Written Evidence from Cognaris Ltd (PMA0055)
cognaris.ai | https://cognaris.ai/careguide.html
Date: April 2026
This submission was prepared by Cognaris Ltd. AI tools were used in the drafting and preparation of this submission. All content, research, and recommendations were directed, reviewed, and approved by the submitter, who accepts full ownership and responsibility for its accuracy.
Summary
• The NHS faces a patient navigation crisis: up to three-quarters of patients do not know the appropriate level of care for their need, 1 in 6 GP appointments are clinically unnecessary, and 40% of A&E attendances could have been managed in primary care.[1]
• AI-powered triage and navigation tools represent one of the most immediate, scalable, and cost-effective interventions available. Independent analysis by the Tony Blair Institute for Global Change estimates that AI navigation could free 29 million GP appointments per year and deliver £340 million annually in productivity gains — without expanding the workforce.[1]
• Cognaris has built NHS CareGuide AI: a production-grade, voice-and-text patient triage and navigation platform grounded exclusively in NHS.uk content, with privacy-by-design architecture, a freephone telephone channel for digitally excluded patients, and a GP pilot being explored in partnership with Dr Pragasen Moodley, Clinical Director of Stevenage North Primary Care Network.
• The NHS 10-Year Health Plan explicitly backs the rollout of digital triage tools within two years.[2] The technology exists. The clinical evidence exists. The policy commitment exists. What is missing is a clear, fast, and trusted pathway for proven AI tools to reach patients at scale.
• This submission focuses on Questions 2 and 5: the role AI can realistically play in improving patient navigation (Q2), and the systemic barriers that currently prevent or delay deployment (Q5).
Policy recommendations:
1. Establish a fast-track DTAC pathway for advisory-only AI navigation tools that do not generate independent clinical diagnoses, with a 60-day target assessment timeline proportionate to their risk profile.
2. Mandate every ICB to commission at least one AI patient navigation pilot by March 2027, with ring-fenced funding from the £3.4 billion NHS technology budget.
3. Designate AI patient navigation as a core component of the NHS App, with open integration standards for NHS-validated third-party tools.
4. Appoint a national clinical lead for AI navigation within NHS England, responsible for setting quality standards, coordinating pilot evidence, and accelerating adoption.
5. Create a proportionate regulatory classification for advisory-only AI navigation tools that do not generate independent clinical diagnoses — distinct from diagnostic or prescriptive clinical decision support, with a 90-day assessment pathway accessible to UK SMEs.
6. Mandate transparent, auditable decision-making for any AI navigation tool operating in NHS-facing contexts — ensuring the NHS retains meaningful oversight of the clinical logic that directs its patients, with a clear audit trail for post-market surveillance.
About Cognaris Ltd and NHS CareGuide AI
1. Cognaris Ltd is a UK AI product studio registered in England and Wales (Companies House no. 16436570), based in Ruislip, London. We build AI-powered health and care navigation tools combining voice, natural language, and intelligent triage.
2. NHS CareGuide AI is a production-grade voice-and-text patient guidance platform grounded exclusively in NHS.uk content. During every conversation, the platform runs silent urgency triage — red, amber, green — routing patients to the appropriate service: self-care, pharmacy, self-referral, or GP. Self-referral pathways — including physiotherapy, talking therapies, optometry, audiology, and sexual health — are surfaced directly where available, routing patients to the right service without an unnecessary GP intermediary. When a GP referral is indicated, the platform automatically generates a structured pre-consultation brief — symptoms, duration, medical history, patient concern, triage flag, and condition matches — and delivers it to the GP system before the patient arrives. No clinical claims are made beyond the NHS.uk source material. A freephone telephone channel — delivering the same NHS-grounded triage intelligence to patients without smartphones or internet access, with automatic escalation to 111 or 999 where indicated — has been technically validated and is on the platform roadmap, ensuring CareGuide can reach the full patient population rather than only those with digital access.
