AI Doctor Simulator: pharma reps rehearse detailing calls with an AI doctor who pushes back
Every session is scored against your approved key messages. The doctor answers only from your SmPC and approved claims. Browser first, optional Quest, with a manager dashboard from day one.
Quick answer
The AI Doctor Simulator is a training product for pharmaceutical field forces. A medical rep opens it in the browser, picks a doctor persona and runs a detailing call. The AI doctor asks questions, raises objections and pushes back. When the call ends, the rep gets a scored debrief: which approved key messages landed, which objections were handled, and any statement that was not supported by the approved materials. The simulator runs only on content your medical, legal and regulatory reviewers have approved. VR Express, an XR studio working in Sofia since 2015, builds it as a fixed-price pilot that is credited to the full build.
Who it is for
The simulator is built for the people accountable for how reps perform in front of healthcare professionals:
- Sales Excellence and Commercial Training managers who need every rep to practise the hard call before a launch or a new cycle, not just once in role-play.
- Brand managers who want the approved key messages delivered the same way in every territory.
- First-line sales managers who cannot ride along on every call but need to know who is ready.
- Medical affairs and compliance reviewers who want training content they can approve, version and audit.
Typical moments to start are a new product or indication, a new rep intake, a change to the SmPC or key messages, or a cycle meeting where you want measurable practice instead of a workshop.
How the AI doctor simulator works
- You give us the approved pack. The SmPC, the key messages and the approved claims for one product. Nothing else goes into the doctor's knowledge.
- We design the doctor personas. For example a sceptical specialist, a busy GP with four minutes, and a cost-focused prescriber, each with realistic objections drawn from your field experience.
- Your reviewers approve everything. Corpus, personas, objections and the scoring rubric go through your medical, legal and regulatory review before launch. The content approval workflow is part of the product, not an afterthought.
- Reps rehearse. In the browser, by voice or text, as often as they like. Each session takes about ten minutes. An optional Meta Quest version runs the same scenarios for training centres and cycle meetings.
- Every session is scored. Key-message coverage, accuracy against the SmPC, objections handled, and any unsupported statement flagged, with a debrief the rep can act on before the next call.
- Managers see the dashboard. Adoption, scores by rep and by key message, the objections that break the team, and progress from the first session to the fifth.
Approved content only, by design
The question every pharma team asks first is whether an AI will say something it should not. The simulator is designed around that question. The doctor and the coach answer only from the approved corpus. A question outside it is declined or redirected rather than improvised, so there is no off-label output by design. When a rep makes a claim the corpus does not support, the session flags it in the debrief, which is exactly the habit you want to correct before a real HCP hears it.
Before go-live we run a set of off-label and out-of-scope test prompts against the build and fix anything that slips through. When the SmPC or key messages change, the corpus is updated, re-approved by your reviewers, and the version in use is logged. To be clear about what this is: a product designed to keep your approval process in control, not a regulatory certification. Your review procedures still decide what goes in.
What you get: pilot, full build, run
- Pilot (3 to 4 weeks). One product, one language, three doctor personas, browser only. Scored debrief per session, a manager dashboard and a pilot readout with the numbers. Fixed-price pilot, credited to the full build.
- Full build. More products or indications, more personas and languages, a photoreal avatar if you want one, export to your LMS, and the optional Quest version.
- Run. Hosting, content refresh whenever the SmPC or key messages change, new personas for each cycle, and a monthly KPI report, on a care plan with licences per rep.
You can see the full flow before you share anything: the demo runs on a fictional molecule with a mock SmPC, so no client data is needed. For the budget logic behind rep training, see the cost and ROI model for medical-rep training.
The flip side: an SmPC-grounded assistant for your HCP site
The engine that keeps the AI doctor inside the approved corpus can also face outward. On a brand's HCP website it becomes an assistant that answers healthcare professionals' questions from the SmPC and shows the citation for every answer. It goes through the same approval workflow and is scoped as its own deliverable, so a brand team can start with rep training, the HCP assistant, or both.
AI doctor simulator vs classroom role-play
Most field forces practise detailing in role-play at the cycle meeting: a trainer or a colleague plays the doctor, once or twice, and the feedback is an impression. It is useful, but it does not scale and leaves no record.
