For medical and healthcare schools
Developed with senior applied-AI researchers and clinical simulation leads, and running in a live Medical school curriculum. Students practise as often as they like. Every consultation is marked against your criteria, with a coach that explains the reasoning.
Amelia Carter
Adult Woman · Headaches
Clinical Reasoning Coach
The coach reads the marked consultation and talks the student through the reasoning behind it. It only runs when asked, so the results page stays quick.
Six stages of a consultation, every time.
- 1Opening and rapportstrong
- 2Cue gatheringdeveloping
- 3Hypothesis formationdeveloping
- 4Risk and escalationnot shown
- 5Synthesis and shared understandingnot shown
- 6Closing and safety netdeveloping
Opening and rapport. Cue gathering. Hypothesis formation. Risk and escalation. Synthesis and shared understanding. Closing and safety net. Each one is rated and quoted back in the student's own words, followed by what to do differently.
Hidden cue
“I'm alright, thanks, just a bit tired to be honest.”
- What this might have signalled
- Possible poor sleep or overwork contributing to headaches and general wellbeing.
- Try saying
- “You mentioned feeling tired. How has your sleep been recently?”
Differential
- Migraine without auraConsidered partially
- Tension-type headacheNot considered
- Headache secondary to stress and lifestyleNot considered
- Medication overuse headacheNot considered
Every plausible diagnosis for the case is listed as considered fully, considered partially, not considered, or anchored on too early, with the question that would have tested it.
How it was built
Six stages of reasoning, in the order clinicians asked for.
"A tone a nervous student would not find harsh", from senior clinician feedback.
The three things to work on before the next consultation, in order of priority.
Student experience
In our pilot at Aston University, medical students practised realistic patient consultations. Each completed two full consultations, then reviewed their scores, rubric and feedback.
Pilot evaluation, Aston University
100%
see a valuable role for DialogicAI alongside existing teaching and simulated-patient practice.
94%
found the platform easy to navigate.
94%
found the clinical content accurate and appropriate for their level of training.
Feedback
“The detailed feedback… pinpointed my weak areas and also gave actual explanations as to what I did wrong… I really liked the fact that it also suggested better phrases and questions I could ask the patient in the future. Not just telling me what could be better, but how I can make it better.”
Red flags
“It allowed me to see where I was going wrong and whether or not I missed any red flags. I was able to use the feedback to learn how to have smoother consultations.”
Phrasing
“Comprehensive feedback with examples of lines that would have made the consultation better. Very useful and would help a lot with practice.”
Realism
“The encounters aren't too easy like other platforms, which adds to the realism. I particularly liked the evidence and it linking back.”
Overall
“This is a very amazing project… I can't wait for more cases to do.”
Most students completed both consultations and their feedback review in under 45 minutes.
Academic output
The simulation and marking approach behind DialogicAI is developed with Aston University and peer-reviewed, not just demoed.
From Dialogue to Feedback: A GMC-Aligned Generative AI Simulation Platform for Clinical Communication Training
Fouad, S., Muskwe, L., Harib, W. and Hassan-Smith, G.
International Conference on AI in Healthcare (AIiH 2026), pp. 163-176. Cham: Springer Nature Switzerland.
Describes the GMC-aligned simulation platform behind DialogicAI, from patient dialogue generation through to structured, criterion-level feedback.
Generative Pretrained Transformers for Medical Communication Training and Automated Performance Assessment
Ilyas, S.R., Fouad, S. and Hassan-Smith, G.
UK AI Conference, 2025.
Evaluates fine-tuned GPT models for simulating patient interactions and automatically assessing student communication skills against marking rubrics.
Questions
A chatbot answers questions. Our patients hold things back until they are asked properly, only follow the language their character would follow, react to how they are spoken to, and will not reveal the mark scheme. Then every consultation is marked against your criteria with the evidence quoted.
Yes. Build Performance Frameworks from your own criteria, nest domains up to four levels, bundle them into Assessment Blueprints, and publish. Or start from the 121 criteria we ship.
Yes. Students hold to talk and hear the patient reply in a chosen voice. Typing works too. Voice input uses the browser's speech recognition, so it depends on browser support.
Free practice is unlimited. On assessed assignments, faculty set the cap.
In the EU with OVH, on a dedicated server and a dedicated managed database.
Pricing is per institution. Book a demo and we will scope it with you.