‘Using generative AI as a copilot is the sweet spot’: A look at Nationwide’s AI approach

When Nationwide Building Society – or the Southern Co-operative Permanent Building Society as it was originally known – arrived on the scene in 1884, it was part of a fast-growing member-owned mutual movement in the UK. And as Nationwide embarks on new experimentation with technology and the drive for growth in the 21^st century, it is doing so in recognition of these roots.

The current tech transformation at Nationwide, specifically its expanded use of generative AI in 2025, is being pursued in the interests of its circa 15 million members and 18,000-plus staff. It’s not using tech for tech’s sake – there is “always a human in the loop as a principle” as use-cases for generative AI deployment are identified, according to Nitin Kulkarni, chief information officer (CIO) for data platforms, engineering & AI centre of expertise (CoE).

The CIO joined Nationwide in July 2023 after nearly ten years at Barclays where he was latterly head of data platforms. He tells ITPro the society’s adoption of generative AI is centred on using it as an aid for staff not a people replacement.

“We are using it as a copilot as we believe this is the right sweet spot,” Kulkarni explains, adding that it follows a framework for using generative AI to ensure activities in this emerging field are controlled and secure.

“We are a building society and we want to make sure the right principles are in place, so everything we’re delivering to customers is abiding by those principles.”

According to Kulkarni, Nationwide’s ambition is to be seen as sector leading when it comes to data and AI strategy. Initially, though, the plan has been to take “baby steps”, test generative AI “end to end” on a case by case basis, and stay close to those guardrails.

Tech to empower employees

Nationwide announced in January it was expanding its use of generative AI with Microsoft to deliver even “better experiences for customers and colleagues”.

By using the GPT-4 model within Microsoft’s Azure OpenAI to help generate customer letters, Nationwide has reported a drop in response times from 45 to 10-15 minutes – which it calculates as an efficiency improvement of 66%.

The society is hopeful such back office use of the new tech tools available to them through their Microsoft partnership will give staff more time to deal with customer queries faster and allow them to better address more complex member queries.

Kulkarni stresses that none of this is possible without major data migration and centralization. Nationwide has had to put in the hard yards, using Microsoft Azure to bring all its data onto one platform, before applying open data analytics platform Azure Databricks and Teradata’s VantageCloud to analyse and maintain the information.

“For AI to be successful all data has to be in one place – that is a fundamental thing you have to do,” says Kulkarni.

The next big decision is selecting the right modelling for optimizing the use of that information.

“This is where Microsoft comes in,” he explains. “They have a very good ecosystem that makes sure models and prompts are validated as part of the process. Microsoft also provides a singular platform to do all these particular jobs.”

Nationwide employees now have better access to the information required to do their jobs. The work also supports the society’s ambition for greater personalisation of customer services and, importantly, it gives Nationwide a foundation for more Azure OpenAI applications in future.

What does that future look like?

Nationwide has already eyed its contact center, credit risk assessment, economic crime monitoring, its virtual assistant ‘Arti’, and CO2 emissions reporting as areas where generative AI might play a greater role in supporting staff.

Fundamentally, the aim is to improve service, and optimize decision making, risk management and operational efficiency. But for Kulkarni, another interesting area of is in helping manage legacy code and modernise the building blocks of tech infrastructure.

In 2024, for example, it was used as part of a GitHub Copilot launch which now has more than 850 active users in the business, and has reportedly empowered developers to deliver new code, and new customer experiences, faster.

But the data foundation is key. The CIO says: “This is in the interest of us as an organization because whatever strand of technology comes next, the need for data-driven decisions is going to be more important.”

Power and responsibility

Kulkarni says Nationwide is working with Microsoft to monitor the performance of AI.

This plays into using AI responsibly mantra that Kulkarni and the wider Nationwide leadership team have embedded into strategy. Both Microsoft and Nationwide have committed to a joint responsible AI framework while the society established an AI CoE, in association with IBM Consulting, to oversee and deliver all of its AI development.

There is an AI Council to ensure all AI applications conform to responsible principles, covering areas such as fairness, transparency, inclusiveness and security.

“It is a responsible body that vets each AI user case we look to implement,” Kulkarni says of the council which includes representatives from the risk, procurement, security architects, compliance, data protection, legal, and people teams.

Use of generative AI at Nationwide is deemed “organizational change”, so the wider business needs to be represented in its implementation, he explains.

According to Kulkarni, Nationwide has made good progress in testing generative AI since ChatGPT’s November 2022 launch “woke up the whole world to doing something with AI”.

“When generative AI started it was how you ask a question, and you get a response,” he notes.

“Enterprise applications need to be much more rigid, they can’t just rely upon prompts. The expansion we’re doing in this space is making sure we’re using the latest and greatest tech plans so we can scale to a position that was not possible in the past.”

He adds: “It’s such a new area and I’m sure new practices will unfold, but with agentic AI knocking on the door we need to continue challenging ourselves.”