The regulatory milestone that changes the competitive landscape
In October 2025, a regulatory decision was made that will have long-term consequences for physiotherapy practice in the UK — and for every professional working in allied health.
Flok Health received Class IIa Medical Device approval from the MHRA, making it the first artificial intelligence system in the UK — and Europe — authorised to deliver fully autonomous physiotherapy. (Digital Health)
“Fully autonomous” is the critical qualifier. This is not an AI that assists a human physiotherapist, generates reports for clinical review, or provides exercise guidance under supervision. It is a system approved to manage an entire care pathway — from initial assessment through treatment to discharge — without human physiotherapist supervision at any stage. (Digital Health)
This is a genuinely different category of AI. Most healthcare AI deployed to date sits in the assistive category — supporting clinicians, not replacing their oversight. Class IIa approval for autonomous clinical decision-making represents the frontier of AI in healthcare regulation.
The NHS pilot results
The regulatory approval would be significant on its own. The clinical outcomes data makes it urgent.
An NHS pilot of Flok Health at Cambridgeshire Community Services delivered the following results:
Back pain waiting lists more than halved during the pilot
2,500+ patients treated in 12 weeks
80% of patients rated their experience as equivalent or better than face-to-face physiotherapy
(Cambridgeshire Community Services NHS Trust)
The waiting list reduction is particularly striking. Musculoskeletal conditions — back pain, joint pain, and soft tissue injuries — are among the highest-volume referrals in NHS physiotherapy, and waiting times have been a persistent problem. Halving a waiting list in 12 weeks is an outcome that NHS commissioners will find difficult to ignore.
The 80% patient satisfaction equivalence figure also matters. It directly addresses the most common objection to autonomous AI in clinical care: that patients will feel short-changed by technology instead of people. A majority of patients found the AI-delivered care comparable to or better than face-to-face — a finding that shifts the debate.
What Flok Health actually does
Understanding Flok Health's clinical capability helps explain both its approval and its implications for private practitioners.
The system works through a digital interface (app or web) that:
Conducts a structured initial assessment through symptom questionnaires and movement screening
Generates a diagnosis and treatment plan using clinical decision algorithms
Delivers a personalised exercise programme with video guidance and progression tracking
Monitors recovery through regular check-in questions and adapts the programme accordingly
Makes discharge decisions based on recovery trajectory data
The pathway from assessment to discharge operates without a human physiotherapist reviewing individual patient decisions. The MHRA's Class IIa approval means the regulator is satisfied that this autonomous pathway is safe for musculoskeletal conditions within its defined clinical scope.
The clinical scope matters. Flok Health is approved for the conditions it was validated on — primarily common musculoskeletal presentations. It is not approved for complex neurological rehabilitation, post-surgical recovery requiring clinical judgment, or high-acuity presentations where autonomous assessment is unsafe. The regulatory boundaries define where the competitive threat lands.
The price and access disruption
For private physiotherapy practices, the competitive implications are straightforward to understand and difficult to absorb.
A typical private physiotherapy appointment in the UK costs £50–£90, with most conditions requiring 4–8 sessions. A complete musculoskeletal episode costs a self-paying patient £200–£720. An NHS referral is free but involves waiting times that Flok Health's pilot has demonstrated it can eliminate.
An AI physiotherapy service operating at Flok Health's scale faces marginal costs per patient far below what any human-staffed practice can achieve. If NHS commissioners deploy this model at scale — which the pilot results make plausible — the referral pathway for a significant volume of musculoskeletal presentations shifts to AI delivery.
For private practices, the relevant question is: what proportion of their private patient caseload involves the kind of uncomplicated musculoskeletal presentation that Flok Health can manage autonomously? If the answer is substantial — and in most MSK-focused practices, it likely is — the competitive landscape has changed materially.
What remains outside AI's autonomous reach
The Class IIa approval defines Flok Health's scope precisely, and that precision is as important for practitioners as the approval itself.