3. A defining feature of CareGuide AI is its visual clinical pathway architecture. Every condition area has its own NHS-authored decision pathway — designed by clinicians, visible in full, and editable by any NHS department to reflect local services and commissioning priorities. The AI executes these pathways transparently: every routing decision is traceable back to an authored, reviewable logic step. This is not a black box. It is the direct answer to the governance concern that NHS England’s own AI guidance raises — that opaque AI decisions undermine clinical trust and make post-market surveillance impossible.[4] CareGuide’s pathway model means the NHS owns and controls the clinical logic, AI delivers it at scale, and any clinician can audit, challenge, or update any decision the system makes. Critically, pathways can be created or updated in hours rather than weeks — without a software release cycle. An ICB facing a respiratory surge can push updated triage guidance the same day. A national health campaign can be reflected immediately across every deployment. This is the direct answer to the NHS’s need for AI that is both responsive to local services and governable at pace — neither the rigidity of legacy telephone triage logic nor the opacity of unconstrained generative AI.
4. A critical but often overlooked factor in NHS digital tool adoption is patient experience. Systems that retrieve answers from a fixed knowledge base, and pathway-driven tools, tend to produce transactional, question-and-answer interactions that can feel clinical and impersonal. CareGuide AI is designed around genuine conversation: the clinical pathway runs silently in the background while the patient simply talks. The AI understands context across the exchange, asks follow-up questions naturally, and responds in plain language. A patient may simply want to ask a health question, or they may be seeking an appointment — CareGuide accommodates both naturally, without forcing an early commitment or decision. Patients do not feel triaged — they feel heard. A tool patients do not use delivers no clinical value regardless of its technical sophistication.
5. Cognaris is an established AI product studio with proven platform infrastructure — the same AI assistant, voice-narrated education, and intelligent tools architecture that powers live consumer products across multiple domains — applied to NHS patient navigation. The infrastructure is production-tested, the engineering is mature, and the deployment risk is accordingly lower than a bespoke NHS build. Cognaris is a registered NHS supplier with an ODS code issued and NHS Digital onboarding submitted for NHS Login integration. Clinical oversight is provided by Dr Pragasen Moodley, Senior Partner at Stanmore Medical Group, Stevenage, and Clinical Director of Stevenage North Primary Care Network. Dr Moodley formerly served as Chair of East and North Hertfordshire CCG and brings over 20 years of frontline GP and commissioning experience. Stanmore Medical Group is confirmed as a Phase 1 pilot site, subject to completion of NHS integrations.
The AI navigation opportunity
Addressing Question 2: the role AI can realistically play
6. The Committee's framing around personalised medicine and genomics represents the frontier of what AI may eventually deliver. But there is a nearer-term, equally significant opportunity that this submission draws attention to: AI's role in patient navigation — getting the right patient to the right service at the right time, first time.
The uncontrolled AI problem
7. The NHS does not have an AI adoption problem. It has an AI exposure problem. Millions of patients are already using general-purpose AI for health decisions — today, at scale, with no NHS oversight, no clinical grounding, and no accountability. Over 230 million people ask health-related questions on ChatGPT every week globally,[8] and 59% of UK GPs anticipate that more patients will rely on AI tools instead of seeking medical attention.[9] This is not a future risk. It is a present cost.
8. General-purpose AI has no access to NHS guidelines, no patient history, and no clinical accountability. When a patient asks ChatGPT about a minor symptom, its safety rails default to recommending they seek medical advice regardless of clinical need — a necessary liability position for a general-purpose tool, but one that systematically drives avoidable GP appointments. The appointment is booked. The slot is wasted. Uncontrolled AI is not solving the demand problem — it is amplifying it. The question for the Committee is not whether AI should be in the NHS. Patients are already using it, at scale, without NHS oversight. The question is whether the NHS takes control of that channel.
9. The result is a self-reinforcing cycle: uncontrolled AI amplifies clinical anxiety, drives unnecessary appointments, reduces GP availability for patients with genuine need, and pushes more patients back to AI. The policy challenge is replacing unaccountable, ungrounded AI with NHS-validated alternatives that resolve clinical uncertainty rather than amplifying it.
