GP Triage and Rapid Health Smart Triage both aim to move eligible patient requests from assessment to an appropriate booking without routine manual triage. The useful comparison is not which company uses more AI. It is how each service defines eligible requests, connects clinical urgency to local capacity and handles the patients who cannot complete the automated route.
Public documentation reviewed on 9 September 2026 gives enough information for an initial shortlist, but not a defensible universal winner. This comparison is published by iatroX and includes its clinical-reference service separately below. Ask-iatroX is not a substitute for either company's patient-access and booking software.
The important architectural difference
GP Triage's relationship with Infermedica makes its division of work relatively explicit. Infermedica supplies clinical assessment; GP Triage supplies the surrounding NHS access and operational workflow. That relationship was publicly announced on 9 September 2026 and was already described in GP Triage's April product update. GP Triage and Infermedica announcement
Rapid Health describes its own assessment and workflow platform operating within clinician-defined parameters. Its public material connects structured history-taking, urgency assessment, navigation and booking. The description establishes the advertised workflow, not a fully disclosed technical comparison of the two systems' inference methods. Rapid Health product information, reviewed 9 September 2026
A practice does not need to adjudicate proprietary architecture to start evaluating either service. It does need a clear description of the decisions being delegated, the configuration it controls and the exceptions it retains.
What the public material does and does not establish
The table summarises published descriptions available on 9 September 2026. It is a procurement starting point, not a hands-on test or confirmation that every capability is enabled in every practice.
| Decision point | GP Triage | Rapid Health Smart Triage |
|---|---|---|
| History collection | Adaptive assessment using Infermedica, with local workflow around it | Structured questioning within its assessment workflow |
| Urgency and navigation | Clinical output informs configured pathways | Urgency assessment linked to clinician-defined navigation rules |
| Appointment booking | Described as part of the practice-integrated journey | Described as part of the automated patient journey |
| Record-system integration | Public pages describe interoperability and clinical summaries; confirm the exact EMIS or SystmOne connection, permissions and writeback with the supplier | EMIS and TPP SystmOne integration explicitly described by the NHS Innovation Accelerator |
| Human involvement | April documentation describes flexibility between automated routing and inbox review | Supplier describes configurable automation and clinical review arrangements |
| Children and other exclusions | Obtain the deployed product's intended population and local handling rules | Published child pathways do not establish that every paediatric request receives autonomous clinical assessment |
| Operational evidence | Named supplier case studies with workload estimates | An externally conducted, commissioned implementation evaluation as well as supplier claims |
| Clinical safety assurance | Supplier describes DTAC, DSPT and DCB0129 work | Supplier describes DTAC, DSPT and DCB0129 work |
| Medical-device description | Infermedica MGP is described as EU MDR Class IIb; establish the host product's separate regulatory boundary | Supplier describes UKCA Class I and MHRA registration, with a separate EU MDR Class IIb assessment in progress |
Sources for the current feature descriptions are GP Triage's website, its 30 April 2026 update, Rapid Health's primary-care page, the NHS Innovation Accelerator profile and Infermedica's regulatory documentation.
An integration claim should trigger a demonstration in the buyer's environment. Ask to see the appointment created, the clinical summary written back, the booking cancelled and the failed transaction recovered. A logo on a product page does not show which of those actions is supported in the proposed deployment.
Rapid Health has useful external evaluation, with limits
Unity Insights evaluated Smart Triage at The Groves Medical Centre in work commissioned by Health Innovation Kent Surrey Sussex and Rapid Health. The post-implementation period ran from 30 October 2023 to 29 February 2024. The evaluation reported that 91% of appointments made through the system were booked automatically. This is an appointment measure, not the proportion of all registered patients managed without staff involvement. Evaluation overview
The fuller evaluation also records mixed experience. The patient survey had only 20 respondents, and most did not judge the new route easier to use or less stressful than its predecessor. These findings are too small and context-dependent to establish a general dissatisfaction rate, but they should not disappear from a favourable product summary. Evaluation findings and limitations, September 2024
The useful inference is modest: there is evidence that substantial booking automation was achieved in this implementation, alongside practical challenges worth investigating. An externally conducted evaluation is valuable, but commissioning, study design and local context still belong in the account.
