Medigent.ai
An on-premise AI pre-consultation platform that structures patient information before appointments — so doctors arrive prepared, consultations run faster, and more patients receive care each day.
40%+
More Patients/Day
6M+
UK Referral Backlog/yr
HIPAA
& UK Compliant
0
Data Egress
The Problem: A Capacity Crisis in Primary Care
Healthcare systems in the UK and USA are operating under severe structural pressure. In the UK alone, over 6 million GP referrals and patient cases go unprocessed every year — not because of a lack of clinical intent, but because there are not enough appointment hours to absorb them. The average GP consultation runs 10–15 minutes, but a significant portion of that time is spent gathering information the patient could have provided in advance: current symptoms, medical history, duration, medications, prior diagnoses, and the reason for the visit.
In the United States, the same dynamic drives a different but equally critical outcome: hospitals and outpatient clinics measure throughput directly. More completed appointments per physician per day translates to more patients receiving timely care — and a more sustainable revenue model that allows providers to invest back into facilities, staff, and equipment.
Medigent.ai was built to address this directly. Their platform needed an AI layer that could handle structured patient intake before the appointment — so that when the physician opens the consultation, the work of assembling patient context is already done.
What Medigent AI Does
The Medigent pre-consultation system guides patients through a structured intake process before their appointment — via a web or mobile interface, available from home or in the waiting room. The AI does not make diagnoses and does not recommend medications. It follows clinically approved intake procedures and protocols to systematically gather and organise patient information that the physician needs.
When the doctor opens the patient file, they receive a structured pre-consultation brief: chief complaint, symptom timeline, relevant history flags, current conditions, and any items the intake protocol identified as requiring direct clinical attention. The physician can review this in 60–90 seconds rather than spending the first 4–6 minutes of the appointment gathering it manually.
- —Structured symptom intake: Guided questioning adapts based on patient responses, following clinical intake frameworks — not free-form chat. The output is a structured summary, not a transcript.
- —Medical history parsing: Patients input or confirm existing conditions, prior surgeries, known allergies, and current medications. The AI flags relevant history against the presenting complaint for the physician.
- —Referral context structuring: For GP referral cases, the system parses the referral document and links it to the patient's intake responses — giving the specialist the full picture before the appointment begins.
- —No clinical recommendations: The system explicitly does not suggest diagnoses, treatments, or medications. It operates strictly within the scope of intake and information structuring, following procedures defined and approved by clinical leads.
- —Physician-ready brief: The output is a standardised pre-consultation document — consistent in format across all patients, reducing cognitive overhead and making it faster to identify what actually requires clinical judgement.
Why On-Premise Matters in Healthcare
Patient data is among the most sensitive personal data that exists. Transmitting it to external cloud AI providers — including large consumer AI platforms — creates compliance exposure under HIPAA in the United States and UK GDPR and NHS data governance standards in the United Kingdom. Beyond the legal risk, it creates genuine trust risk with patients and regulators.
The Medigent system runs entirely on infrastructure controlled by the healthcare provider. Patient intake data never leaves the organisation's network. The AI models — including the NLP components that parse and structure patient responses — run locally. No query, no response, and no patient record is transmitted to any external service.
This also eliminates dependency on third-party AI provider uptime, pricing changes, and policy shifts — all of which have caused disruption for healthcare organisations that built workflows on top of cloud AI APIs. The Medigent system is self-contained and operationally stable.
Infrastructure & Technical Specifications
The deployment runs on dedicated on-premise hardware managed by the healthcare provider's IT infrastructure team. Key specifications:
The intake NLP pipeline uses a fine-tuned language model quantized to 4-bit precision, optimised for structured extraction rather than open-ended generation. Response latency for intake processing is under 800ms per patient interaction step. The system is load-tested to handle concurrent intake sessions across all appointment slots without degradation.
Deployment Phases
Phase 1 — Clinical Protocol Definition (Weeks 1–2)
Safe4AI worked with Medigent's clinical advisory team to define the intake procedures the AI would follow. This included mapping the question trees for general practice, specialist referral, and urgent care contexts. The system was designed around these protocols — it follows clinical logic, not open-ended AI reasoning. No intake flow was implemented without explicit clinical sign-off.
Phase 2 — NLP Pipeline & Intake Engine (Weeks 3–6)
The core AI engine was built to parse free-text patient responses into structured clinical fields (symptom onset, severity scale, associated symptoms, relevant history, current medications). A local language model handles entity extraction and response normalisation. The structured output schema was validated against the intake requirements defined in Phase 1.
Phase 3 — EHR Integration & Pre-Consultation Brief Generation (Weeks 7–9)
The system was integrated with Medigent's existing electronic health record system. Patient demographic data is pulled automatically — patients do not re-enter what is already on file. After intake is complete, the system generates a standardised physician brief in the format the clinical team specified: chief complaint, key history flags, symptom summary, and any intake flags requiring direct clinical follow-up.
Phase 4 — Pilot & Clinical Validation (Weeks 10–12)
A pilot cohort of 15 physicians ran the system in parallel with their standard workflow for four weeks. Pre-consultation briefs were reviewed and rated by clinicians for completeness and accuracy. Feedback was used to refine the intake question logic for edge cases. Physician-reported time savings averaged 4.5 minutes per consultation during the pilot period.
Phase 5 — Full Rollout & Training (Weeks 13–14)
Platform rolled out to all physician accounts. Patient-facing onboarding was kept to under 3 minutes — accessible via a link sent with the appointment confirmation. Staff training covered brief review workflow and how to flag intake responses for quality review. Adoption reached 81% of booked appointments completing pre-consultation intake within the first month.
Before & After: Consultation Workflow
| Step | Without AI | With Medigent AI |
|---|---|---|
| Patient intake | Doctor gathers history manually, 4–8 min | Completed pre-appointment, doctor reviews 60–90 sec brief |
| Symptom history | Verbal Q&A, inconsistent depth per patient | Structured, protocol-driven — consistent across all patients |
| GP referral context | Doctor reads referral letter mid-consultation | Referral pre-parsed and linked to intake before appointment |
| Relevant history flagging | Reliant on patient recall | AI flags history items relevant to presenting complaint |
| Consultations per doctor/day | Typically 15–20 | Increases to 22–28 with recovered intake time |
| Unprocessed referral risk | High — backlog builds when capacity is fixed | Reduced as each appointment becomes more time-efficient |
Impact
The core outcome of the Medigent AI platform is structural: by removing information-gathering overhead from the consultation itself, each appointment can be shorter without reducing quality — and the time recovered is returned directly to clinical care. A physician who can complete a consultation in 8–10 minutes rather than 13–15 can see 40% more patients in the same working day. At scale, this is a meaningful intervention against the access gap that leaves millions of patients waiting.
- —40%+ increase in patients seen per physician per day — recovered from intake time alone
- —4.5 minutes average time saved per consultation during pilot validation
- —81% of booked appointments completing pre-consultation intake within first month of rollout
- —Physician-reported improvement in consultation focus — more time on clinical decision-making, less on administrative gathering
- —Standardised intake quality — consistent data depth across all patients regardless of how they communicate
- —Zero patient data transmitted externally — fully HIPAA and UK GDPR compliant at the infrastructure level
The platform is designed for deployment across primary care, specialist outpatient, and referral-heavy clinical environments. It is currently in use and available for licensing to healthcare providers in the US, UK, and internationally via Medigent.ai.