10. The technology underpinning AI patient navigation is proven. The Tony Blair Institute for Global Change estimated in March 2025 that properly deployed AI navigation could free 29 million GP appointments per year and deliver £340 million in annual productivity gains — at an estimated national implementation cost of £10–100 million: a fraction of the projected savings.[1] The barrier is not capability. It is adoption.
11. The scale of the underlying problem justifies the ambition. In November 2025, 32.1 million GP appointments were delivered in England in a single month.[3] NHS 111 handles approximately 1.67 million calls per month, with over 200,000 abandoned each month after waits exceeding 30 seconds.[1] Between 2012 and 2023, the proportion of patients who found it easy to get a GP appointment fell from 81% to 50%.[1] This is not a capacity problem alone — it is a navigation problem. AI can address it directly.
12. NHS CareGuide AI addresses this problem through two deliberate design choices. First, grounding entirely in NHS.uk content: every triage output, every service recommendation, and every piece of guidance is sourced from NHS.uk. The platform cannot hallucinate treatment pathways or generate clinical claims beyond NHS-validated guidance. Second, the visual pathway architecture described in paragraph 3: clinical logic that is authored, visible, and auditable — not opaque generative output. Together, these two design choices produce an AI that the NHS can trust, deploy, and govern.
13. The personalisation argument extends to navigation. A patient presenting with chest pain and a history of cardiovascular disease should not receive the same navigation pathway as a patient presenting with chest pain arising from anxiety. AI can parse this complexity consistently and at scale in a way that decision-tree models and non-clinical call handlers cannot. As NHS England's own guidance acknowledges, symptom checkers including NHS 111 online are already trialling AI for triage, while noting the risk that risk-averse models could "paradoxically increase clinical workload" if poorly calibrated.[4] Getting this calibration right — which requires clinical oversight, real-world piloting, and iterative refinement — is precisely the work CareGuide AI has been designed to support.
Barriers to deployment
Addressing Question 5: systemic barriers that prevent or delay deployment of proven innovations
14. Cognaris has developed a detailed understanding of the barriers the Committee is examining, informed by direct engagement with NHS Digital onboarding, regulatory frameworks, and the guidance of our clinical advisor Dr Pragasen Moodley, Senior Partner and PCN Clinical Director with over 20 years of frontline GP and commissioning experience. The government has set clear objectives for AI in the NHS.[2] The technology to meet them exists. What is needed is a fast-track mechanism that allows proven, clinically supported tools to be tested and deployed in line with those objectives, without navigating a system designed for a slower era. We set these barriers out plainly, as they are illustrative of the wider challenge facing AI health technology companies — particularly early-stage, UK-based SMEs.
Barrier 1: Data trust and architecture
15. Data trust is consistently cited as the single biggest barrier to NHS AI adoption, and rightly so.[4] The controversy surrounding early NHS AI data partnerships cast a long shadow over the sector. Yet the architectural response to this concern is rarely examined at the level of technical design. Cognaris has built a specific solution: personal data never leaves the patient's device. Patient identifiers are detected and stripped client-side before any message is transmitted. Cognaris servers never receive, process, or store personal data. The AI only ever processes anonymised clinical content — symptoms, conditions, medications. Even in the event of a breach, there is no personal data to leak.
16. This privacy-by-design architecture means CareGuide AI processes anonymised data, substantially simplifying compliance with GDPR, DSPT, and NHS data residency requirements. NHS Login handles authentication; Cognaris never sees the NHS number, patient identity, or GP registration. The Committee will hear that data governance is a barrier to AI adoption. We submit that it is a solvable engineering problem — if the regulatory and commissioning framework creates space for solutions to be evaluated on their technical merits rather than assessed by category alone.
Barrier 2: Regulatory uncertainty and disproportionate burden
17. The regulatory pathway for an AI tool that provides patient navigation based exclusively on published NHS guidance is currently unclear. The MHRA's Software and AI as a Medical Device framework[5] was not designed with this category of tool in mind. An AI triage platform that routes patients to GP, pharmacy, or self-care — without generating independent clinical diagnoses — occupies a grey zone between general wellness software and Class IIa medical device. That uncertainty alone is sufficient to deter commissioners and delay pilots.
