Somewhere in an American hospital this week, a clinician requested authorisation for a scan, a procedure, an admission, and waited to find out whether an algorithm would agree it was necessary. That wait, and the administrative apparatus built around it, is the entire premise of the deal announced by R1 on 18 August, in which the revenue management giant agreed to acquire Humata Health and fold its AI-driven prior authorisation technology into a platform called Phare OS. The pitch is seductive in its simplicity: a 96 per cent first-pass approval rate, staff touches cut by 45 per cent, rescheduled appointments down by 83 per cent. Friction, the argument runs, is the enemy of good care, and machines are better at removing it than people.

Britain does not have prior authorisation in the American sense, and it would be a mistake to import the term wholesale into an NHS context. There is no insurer standing between clinician and patient deciding whether treatment will be paid for. But there is a structural cousin operating quietly across English general practice, and it has just been the subject of an uncomfortable interim finding from the Health Services Safety Investigations Body. Advice and Guidance, the mechanism by which GPs seek specialist input before or instead of referring a patient to hospital, was introduced to sharpen communication between primary and secondary care and to keep unnecessary referrals out of an overstretched system. HSSIB's investigation found it has also, in a number of documented cases, delayed or contributed to missed cancer diagnoses, largely where local systems required GPs to route through A&G rather than refer directly, or where requests for specialist assessment were declined despite continuing clinical concern.

The parallel with R1's acquisition is not that Humata's technology will ever touch an NHS referral pathway. It is that both systems are answers to the same underlying pressure: too many requests for scarce specialist capacity, and a belief that intelligent automation can adjudicate which of them are urgent without slowing everything down. R1's own marketing describes what its technology optimises for with unusual candour, fewer staff touches, fewer rescheduled appointments, faster approval. Those are throughput metrics. They say nothing about whether the requests that get declined, or downgraded, or redirected, were the right ones to decline. HSSIB's report makes essentially the same point about A&G in more measured language, noting that where the service lacked digital safety checks and standardised escalation routes, clinicians with continuing concerns had no reliable way to override the system's default.

This matters for NHS leadership beyond the immediate safety findings. The service is under real pressure to expand automated triage and demand management ahead of the October Budget, where productivity gains are expected to do much of the work that additional funding cannot. Ambient voice technology, AI-supported coding, and now moves toward wider algorithmic gatekeeping of referrals are all being pursued for legitimate reasons, and A&G itself, HSSIB was careful to note, delivers real benefit where it is properly resourced and monitored. The Royal College of Physicians' own survey found that seven in ten physicians sometimes or often receive A&G requests they consider clinically inappropriate for that route, and that the majority have had no formal training in using it. That is not evidence against automation or advice-based triage as such. It is evidence that the safety architecture around these systems, the audit trails, the override mechanisms, the incident reporting, tends to be built after the efficiency case has already been made and the rollout is already under way.

R1's Humata deal will be read by NHS digital and life sciences leaders as a signal of where agentic AI in administrative healthcare is heading, and it should be. But the more useful reading is cautionary rather than aspirational. The industry's own success metrics for authorisation automation are efficiency metrics. The NHS has just been reminded, in its own more modest system, that efficiency and safety are measured differently, and that the gap between them tends to surface only after the harm has already occurred.