Swillklitch/iStock/Getty Images Plus via Getty Images
Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- Successful AI projects are built on strong foundations.
- Find the value, measure it, and discuss aims with your peers.
- Take an iterative approach to build confidence and deliver results.
Turning what seems like a good idea for an AI system into a practical service is proving to be a tough challenge for many professionals.
Research suggests most AI projects get stuck at the starting blocks, and even the initiatives that progress down the track don’t necessarily deliver great results.
Also: Dreading AI job cuts? 5 ways to future-proof your career – before it’s too late
Paul Neville, director of digital, data, and technology at The Pensions Regulator (TPR) in the UK, is eager to ensure his agency turns AI into a competitive advantage. Here, he shares his five best-practice lessons from exploiting AI with ZDNET.
1. Start with foundations
Neville said the starting point for exploring AI successfully is the foundation underlying your emerging technology initiatives.
Late last year, he launched a Digital, Data, and Technology Strategy, a series of missions in a five-year plan to renew TPR’s capabilities.
In March this year, he initiated the data component of the strategy, which outlines a collaborative plan to drive adoption of new data technologies and standards.
Also: 5 ways to prevent your AI strategy from going bust
In addition to cybersecurity and data governance, his team has focused on service management projects that help rationalize TPR’s application estate.
“If you don’t have your data in good order, that’s a problem,” he said. “Of course, there are AI tools to help you do that. But ultimately, strong foundations are about good practice, good governance, and data ownership.”
2. Focus on end users
Neville said AI success is directly related to your ability to exploit the technology for the benefit of the people who work in your organization.
Neville: “User focus is absolutely critical to everything you do.”
The Pensions Regulator
“User focus is absolutely critical to everything you do,” he said. “You’re not here for the tool, you’re here to deliver value. So, find that value, measure it, and talk about it.”
Case management is a good example. When joining TPR in late 2023, Neville inherited a situation where cases were managed on spreadsheets or via spot technology solutions.
Also: 5 ways to feed your AI the right business data — and get gold dust, not garbage back
His team implemented Microsoft Dynamics 365 as an integrated case management system, which demonstrated the benefits of automation and the potential for applying other AI tools and data science techniques.
“When you think about automation, you should automate the simpler things,” he said. “You want talented human beings working on the outliers, the difficult cases, and that’s the way we do it. Our approach to automation helps us to be much more joined-up.”
3. Integrate with the business
Neville said that understanding the demands of the people who work across your organization means you must act as part of the wider enterprise.
“In the old days, IT teams were like a service team that was just told what to do. Now, IT teams are part of the business,” he said.
Also: 92% of young professionals say AI boosts their confidence at work – how they use it
As a member of TPR’s C-suite, Neville ensures that other executives understand how digital, data, and technology can help drive internal change and transform the pensions industry more generally.
“We need to be integrated with great business sponsorship as part of a single team across the organization,” he said. “That’s certainly what we’ve tried to do, and that really makes the difference.”
4. Develop a learning culture
Even once an aim is identified and targeted with AI, IT professionals shouldn’t sit back and assume the job is done.
Neville said successful digital leaders establish a culture that ensures people in their team work iteratively, helping to spread the benefits of new ways of working beyond the confines of the IT department.
Also: 3 ways AI agents will make your job unrecognizable in the next few years
For TPR, this agile, product-centered approach means the organization focuses on developing reusable capabilities for flexible services in key areas related to pensions governance. Neville said other professionals can benefit from an iterative approach to AI projects.
“It’s about having that learning culture, where you’re agile, measuring, and improving along the way,” he said.
“The whole point of iterative work is that it builds confidence in the organization. People you work with start to believe you can drive change. Agile approaches help to build confidence.”
5. Tell great stories
Neville said the final component is storytelling, which he described as a big part of his job: “The language you use to talk about AI is so important.”
His organization is always on the lookout for new ways to tell effective stories. Neville recently launched the Pensions Data and Digital Working Group, which aims to ensure that TFP and the pension industry create a joined-up narrative for digitalization and automation.
Also: Enterprises are not prepared for a world of malicious AI agents
This working group has 15 cross-sector members, including trustees, actuaries, lawyers, and technologists. The aim is to bring people together to consider how emerging technology can solve intractable challenges.
“To do all these amazing things, the industry itself needs to change, and we need to work with the industry to do that,” he said. “However, we want to incentivize people to digitize and take advantage of new technology. So, instead of being a regulator that imposes things on the industry, we agreed to set up an official working group for innovation.”
Delivering benefits from AI
Neville said his organization is always looking for new opportunities to exploit emerging technology. He detailed two emerging AI use cases for ZDNET.
First, using AI to assess pension schemes. TPR’s relationship teams have regular conversations with scheme providers to analyze potential risks. It’s an effective but manually intensive process, and Neville saw AI could provide a helping hand.
Also: Is your company spending big on new tech? Here are 5 ways to prove it’s paying off
“We wanted to ensure we don’t miss anything,” he said. “So, we have a tool that calls all the news sites across the internet and then links information it finds with the pension scheme. If there’s a scheme in distress, we can flag it as a risk and find out what’s happening.”
The second use case is based on OpenAI and Azure technology. Neville said the companies his organization works with must submit climate-related statements to comply with legislation. Once again, sifting through all those statements involves tedious work.
(Disclosure: Ziff Davis, ZDNET’s parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
“It’s certainly very time-consuming, and we weren’t getting through all the information,” he said. “So, we’ve deployed gen AI to read all this information and turn it into analysis, and then we intervene where we think there is a requirement.”
Also: Microsoft Copilot AI can now pull information directly from Outlook, Gmail, and other apps
Other AI-enabled projects are at various stages of gestation. TPR is rolling out Copilot internally, and the technology is being tested across roles, including for developers. He painted a picture of the data-enabled organization two years from now.
“If you work in TPR, the systems that manage your everyday work will feel joined up,” he said. “You won’t have to do too much manual work, and the work you do will be value-adding.”
Neville said this approach will extend beyond the enterprise firewall to the organizations and people TPR engages with: “We’ll be much more streamlined and will, therefore, be easier to work with. Because if you’ve told us something, we’ll know it.”