AI autonomous physiotherapy is not currently appropriate for:
Post-surgical rehabilitation: Cases where the treatment plan must be adapted in real time based on wound healing, hardware tolerance, or surgeon preference require clinical judgment that an autonomous AI cannot safely exercise.
Complex neurological presentations: Stroke rehabilitation, spinal cord injury recovery, and Parkinson's physiotherapy involve the kind of adaptive, observational clinical reasoning — watching how a patient moves, reading subtle signals of fatigue or pain — that current AI systems cannot replicate at the bedside.
Paediatric physiotherapy: Developmental assessment and therapeutic intervention in children requires specialist clinical knowledge and the ability to work with families in ways that autonomous AI is not equipped for.
Rare or novel presentations: AI clinical decision systems are trained on data. Unusual presentations, rare conditions, or atypical symptom combinations may fall outside their reliable operating envelope.
These categories represent significant practice areas for many physiotherapists — particularly those who have developed specialist expertise. The practices most disrupted by Flok Health are those primarily treating high-volume, lower-complexity MSK conditions. Specialists are better protected.
The MHRA regulatory sandbox and what comes next
Flok Health's approval is a single data point in a rapidly evolving regulatory landscape. The MHRA's AI Airlock regulatory sandbox — Phase 2 running until March 2026 — is designed to evaluate AI medical devices in real-world NHS settings before full approval, and has multiple AI health products progressing through it. (Digital Regulations Innovation Office)
The National Commission on AI Regulation in Healthcare, launched in September 2025, is developing a new regulatory framework due in 2026. (GOV.UK) This framework will address the governance questions that Class IIa approvals open: what oversight is required for autonomous AI clinical systems, how adverse events are reported and investigated, and how accountability is allocated when an AI clinical pathway produces a poor outcome.
The GMC's position is clear: clinicians remain responsible for all decisions when using AI. (iatroX) For autonomous AI systems where a clinician is not involved in individual patient decisions, this framing creates unresolved liability questions that the 2026 regulatory framework must address.
Strategic implications for private physiotherapy practices
The right response for private physiotherapy practices is not panic — but it is not complacency either.
Differentiation on complexity and relationship: Practices that have positioned themselves as specialists in complex rehabilitation, neurological physiotherapy, post-surgical recovery, or specific sports injury management have durable competitive advantages. Autonomous AI cannot compete here.
The human dimension as premium: 80% patient equivalence is not 100% patient preference. A meaningful minority of patients will always prefer human care — either from genuine therapeutic preference or from conditions where the relational dimension of treatment is clinically important. This segment supports premium pricing for high-touch physiotherapy.
Integration rather than competition: Some private practices are beginning to partner with digital physio platforms, using AI for initial assessment and home exercise monitoring while reserving clinic time for hands-on treatment, complex decision-making, and cases where human assessment adds clear value. This hybrid model is emerging as a viable middle path.
NHS partnership: Practices with strong NHS relationships should engage with their local commissioners about the integration of autonomous AI pathways. Understanding which patients are being triaged to AI delivery — and which remain in human-delivered pathways — is essential for planning.
Key statistics at a glance
Flok Health received Class IIa Medical Device approval — first autonomous AI physiotherapy in UK and Europe (Digital Health)
NHS Cambridgeshire pilot: back pain waiting lists more than halved (Cambridgeshire Community Services NHS Trust)
2,500+ patients treated in 12 weeks during pilot (Cambridgeshire Community Services NHS Trust)
80% of patients rated AI physiotherapy as equivalent or better than face-to-face care (Cambridgeshire Community Services NHS Trust)
MHRA AI Airlock Phase 2 running until March 2026 with multiple AI health products in evaluation (DRIO)
National Commission on AI Regulation in Healthcare new framework due 2026 (GOV.UK)
MarGen builds AI-visible content strategies for healthcare, legal, and financial services firms. Explore our approach.