CIOs face challenges turning AI ambition into measurable value, roundtable finds

Trust, control, governance and sovereignty concerns delaying AI adoption

Image:
Delegates at roundtable discussion

A Lenovo and Intel roundtable heard that enterprises must build two-way learning cultures and stronger governance if they want to turn AI pilots into lasting productivity gains

Enterprise IT is under pressure to innovate with AI, but a recent roundtable, sponsored by Lenovo and Intel, found that the path to greater productivity and a reshaped workforce is, in many organisations, obscured by a fog of questions and concerns.

Around a third of the CIOs and CTOs from a range of industries, many of which were subject to strong regulation, with finance, the public sector, critical national infrastructure and medical technology all represented, said they saw a gap between the promises of what AI can deliver, and the reality in their organisations.

“Maybe call me a bit of a sceptic,” said one Head of IT from the diagnostics industry, but I’ve yet to see any true, real value around the use of AI. I still sort of believe it's a bit of a buzzword.” His concern wasn’t just about maturity. It was about trust and control, especially around intellectual property and data governance. “How do you police AI?” he asked. “The implications of losing your IP publicly could be massive.”

That caution was echoed by others. A delegate from the public sector described the dual challenge of defining governance while proving value in real-world workflows. Local government, he said, is wrestling with the promise that “AI will save us money” versus the practical reality of implementation.

Some of these organisations pilot projects such as transcription tools that free up employees to focus “on people rather than paperwork”, show clear benefits. But scaling that into predictive interventions, such as identifying residents at risk of homelessness from council-tax data, raises difficult ethical and legal questions. “How much do we trust a system that says we should intervene? And how do we justify spending public money now to avoid higher costs later?”

For one bank, the conversation has already shifted from pilots to production. The bank has developed several in-house foundational models, hosted in a private cloud environment with strict governance and human oversight.

“There’s no outside interaction from the bank with any external AI facility,” a CTO explained. Every AI-generated output requires human validation. Yet even within such control, the operational realities are complex: “We’re now dealing with token-spend management, model-execution costs, and training. Everyone must be enabled to use it appropriately.”

For this bank there is also a social responsibility dimension. “The bank has a strong view on how to adopt technology without just removing people. It’s about looking at how we can use that technology to enhance, rather than replace people.”

Defining efficiency and finding skills

The cost question looms large across sectors but some around the table were starting to see some serious wins in terms of productivity. A bank had seen, from code development assistance, roughly a 40% increase in efficiency.

However, an IT Director of a global business pointed out the difficulty of trying to quantify ROI on AI investment saying:

“How do you put a number on the ability of Copilot to bring me up to speed with a meeting if I’m late? Or translate some emails for me?”

However, the cost benefit calculation is unique to each organisation and as one IT leader put it, “Once you start using agentic AI frameworks, the costs skyrocket. So where do we find the benefits? There’s no point giving everyone a Copilot license.”

The question of how best to help employees use the tools available to them was also discussed at length, with most focusing on carrots rather than sticks. Nobody was keen on the idea of mandating AI use in the way some Big Tech employers have.

“You’re not going to be replaced by AI,” one delegate warned, “but you might be replaced by someone who understands AI when you don’t.” That message is especially relevant for IT leaders rethinking skills pipelines.

Several delegates stressed the importance of nurturing younger technologists.

“We’re going to struggle if we stop employing juniors,” said one, whilst another noted that younger generations “are more able to grasp new paradigms” and have fewer anxieties about AI than many of those potentially employing them.

A delegate representing a global business said of younger Gen Z employees: “What we see as guard rails they see as barriers.”

Many businesses are presenting a confused message on the topic because they haven’t bought their own businesses up to date. As one bank CTO explained, “When we interview new recruits, we frown on them using AI in an interview process and yet the first thing we do if we employ them is to inject AI into their working lives. Should we not be testing them on using AI or the appropriate use of it to discern how they can us technology?”

The most successful IT organisations, they agreed, will foster a two-way learning culture where seasoned professionals bring experience and younger talent brings instinctive digital fluency which their older colleagues may lack.

Infrastructure and sovereignty

The question of sovereignty also arose during the discussion. One delegate pointed out that there was more to sovereignty than data.

“Sovereignty isn’t just about data,” he said, “it’s about whether you can own, operate, and replace the technologies you need. I can build my datacentre so I can keep my data and my processing local, but ultimately, I have to rely on the technology that's provided by a non-sovereign state to me on a commercial basis.”

For many around the table, hybrid models and private-cloud partnerships are emerging as pragmatic middle grounds between control and agility.