Nationwide warns returns from corporate AI are still hard to measure
Nationwide has warned that the financial returns from workplace AI remain difficult to quantify, even as the building society rolls out the technology to around 16,000 staff.
Around 1,000 employees across the combined Nationwide and Virgin Money finance team now have premium Microsoft 365 Copilot licences, while some 2,500 engineers are using Github Copilot.
Paul Ballard, Nationwide’s technology transformation director, told Morning Wire the building society could see staff using the technology more regularly and completing work faster, but had yet to establish exactly what that meant for its bottom line.
“The thing that we’ve not nailed yet… is what’s the value that we’re receiving from it?” he said. “We can see the value, but how does that translate?”
Nationwide joined Microsoft’s early access programme with around 300 users and had only around 1,000 premium Copilot licences at the start of this year, despite larger deployments being announced elsewhere.
“Every pound that we spend is a pound of our customers’,” Ballard said. “We were quite mindful that we really understand where the value is.”
Finance was among the first divisions to expand its use, alongside technology and retail operations.
Staff are using Copilot to analyse spreadsheets, review documents, draft emails and summarise meetings.
Nationwide has also previously used GPT-4 through Microsoft’s Azure OpenAI service to help generate customer letters, cutting the average time taken from around 45 minutes to between 10 and 15 minutes. Staff review the output before letters are sent.
Ballard claimed the rollout was aimed at increasing the productivity of Nationwide’s existing workforce rather than reducing headcount.
“We very much see this as giving the existing workforce better tools so they produce more output for our customers,” he said.
‘Nowhere near’ AI agents
Their building society is now looking at AI agents for areas like anti-money laundering, customer due diligence, business analysis and mortgage operations.
But Ballard said Nationwide was “nowhere near autonomous agents”, with human oversight remaining in place across its AI use cases.
“The biggest risk is complacency with this stuff,” he said. “If it works 99 times, brilliant. If the hundredth time it doesn’t, how do you make sure you spot it?”
Nationwide has set up an AI council spanning technology, security, legal and procurement to assess new uses of the technology. It has also consolidated 19 separate data stores onto Microsoft Azure and Azure Databricks.
Its expansion comes as Microsoft looks to turn growing corporate use of AI into returns on the billions it is spending on infrastructure.
Microsoft 365 Copilot passed 30m paid seats in its latest quarter, while the number of enterprise customers with more than 50,000 seats increased more than seven-fold year on year.
Charles Lamanna, Microsoft’s executive vice president for Copilot, Agents and Platform, told Morning Wire that financial services was among the fastest sectors to adopt the technology, alongside technology and pharmaceuticals.
Microsoft’s revenue rose 18 per cent to $90bn (£67bn) in the three months to June, while Azure revenue jumped 43 per cent.
Competition for that corporate spending is also growing, with Google, OpenAI and Anthropic pushing their own workplace AI products and businesses increasingly able to choose between different underlying models.
Despite its extensive use of Microsoft, Nationwide does not intend to rely on a single AI model.
Ballard said the lender was looking at “different models for different use cases”, and wanted to retain “optionality” as costs and regulation changed.
Nationwide has also adopted what Ballard called a “no FOMO” approach to AI investment.
“AI isn’t the answer to everything,” he said. “You just need to be really mindful of what’s the opportunity that you’re trying to solve.”