South Tyneside and Sunderland NHS Foundation Trust has deployed robotic process automation software to handle incoming GP referrals, cutting the time between a referral being made and it reaching a clinical team to within 48 hours. In many cases, that window is now 24 hours.

The results are measurable. According to national figures for March 2026, 77 per cent of patients at the trust are now receiving planned treatment within 18 weeks of a GP referral. The national average sits at 63.5 per cent. The gap is not marginal.

The 18-week referral-to-treatment standard has long been a central measure of NHS performance. Trusts across England have struggled to meet it consistently, particularly in the years following the pandemic. South Tyneside and Sunderland's figures suggest that automating one part of the administrative process has had a concrete effect on overall throughput.

The trust's RPA software operates as what it calls "digital workers." These are not AI systems making clinical decisions. They perform administrative tasks: receiving referrals, processing them, and routing them so they are ready for clinical review. The work previously required manual handling, which introduced delays. Removing that step has shortened the queue before a clinician even looks at a case.

RPA is part of a wider programme of digital investment at the trust. New appointment booking systems have been introduced to reduce the time between a referral being reviewed and a date being confirmed. Investment in diagnostic technology has also expanded the number of patients who can be seen for tests and scans, addressing a separate bottleneck that delays treatment decisions.

Alison King, the trust's Director of Performance, said that access to care remains the top priority for patients. "As we look to the future, we want patients to have greater certainty about when they will be seen and technology is key to this," she said. "Our automated digital workers are just one aspect of this." The framing is deliberate. The trust is not presenting automation as a solution in itself, but as one instrument within a longer-term effort to improve reliability for patients awaiting planned procedures.

The development sits within a broader pattern of NHS investment in automation and artificial intelligence. The UK government has backed a £500 million sovereign AI fund aimed at early-stage AI companies, offering access to supercomputing infrastructure with up to one million GPU hours per startup and early-stage investment of up to £20 million. The fund is designed to help UK providers scale and compete internationally. A further £282 million has been earmarked to support AI research and development, including the creation of new datasets.

At the clinical level, Barts Health NHS Trust has reported outcomes from a six-month AI pilot at Newham Hospital focused on patient complaints. Using tools within Microsoft Copilot, which is approved for NHS use, the trust reduced the time spent drafting complaint response letters by more than half. The number of cases needing to be reopened fell significantly. All responses were reviewed and approved by staff before being sent.

The question of scale is harder. Individual trusts have introduced automation on their own terms, with varying infrastructure, governance frameworks, and technical capacity. What works in Sunderland requires a degree of digital readiness that is not uniform across the health service. Staff training, procurement, and integration with existing systems all represent friction. The NHS has a long record of technology pilots that do not translate into consistent national practice.

For patients, the practical consequence of faster referral processing is straightforward. A referral that reaches a clinical team within a day, rather than sitting in an administrative queue for a week, means earlier assessment, earlier diagnosis, and earlier treatment. South Tyneside and Sunderland has demonstrated that the administrative stage of that journey can be shortened without clinical compromise. Whether other trusts follow, and how quickly, will depend on factors well beyond the software itself.