Add us as a preferred source
Rachel Finegold, founder and chief executive of Frontier Health, argues that the NHS should use investment in AI which can make the systems already in place work harder.
Ask most people what’s wrong with the NHS, and they’ll point to waiting lists, understaffed wards, or poor communications around when you’ll be seen, or when you can leave. Spend enough time inside hospitals, as I have, and you’ll find that all of these complaints share a common source: administration that simply cannot keep pace with demand.
The staff who keep patients moving through care pathways are doing the job with tools that haven’t meaningfully changed in years. They are tracking referrals, chasing results, booking appointments, and updating records across multiple ageing, disconnected systems.
When administration breaks down – and it does, all the time – crucial information never reaches the right person, tasks don’t happen at the right time, and patients wait longer for care.
Most damaging are missed follow-ups and slow escalations, which have a pronounced impact on patient outcomes. Almost 30% of cancer patients breach the 62-day target to start treatment following referral. A four-week delay in that window raises the risk of death by 10%. Yet I’ve seen hard-working staff in cancer care manually managing 300+ patients at once on rudimentary spreadsheets.
These administrative bottlenecks exist across the NHS. A patient can be clinically ready to move forward, but still be waiting because the system hasn’t been updated. The waiting, the chasing, and all too often the complete silence patients experience is a reflection of an under-resourced administrative function doing its best with what it has.
Broken administration won’t fix itself
Non-clinical hiring is frozen across much of the NHS, so simply recruiting more staff isn’t a realistic answer to the capacity problem. Operations managers are too busy firefighting breaches in real time to make inroads into the root causes, while fragmented systems make it almost impossible for them to understand where spare capacity is and what bottlenecks are slowing patient flow.
Most Trusts have accumulated layers of software over years, with different systems used by different departments for different parts of the patient journey. The staff who understand these processes already work under enormous pressure, so there is little spare capacity for them to redesign workflows and migrate systems. Equally, the NHS can’t afford to just rebuild the administrative layer from scratch. The cost, disruption, and implementation time would be enormous.
Much has been written about where AI might play a meaningful role in solving some of these longstanding problems. I’d argue that what the NHS needs from any AI solution is three-fold.
Firstly, can it work with the existing infrastructure without the need for major 6-12-month integration projects just to get the tech into anyone’s hands?
Secondly, can a typical member of the administrative workforce be easily trained to operate and oversee the technology safely?
Thirdly, will it deliver measurable near-term benefits – productivity gains, cost reductions, incidents reduced, waiting lists shortened – as well as harder to measure, but nevertheless desirable goals such as improved morale and team wellbeing?
Or to put it more bluntly, the NHS needs solutions to longstanding problems that simply work, and with near-immediate effect.
Building AI to resolve bottlenecks
AI should do more than just speed up existing processes. Taking repetitive work off the desks of overloaded NHS staff won’t address the much larger administrative bottlenecks.
Take patient tracking. Rather than focusing on simply completing individual tasks as a human would, AI can coordinate care pathways in ways that are difficult to achieve when information and responsibility is fragmented across individual people and disconnected systems. A new orchestration layer that connects AI agents will increase workflow visibility across the entire hospital and enable agents to independently identify the next steps on a patient’s care pathway.
Going beyond individual workflows and using AI to manage the end-to-end process will unlock meaningful increases in capacity at a time when non-clinical hiring freezes are likely to remain in place for the foreseeable future.
AI can make these improvements to patient care pathways while keeping accountability and clinical decision-making with NHS staff. Trust in new AI tools should be built by ensuring every step an agent takes must be validated as it runs, every action logged and auditable, and if something doesn’t check out, it stops and flags it rather than guessing.
The importance of inside knowledge in shaping NHS AI
For all my on-the-ground experience within NHS Trusts, I knew that internal buy-in and collaboration were essential to get any new technology project off the ground. So from a very early stage of inception, we’ve been working with East Sussex Healthcare NHS Trust to design, test and iterate on our own technology in high-stakes administrative environments.
By working in close partnership with the Trust’s admin team, we were able to fully understand the nature of the work staff were already doing and identify where AI could safely take on repetitive tasks. The results, over just eight weeks, were seven months’ worth of work (221 days) saved on repetitive tasks and a 22% reduction in the median time patients spent on the pathway.
What I’ve learnt from the experience – and our subsequent conversations, deployments and partnerships with many other Trusts, is that investment in AI can make the systems already in place work harder and unlock capacity without requiring the NHS to rebuild from the ground up. The NHS spends a lot of energy debating what AI should do in healthcare. I’d start with the unglamorous work: keeping track of patients, so nobody gets lost in the system and experiences avoidable delays in their care.