18. DTAC (Digital Technology Assessment Criteria) is the NHS's primary gateway for digital health tools. For a company seeking to pilot with a single PCN, the DTAC self-assessment is a substantial resource commitment with no guaranteed outcome and no defined timeline. We are not arguing for lower standards. We are arguing for proportionality: a tool grounded entirely in NHS.uk content, advisory-only, with no prescriptive clinical output, should have a clearly defined, expedited regulatory status that reflects its actual risk profile.
19. The NHS 10-Year Health Plan commits to a new regulatory framework for medical devices including AI, to be published in 2026.[2] This is the right moment to embed proportionate, risk-stratified regulation for AI navigation tools into that framework from the outset.
Barrier 3: Commissioning structure and ICB inertia
20. NHS commissioning structure creates a fundamental asymmetry for AI navigation tools. The benefits accrue systemically — reduced A&E pressure, freed GP capacity, improved patient outcomes — while the commissioning responsibility sits at ICB or PCN level, where budget holders do not have visibility of or accountability for the downstream savings their decisions generate. A PCN that pilots an AI navigation tool and reduces its own appointment volume may find its funding adjusted downward. There is no incentive architecture for adoption.
21. The result is that AI navigation tools with strong evidence bases remain in "perpetual pilot" — funded by innovation budgets, evaluated by Health Innovation Networks, celebrated in case studies, and never commissioned at scale. This is not a failure of individual commissioners. It is a structural failure. The NHS 10-Year Plan's commitment to rolling out digital triage within two years requires a commissioning framework that can deliver at pace, not one that returns each tool to the start of a procurement cycle in each ICS.
22. The House of Commons Health and Social Care Committee's 2023 report on digital transformation noted that successive Governments have attempted digital transformation of the NHS with progress that has been "slow and uneven".[6] The problem is not primarily technological. It is governance.
Barrier 4: Fragmentation and the absence of a national front door
23. There is currently no standard for what NHS-validated patient navigation looks like. NHS 111, GP online consultation tools, symptom checkers, and AI assistants operate in parallel with no common data standard, no shared triage outcome definitions, and no interoperability requirement. A patient who uses an AI navigation tool at 9pm and then calls 111 at midnight is assessed from scratch. This is not just inefficient — as the TBI report notes, NHS 111 decisions are routinely challenged by clinicians in care settings, creating duplicate clinical workload.[1]
24. The NHS App is the natural integration point. NHS England has connected 85% of acute trusts to the NHS App, with 100% targeted by the end of this year.[2] The NHS 10-Year Plan explicitly envisages AI virtual assistants accessible through the NHS App.[2] Designating AI navigation as a core NHS App function — with open integration standards for NHS-validated third-party tools — would create the national front door that currently does not exist.
25. Open integration also resolves the procurement lock-in problem. The NHS should not be committed to any single AI navigation model — whether built in-house, procured from a large supplier, or developed through an innovation programme. A platform architecture with open API standards allows NHS-developed models, nationally procured models, and validated third-party tools to operate within the same framework and be evaluated against shared clinical outcome metrics. Cognaris is built on exactly this model: the platform integrates with any NHS system and any AI model via standard APIs, and is designed to run alongside or on top of models the NHS has already commissioned or developed — not to replace them.
26. This makes systematic comparative evaluation possible. A/B testing of competing navigation models against shared metrics — triage accuracy, appropriate service routing, patient satisfaction, GP appointment reduction — would give commissioners, clinicians, and regulators the evidence base they currently lack to make informed adoption decisions. Cognaris can provide this: structured evaluation reports and outcome metrics across model variants, giving the NHS the quality assurance framework it needs to deploy AI navigation with confidence and continuously improve it over time. The NHS should not adopt a single AI navigation solution and consider the problem solved. Performance varies significantly between models, and the field is advancing rapidly. A framework that mandates open integration standards and systematic comparative evaluation ensures the NHS can continuously adopt better-performing tools as they emerge — rather than locking in an early incumbent, whether a large technology supplier or an in-house build, regardless of its ongoing performance relative to alternatives.
