, who was taking part in a customer panel. "I wanted to see the same transformation that we've seen in the rest of our lives when it comes to customer experience.”
Starling was already using AI in back-office functions, like call handling, but last month it launched its first customer-facing generative AI application, Spending Intelligence. Now, customers can use natura language to query their spending (“How much did I spend at Greggs last month?").
“I can tell you that you might not like the answers to all of the questions that you ask,” Harriet quipped. Despite that, she sees AI as a “mainstay” to improving financial literacy in the UK, which is “very low.” She expects company AI adoption to not just speed up, but to become the default expectation from customers.
“We have to remember that the world we are living in now is one in which people are using tools like Gemini in their everyday lives. We've evolved the search world. People are no longer searching through 10 or so links before they get to an answer; they're asking an intentional question, and they're being given the answer to that question. I think we have to accept that customers will expect this in all aspects of their life.”
Morrisons CDO Peter Laflin also sees AI as a customer-pleasing tool. The supermarket chain has just launched an addition to its MoreCard app, which Peter demonstrated to me after the panel, called Product Finder. Customers can now find exactly where specific products are located in their local store using natural language.
“We can now confidently answer customer questions...even for those of us who aren’t in that store every day. This was about trying to make sure that that consistent experience was there, regardless of who you ask.
“We’ve only been able to deliver that solution because Google’s Vertex and Gemini models have taken away some of the pain of interpreting what customers are asking for.”
Peter also talked about Google’s integrated stack, in essence a single product, which makes it “so much easier” to deploy quickly.
Other panellists were also pleased with the single platform approach. Rob Passmore is the CEO of Additive Catchments, which is building a national civic infrastructure to monitor the health of the UK’s environment, starting with its rivers.
“The main difference [Google offers versus other hyperscalers] is the integration. It's a single product...and you interact with it as a single product.
“We are looking to create civic infrastructure that governments, utilities, investors and customers are going to depend on the data for; it's a very high bar that you need to meet, and a single, integrated product was absolutely what we're after.”
The combination of AI and the Google stack “has allowed us to move from concepts to proof of concepts, and prove ideas very quickly,” said Capgemini CTIO Steven Webb.
“The game changer for me is the power of something like a Gemini Vertex [AI] stack, and the ability to take a concept and develop, prove and test that with a board at lightning speed.”
Capgemini has been a Google Cloud partner for 10 years and recently worked on a project with the Imperial War Museum to transcribe 20,000 hours of audio recordings using AI (watch for our coverage soon).
Managing risk
There’s a bit of an ‘eggs in one basket’ situation here, but the risk is only really critical if “somebody pulls the plug out at Google,” said Peter (unlikely at best). The benefits, overall, far outweigh the risks.
Panel host Hayete acknowledged that being locked into one vendor for your AI strategy would be “crazy.” For that reason, Google does not lock customers into only using its own models and agents.
“We are very committed to that, because we understand that if we lock customers in, they'll be limited to the innovation we can bring.”
Harriet of Starling Bank said the “plethora of models” available in Google’s model garden, and ability to use a non-Google models deployed in the Vertex AI pipeline, was “key” to the companies’ partnership.
“Any CIO will know that resilience is something that we think about all the time. So, it also allows us to have a primary model and then a secondary model that we pull in in case of failure.
“That's very much part of the reason that we chose to go to go with Google: in particular, the fact that open-source models, which are developing at pace and evolving in a really quite impressive manner, are available for us to road test.”
While agentic AI has been the big news for 12 months or more, research shows that its use doesn’t match the hype. It makes sense, then, that IT leaders seeking return on all that AI investment are increasingly turning to the foundation of every company’s success: the customer experience.