The 2025/26 and 2026/27 GP contracts describe a version of general practice that is measurably more digital than what came before. Cloud-based telephony. Mandatory online consultation access. GP Connect interoperability. Risk stratification using digital tools. Data collection against standardised access metrics. QOF alignment with current NICE guidance. Same-day clinically urgent access for every patient.
None of these requirements mention AI explicitly. But each creates a problem that AI tools are increasingly well-positioned to solve — and together they describe a practice operating model where technology is not optional but structural.
The Access Layer: Managing Demand
The contract requires online consultations open throughout core hours and all clinically urgent requests managed same-day. This creates an open channel of patient demand that must be triaged, prioritised, and managed without overwhelming clinical capacity.
AI-powered triage tools address this directly. Rapid Health Smart Triage autonomously assesses patients and books appointments without staff intervention — the most complete solution, deployed across 30+ ICSs with published evidence showing 91% autonomous allocation and 73% reduction in waiting times. Anima provides AI-assisted prioritisation of the online queue. QuantumLoop EMMA and InTouchNow handle phone demand with AI voice agents that eliminate call queuing. Accurx and eConsult collect online requests for manual triage.
The access layer is where AI adoption is most visible to patients and most immediately impactful for practice workload. It is also where most NHS investment and policy attention is currently focused.
The Documentation Layer: Consultation Efficiency
Same-day urgent access requires seeing more patients per session without sacrificing quality. AI ambient scribes directly support this by reducing documentation time — typically saving 2-5 minutes per consultation, compounding across a full clinic into 2-4 additional appointment slots without extending hours.
Heidi is the most widely adopted individual scribe among UK GPs, with customisable templates and dedicated hardware (Heidi Remote). TORTUS has NHS institutional backing through NHS England's ambient voice technology evaluation. Accurx Scribe integrates within the broader Accurx platform. Each reduces the administrative burden that makes general practice unsustainable — the hours spent typing after clinic, the evening documentation sessions, the growing sense that paperwork has displaced patient care as the primary output of a GP's working day.
The documentation layer is where GPs most immediately feel the benefit of AI. Less time typing after clinic, less administrative dread, and more sustainable working patterns.
The Knowledge Layer: Clinical Quality
The QOF changes require explicit NICE alignment. Risk stratification surfaces complex, multimorbid patients who need guideline-informed management. Same-day urgent access produces undifferentiated presentations requiring rapid clinical reasoning across the full scope of general practice.
This is where iatroX fits. It provides the clinical knowledge foundation that makes every other layer safe and clinically effective.
Ask iatroX delivers instant, NICE-grounded guideline reference with citations during consultations — answering the clinical questions that arise when managing the complex patients that risk stratification surfaces or the undifferentiated presentations that same-day urgent clinics produce. When you need to confirm whether your heart failure management includes all four NICE-recommended pillars, or whether your diabetes care process captures all 8 required elements, the answer arrives in seconds with a link to the source guideline.
Brainstorm supports structured clinical reasoning for patients whose complexity defies a single guideline pathway — the multimorbid patient with diabetes, heart failure, CKD, and COPD where the management of one condition conflicts with the management of another.
The Q-Bank maintains clinical knowledge through spaced repetition across all domains — ensuring the GP staffing same-day urgent slots has current, durable knowledge across the full scope of presentations they may encounter. This is not exam preparation; it is professional knowledge maintenance for safe practice.
The CPD module documents professional development from daily clinical practice — turning guideline queries and learning moments into portfolio-ready CPD evidence mapped to GMC domains.
The Knowledge Centre provides structured guideline browsing for systematic coverage of QOF-relevant conditions — enabling clinicians to review the NICE recommendations for each indicator domain before the QOF year begins.
The knowledge layer is the least visible but most clinically important layer in the stack. Access improvements and documentation efficiency are only valuable if the clinical decisions made within each consultation are safe and evidence-based. Without the knowledge layer, faster consultations risk being worse consultations. More patients seen per session is only a good outcome if those patients receive guideline-aligned care.
The Population Health Layer: Proactive Care
Risk stratification tools analyse patient data to identify who needs proactive care. The contract embeds this as a core PCN requirement. The tools — Eclipse, QAdmissions, embedded EMIS/SystmOne analytics, and ICB-provided population health dashboards — generate the intelligence. The clinical team must act on it.
Acting on it requires the clinical knowledge to manage the flagged patients effectively. A risk score without a care plan is a number. A care plan without guideline grounding is guesswork. The population health layer identifies who. The knowledge layer informs what. Both are necessary.
Where AI Does Not Fit
AI does not replace the GP. The contract is clear: clinical urgency is determined by the practice. The therapeutic relationship, the physical examination, the professional accountability, the human judgement that integrates clinical data with patient context and preference — these remain irreducibly human. No AI tool replicates the GP who knows the patient, understands their family situation, and makes a management decision that accounts for factors no algorithm can weigh.
AI does not solve the workforce crisis. Technology can make existing GPs more efficient, but it cannot create the 6,000+ GP partners lost since 2015. The ARRS expansion and practice-level funding help at the margins, but the fundamental gap between patient demand and workforce capacity requires political solutions — training more GPs, retaining more GPs, and funding general practice adequately. Technology optimises what exists; it does not replace what is missing.
AI does not remove professional accountability. Using an AI tool that provides a recommendation does not transfer liability if that recommendation is followed and the patient is harmed. The GP made the decision; the AI provided input. Professional responsibility does not delegate to an algorithm, regardless of how sophisticated the algorithm is.
The Complete Practice Stack for 2026/27
Access and triage: Rapid Health Smart Triage, Accurx, eConsult, or Anima. Usually ICB-funded.
Phone management: QuantumLoop EMMA or InTouchNow for practices with phone congestion.
Documentation: Heidi, TORTUS, or Accurx Scribe. Paid (£50-200/month per clinician).
Clinical knowledge and QOF: iatroX. Free.
Learning and CPD: iatroX Q-Bank and CPD module. Free.
Risk stratification: Eclipse, QAdmissions, or embedded EMIS/SystmOne tools. Usually ICB-funded.
Cloud telephony: Practice infrastructure with reporting capability.
Prescribing: BNF app. Free.
Start with the free layers. iatroX and the BNF cost nothing, require no procurement, and provide the clinical foundation that makes every other tool safe. Build the paid and ICB-funded layers as your practice and budget allow.
The contract describes a digitally-enabled general practice. AI tools make that description achievable. The practices that thrive under the 2026/27 requirements will be those that adopt technology thoughtfully — using AI for what it does well (retrieval, triage, documentation) while preserving clinical judgement for what only humans can do (reasoning, relationships, accountability).