Tech leaders share ideas on turning AI promise into real-world value

Because there’s more to AI than Copilot

At a roundtable, sponsored by Lenovo and Nvidia, technology leaders debated how to move beyond pilots to AI systems that generate real business value — balancing cost, ethics, and the human layer of trust.

As generative AI moves from experimentation to enterprise adoption, IT leaders are confronting the challenge of building infrastructure that doesn’t just amount to a few Copilot licences but delivers measurable business impact.

At a recent Computing roundtable, sponsored by Lenovo and Nvidia, technology leaders from many sectors shared their experiences in turning AI promise into production value — navigating legacy data, ethics, sovereignty, and cost constraints along the way.

For many organisations, the first barrier to AI value creation is data — its quality, accessibility, and relevance. One CIO from a financial consultancy captured this challenge:

“We have 35 years’ worth of client data — documents, contracts, proposals, work orders — all unstructured. For the first time, we have the technology to root out information from diagrams, contracts, even handwritten notes. But then you must ask how we make that data relevant and useful and for today’s compliance rules and laws?”

This sentiment resonated across the public sector as well, with one government representative noting similar struggles: ensuring data is not only useful but also transparent and explainable to build public trust.

“We are trying to make AI visible and explainable so that people can really trust it,” he said.

Some organisations are more mature, with representatives of even quiet conservative, risk averse financial institutions explain how they were extracting value from generative AI tooling.

“Copilot is embedded in our workflows,” explained one, “meetings, transcripts, minutes, actions we’re seeing real acceleration in. We’re also looking at project management, analysing incidents across the technical estate, and improving the software development lifecycle.”

A colleague continued:

“We pump deployment and incident data into an AI model to see themes such as was the issue in design, testing, or release? That accelerates how fast we can find and fix root causes.”

Nonetheless, human oversight remains crucial at this institution.

“You always need humans to validate AI outputs. Copilot assists workloads, but we always keep human intervention to check results.

Ethics, sovereignty, and the human layer

For the Head of IT Operations at a charity, the focus is on using AI responsibly while staying aligned with ethical and sustainability principles.

“We’re asking, what AI tools can we use ethically? How can they deliver real value without compromising our mission?”

In terms of infrastructure, data sovereignty is also a pressing issue. “Our data is stored in North Europe, but with a US company. With the current political climate, that’s not necessarily reassuring. We’re exploring how to back up and protect data outside those services.”

Nonetheless, this person didn’t consider themselves to be a sceptic.

“I’m not an AI sceptic; I’m an AI realist. It’s cool technology, but not inherently transformative like the internet. Maybe as transformative as smartphones. It has power, but also real limitations that people are only starting to understand.”

“It’s people, process, then technology. AI won’t help if you don’t understand your business, clean your data, and fix your processes first.”

Balancing Cost, Scale, and Value

For some leaders, cost remains the biggest inhibitor to scaling AI. The CTO at a hedge fund it put it bluntly:

“I’m a skinflint — I want it cheap. We’ve had quotes for hundreds of thousands of pounds, and honestly, I’m not seeing the result yet. Until I do, I’m waiting.”

In contrast, other organisations are already scaling carefully targeted prototypes. A CTO in another investment business described a phased approach:

“We’re implementing AI in compliance, marketing, sales in small prototypes producing very good results. At the core, we’ve used machine learning for years in quantitative analysis, but I’m especially positive about agentic AI. In the next three or four years, we’ll probably name them as we do humans.”

Ultimately, the consensus across the roundtable was clear: AI-ready infrastructure isn’t about hardware or algorithms alone. It’s about alignment of people, processes, ethics, and data - from the core to the edge of the enterprise. Organisations that succeed will be those that invest not just in AI tools but in the trust, governance, and clarity required to turn them into real business value.