Barrier 5: The pilot-to-scale gap
27. The most acute barrier is the gap between a successful pilot and commissioned deployment. Cognaris, with an active clinical advisor relationship, a GP pilot in exploration, and a production-grade platform, is — by the standards of the sector — well positioned. Yet the pathway from PCN pilot to ICB commissioning remains opaque, under-resourced, and slow.
28. Health Innovation Networks exist specifically to bridge this gap, and we are actively engaging with Health Innovation Network South London and UCLPartners. But HINs are intermediaries, not commissioners. They can open doors; they cannot write contracts. The structural change required is for ICBs to have ring-fenced commissioning capacity — and accountability — for AI navigation tools that have passed regulatory assessment and demonstrated clinical effectiveness.
Policy Recommendations
29. The Committee asked for concrete, actionable recommendations. We offer six.
Conclusion
30. The NHS faces a navigation crisis that costs billions of pounds annually,[1] misdirects millions of patients, and contributes to the access pressures that undermine confidence in the health service. Uncontrolled general-purpose AI is actively making this worse. The technology to address it — NHS-grounded, clinician-authored, privacy-by-design — exists and is proven in NHS settings. What is needed is not more innovation. It is a clear, funded, and fast pathway to bring proven tools to patients.
31. Cognaris is a UK AI company that has built NHS-grounded patient navigation technology that is production-ready and clinically supported. The final integrations required for NHS deployment — NHS Login authentication, GP Connect, and FHIR data access — are funded development work, not research. This submission is itself an illustration of the barrier: the technology exists, the clinical partnership exists, the regulatory pathway is understood, and the remaining step is the funded engagement that allows it to reach patients. We would welcome the opportunity to give oral evidence if the Committee considers it useful.
About Our Clinical Advisor
32. NHS CareGuide AI is supported by Dr Pragasen Moodley, Senior Partner at Stanmore Medical Group, Stevenage and Clinical Director of Stevenage North Primary Care Network. Dr Moodley has practised as a GP for over 20 years and formerly served as Chair of East and North Hertfordshire CCG, holding commissioning leadership responsibility across the ICS footprint. Stanmore Medical Group is confirmed as a Phase 1 pilot site for NHS CareGuide AI, subject to completion of NHS integrations.
References
[1] Tony Blair Institute for Global Change, Preparing the NHS for the AI Era: Why Smarter Triage and Navigation Mean Better Health Care (March 2025). Available at: https://institute.global/insights/public-services/preparing-the-nhs-for-the-ai-era-why-smarter-triage-and-navigation-mean-better-health-care
[2] HM Government, NHS Fit for the Future: 10 Year Health Plan for England (3 July 2025).
[3] NHS England Digital, Appointments in General Practice, November 2025 (Published 18 December 2025). Available at: https://digital.nhs.uk/data-and-information/publications/statistical/appointments-in-general-practice/november-2025
[4] NHS England, Artificial Intelligence (AI) and Machine Learning — Long Read (2024). Available at: https://www.england.nhs.uk/long-read/artificial-intelligence-ai-and-machine-learning/
[5] Medicines and Healthcare products Regulatory Agency (MHRA), Software and Artificial Intelligence (AI) as a Medical Device (2024). Available at: https://www.gov.uk/government/publications/software-and-artificial-intelligence-ai-as-a-medical-device
[6] House of Commons Health and Social Care Committee, Digital transformation in the NHS, HC 223 (30 June 2023). Available at: https://publications.parliament.uk/pa/cm5803/cmselect/cmhealth/223/report.html
[7] NHS England, NHS AI expansion to help tackle missed appointments and improve waiting times, Press Release (14 March 2024). Available at: https://www.england.nhs.uk/2024/03/nhs-ai-expansion-to-help-tackle-missed-appointments-and-improve-waiting-times/
[8] OpenAI, Introducing ChatGPT Health (7 January 2026). Available at: https://openai.com/index/introducing-chatgpt-health/
[9] Kharko A, Garcia Sanchez C, Hagström J, et al. "General practitioners' opinions of generative artificial intelligence in the UK: An online survey." Digital Health. 2025;11:20552076251360863. doi: 10.1177/20552076251360863. Survey of 1,005 UK GPs, January 2025. Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC12276478/