Capita deploys Microsoft Copilot to tackle UK civil service pension backlog crisis

Capita is turning to Microsoft Copilot in an attempt to rescue the UK Civil Service Pensions Scheme (CSPS) from a mounting backlog that has left thousands of civil servants facing payment delays and service disruptions.

The outsourcing giant took over the £239 million contract to manage the pensions of approximately 1.5 million current and former public servants in December last year. Almost immediately, scheme members began reporting problems with the new portal, including unrecognized passwords, broken links, and an interface that appeared unfinished and untested.

By January, Capita and the Cabinet Office were forced to apologize, acknowledging they were managing an inherited backlog of 86,000 cases, "a significant proportion of which was already overdue." The situation has been described as a "fiasco" by some observers, with retired civil servants reporting slashed incomes after payments failed to arrive.

Speaking to Parliament's Public Accounts Committee last week, Chris Clements, managing director of Capita Public Services, outlined how the company is deploying Microsoft's AI assistant to address the crisis. The technology is being used to scan initial contact forms and examine case documents, with the goal of prioritizing urgent cases and improving productivity.

"Caseworkers get a copilot [to] read all the attachments and documents for a case, and that gives them a summary at the start," Clements explained. "You look at the end-to-end handling time of a case, and you go through and you evolve and improve every step along the way. That's the process that we are starting to undergo now [and is a] core part of driving productivity."

The AI system is designed to automatically read and understand incoming website messages, identify priority cases, and place them in the work queue. Clements recommended that scheme members use the online contact form rather than calling, as this allows the AI to immediately prioritize cases based on urgency.

"Copilot automatically reads and understands, identifies priority cases, and puts them in the work queue," he said. "By doing it that way, the AI tool allows us to make sure we are instantaneously identifying those cases with the most detriment. That is a completely different solution from sitting in an unread email inbox, and it is part of the transformation we are making."

The scale of the problem became apparent when call volumes surged. While Capita expected to handle around 7,000 calls per week, peak levels reached 25,000 in a single week, overwhelming the system and staff.

This high-profile deployment of AI to solve a critical public service problem raises interesting questions about the technology's capabilities and limitations. Large language models like those powering Microsoft Copilot are trained on internet text rather than embodied human language, leading to ongoing debate about whether they truly "understand" the information they process.

For civil servants waiting for their pensions or information about their investments, the effectiveness of this AI-driven solution will be the ultimate test. The situation highlights both the potential and the risks of deploying AI in critical public services, where errors or misunderstandings can have immediate and significant impacts on people's lives.

Capita's experience may serve as a case study for other organizations considering AI solutions for complex administrative challenges, particularly in sectors where accuracy and timeliness are paramount. The outcome could influence how public sector organizations approach AI adoption for similar operational challenges in the future.

As the company works to resolve the backlog and restore normal service, the effectiveness of Microsoft Copilot in this high-stakes environment will be closely watched by both the pension scheme members affected and other organizations considering similar AI deployments for critical business processes.