- Repetition: a rep can run the same difficult call many times in a week instead of once a cycle.
- Consistency: every rep meets the same personas and objections, so results are comparable across regions.
- Measurement: each session is scored against your approved key messages, not judged by whoever played the doctor that day.
- Control: the doctor's knowledge is your approved corpus, versioned and reviewed, rather than whatever a colleague remembers.
Role-play still has a place for coaching nuance; the simulator gives it a measured baseline. For the wider picture of immersive rep training, read our guide to virtual medical representative training and the Pharma XR overview.
Full comparison: AI doctor simulator vs role-play training.
Proof: the building blocks we have already shipped
The simulator combines three layers we have delivered before: a lifelike avatar (including a photoreal avatar for a pharmaceutical client and a bot avatar for a bank), a conversational layer grounded in approved documents, and training simulations with measurable outcomes for pharma and banking clients.
- Recordati Eligard: HCP information hub
Pharma · HCP website
A dedicated landing page and information hub that gives healthcare professionals clinical data and patient-support resources. It is the kind of HCP site the simulator's flip side, an SmPC-grounded assistant, is built to sit on.
- Halen VR: training simulation for medical staff
Pharma · VR training
A VR simulation that trains medical staff on a new, complex surgical device in a risk-free environment, so they are proficient before using it in a clinical setting.
- Postbank VR Onboarding
Banking · VR training programme
Gamified VR modules for new hires, including simulated customer interactions: the same idea of rehearsing a conversation safely before doing it for real.
- Zentiva Sideral VR
Pharma · HCP education
An immersive educational experience that visualises the molecular benefits of Sideral for healthcare professionals, built on approved product science.
- Janssen VR
Pharma · Medical congresses
A VR experience that engages healthcare professionals at medical conferences by visualising complex biological processes, going beyond a standard presentation.
- Gaviscon AR Avatar
Consumer health · AR avatar
An augmented-reality avatar experience for the Gaviscon brand: an early example of the avatar layer we now pair with conversational AI.
More work is in the project portfolio.
AI doctor simulator: frequently asked questions
What is an AI doctor simulator for pharma reps?
It is a training tool in which a medical representative holds a detailing conversation with an AI-driven doctor avatar. The doctor asks questions, raises objections and reacts to what the rep says. After each session the rep gets a scored debrief against the brand's approved key messages, and the sales manager sees the results on a dashboard.
Can the AI doctor say something off-label?
The simulator is designed so that the doctor and the coach answer only from the approved corpus you give us: the SmPC, the key messages and the approved claims. Questions outside it are declined or redirected, and unsupported statements by a rep are flagged. This is a design principle, not a regulatory certification; your own review process stays in charge of what goes in.
Who approves the content and the doctor personas?
You do. The approved materials, the persona descriptions, the objections and the scoring rubric all go through your medical, legal and regulatory review before any rep sees them. When the SmPC or key messages change, the corpus is updated and re-approved, and the version in use is logged.
Do reps need a VR headset?
No. The simulator runs in the browser on the laptop or tablet a rep already carries, with voice or text. A Meta Quest version of the same scenarios is optional, for training centres or cycle meetings where you want full immersion.
What does the pilot include?
One product, one language and three doctor personas, browser only, with a scored debrief per session and a manager dashboard. It runs for 3 to 4 weeks with your reps. The pilot is fixed-price and its fee is credited to the full build if you continue.
Do we have to share confidential data to see a demo?
No. The demo runs on a fictional molecule with a mock SmPC and mock key messages, so you can see the full flow (persona, objections, scoring, dashboard) without sharing anything. Your own product pack is only loaded in the pilot.
Can the same engine answer doctors on our HCP website?
Yes. The same grounded engine can sit on a brand's HCP site as an assistant that answers healthcare professionals' questions from the SmPC, with a citation for each answer. It is scoped as a separate deliverable and goes through the same content approval.
See a rep handle a sceptical doctor, then the scored debrief
A 20-minute demo on a fictional molecule, no data needed. Bring your next launch or cycle and we will scope a fixed-price pilot: one product, one language, three doctor personas.