GP Triage's practice reports answer a different question
GP Triage's Langton report describes reduced GP triage work, while its Swanscombe report counts automated appointments and models the corresponding staff time. The former covers October 2025 to February 2026; the latter covers four weeks beginning 23 February 2026. Neither is a head-to-head comparison with Rapid Health. Langton case study and Swanscombe case study
Those reports support conversations with reference sites. Ask what the practice stopped doing, what staff still review and whether the initial improvement persisted. A practice with a large manual GP inbox may have a different opportunity from one already using efficient care navigation and selective clinician review.
Do not construct a league table from GP hours, automatically booked appointments and waiting-time changes. They are different quantities, measured over different periods in different organisations.
Test the empty appointment book
For both suppliers, a revealing demonstration starts after the straightforward booking journey has succeeded. Remove the relevant appointment capacity from the test environment and repeat the request.
The following is an original proposed acceptance test, not a report of completed product testing. The buyer specifies a clinical urgency outcome, a suitable appointment type and the permitted timeframe. The system then encounters an empty calendar. The evaluator records whether it preserves the urgency, selects an agreed alternative, requests human intervention or leaves the patient without a completed next step.
The key is to inspect the handover rather than the wording alone. A message saying the practice will respond is insufficient evidence that a task has reached the right team. A task reaching an inbox is insufficient evidence that somebody will monitor it within the necessary period.
Repeat the exercise with a closed external service, a patient who cannot travel to the offered site and an interrupted booking transaction. These tests explore local implementation, not just the clinical algorithm.
Children, accessibility and exceptions deserve their own demonstration
Do not use a generic claim of paediatric coverage as a substitute for the intended-use documentation. The NHS Innovation Accelerator's current Smart Triage profile mentions custom pathways for children and at-risk patients; a custom pathway could still involve staff review rather than autonomous assessment. NHS Innovation Accelerator, reviewed 9 September 2026
Ask both suppliers to demonstrate excluded ages, uncertain answers, communication difficulties and requests entered by someone supporting the patient. The product's clinical scope and the practice's access obligations are related but different. A patient outside an algorithm's scope still needs a workable route to care.
The practice should also examine its telephone and reception processes. If staff must repeat a long interview for every person unable to use the digital route, that workload needs to be counted rather than described as digital exclusion solved by an alternative channel.
The contract should explain changes as well as launch
Procurement should establish how changes to clinical content, routing and integrations are communicated. Ask which changes the practice can make itself and which require supplier review, testing or a revised safety argument. These are proposed buyer questions, not claims that either company lacks such arrangements.
A named escalation contact, access to incident logs and the ability to disable an unsafe pathway may be more useful than an additional dashboard. The service needs to remain understandable when something goes wrong.
For clinical questions arising during implementation, a reference tool serves a separate role. As described in iatroX's September 2026 product information, free Ask-iatroX offers linked-source clinical reference. It does not certify the local configuration or replace supplier and practice clinical-safety responsibilities.
Which approach is worth shortlisting?
For a practice that prioritises a clearly documented external implementation evaluation and explicitly described EMIS or SystmOne integration, Rapid Health merits close examination. The next step is to establish how closely the evaluated arrangement resembles the practice's proposed use.
For a practice interested in GP Triage's workflow configuration and its use of a specialised clinical assessment partner, GP Triage merits a demonstration centred on the actual Infermedica product and local booking journey. Reference-site conversations should examine sustained workload changes, not only the initial case-study headline.
For a service with substantial paediatric, accessibility or complex exception requirements, neither a brand comparison nor an automation percentage is enough. The preferred supplier is the one whose intended scope and demonstrated handling of those requirements fit the service, supported by acceptable evidence and contractual arrangements.
Frequently asked questions
Are GP Triage and Rapid Health both autonomous triage products?
Their public descriptions reviewed on 9 September 2026 both include assessment, navigation and booking automation. The extent of automation depends on intended scope and the configuration supplied to the practice.
Does Rapid Health's Class IIb assessment mean it already has that certification?
No: its current primary-care page describes UKCA Class I and says a separate EU MDR Class IIb conformity assessment is in progress. An assessment in progress should not be reported as a completed certification.
Which system saves more GP time?
The reviewed evidence does not provide a comparable head-to-head answer. Different sites report different outcomes, so buyers should measure their own baseline, retained work and sustained net change.