Case studies: empathetic AI in practice

These are the deployments and pilots that shaped the product. Each describes the problem, what we built, and the honest limits of the result — no unverifiable numbers and no named customers.

Emotion-aware clinical intake — Healthcare

Intake questionnaires capture symptoms but miss distress. Clinicians repeatedly told us that the patients who most need attention are the ones who understate their state in a form.

An AI USM consultation avatar conducts the intake conversation, records the structured answers and attaches an emotional-state summary — hesitation, flat affect, agitation — to the clinician's handover note. Nothing is scored as a diagnosis; the output is a flag for a human to read.

Clinicians receive a two-part record: the facts the patient reported, and how they reported them. The emotional summary is advisory and always attributed to the model, never to the patient.

  • Status: Pilot design with healthcare partners in the UAE

Multi-avatar consilium for complex questions — Healthcare

A single assistant answering a complex health question tends to produce one confident narrative, with no visible disagreement to warn the user.

The Konsilium feature runs several specialist avatars over the same case in parallel — general practitioner, nutrition, lifestyle — and shows where their reasoning diverges instead of collapsing it into one answer.

Users see disagreement explicitly, which is the honest representation of an uncertain case, and are directed to a clinician when the avatars do not converge.

  • Status: Live in the product at /konsilium

Study companion for university students — Education

Study tools optimise for content delivery and ignore motivation collapse, which is what actually ends most self-directed study.

The student assistant tracks engagement signals during sessions — voice energy, response latency, frustration cues — and adapts pacing and tone, shortening a session rather than pushing through it.

Students get sessions that stop before disengagement rather than after it. The programme runs with discounted access for verified students.

  • Status: Available through the student programme

Live Mode with biometric context — Fitness and wellbeing

Coaching advice that ignores physiological state is either too aggressive or too cautious, and users abandon it quickly.

Live Mode combines real-time vision, voice and available biometric input so the coaching avatar can see the difference between effort and strain, and respond in the moment.

Sessions adjust continuously instead of at the end. All biometric processing is opt-in and can be revoked per session.

  • Status: Live in the product at /live-mode