AI-Powered Consultations

Ceveta's AI consultation feature is available on any appointment. When you start a consultation, the system activates your microphone, streams audio to a medical-grade transcription engine, and produces a real-time transcript with speaker identification.

Once the consultation is complete, AI analyses the transcript to generate structured clinical summaries, extract key findings, suggest clinical codes, identify potential drug interactions, and create follow-up task lists.

All consultation data — transcripts, audio recordings, summaries, findings, and clinical codes — is encrypted using AES-256-GCM before storage. Audio recordings are stored in encrypted cloud storage. No patient data is used to train AI models.

To start an AI consultation, open any appointment and click the Start Transcription button in the AI Consultation card. Your browser will request microphone permission — this is required for real-time transcription.

You can pause and resume recording at any time. When you pause, the transcription stream is paused but your progress is preserved. The elapsed timer tracks total active recording time.

As you speak, the transcript appears in real-time on the left panel. Partial results (shown in lighter text) update as the AI refines its understanding, then finalise into permanent transcript lines.

The waveform visualiser shows live audio input levels, confirming that your microphone is active and capturing speech. A pulsing red indicator in the header confirms recording is in progress.

Alongside the transcript, you can type manual notes in the right panel. These notes are saved with the consultation and included when AI generates summaries, giving you a way to add observations that may not be spoken aloud.

Ceveta automatically identifies different speakers in the consultation. Each speaker is labelled (e.g. Speaker 0, Speaker 1) and their dialogue is visually distinguished in the transcript with different background colours.

Typically, Speaker 0 is the clinician and Speaker 1 is the patient. The AI summary engine uses these speaker labels to better understand the clinical context — distinguishing between what the patient reports and what the clinician observes or recommends.

After stopping the recording, click Finish to generate an AI summary. The AI analyses the full transcript, manual notes, patient profile (including allergies and demographics), and appointment context to produce a structured clinical document.

Built-in summary formats:

You can regenerate the summary in a different format at any time. Summaries can be saved as documents with your organisation's letterhead and exported as PDF.

Beyond the built-in formats, you can create your own custom summary templates. Go to Settings → AI Templates to define templates with custom AI prompts that control the format, tone, and structure of generated summaries.

Custom templates appear in the format dropdown alongside the built-in options. This is useful for specialty-specific formats, medico-legal reports, or organisation-specific documentation standards.

After a consultation is completed, Ceveta's AI analyses the summary to suggest relevant clinical codes. Both ICD-10 and SNOMED CT codes are supported.

Each suggested code includes a description and confidence level (high, medium, or low) to help you quickly assess relevance. Codes are displayed as colour-coded badges — ICD-10 codes in blue, SNOMED codes in purple.

During an active consultation, Ceveta monitors the transcript in real-time and cross-references any mentioned medications against the patient's recorded drug allergies.

If a potential conflict is detected, a prominent red alert banner appears immediately. The system checks over 200 common UK medications and includes awareness of drug class cross-reactivity — for example, if a patient is allergic to penicillin, mentions of amoxicillin, flucloxacillin, or co-amoxiclav will trigger an alert.

Alerts show the specific medication mentioned, the patient's recorded allergy, and the type of match (exact or cross-reactivity). This gives clinicians an immediate safety signal without interrupting the consultation flow.

AI automatically extracts structured clinical findings from the consultation transcript. Findings are categorised into:

Follow-up tasks are also extracted automatically. These include referrals to make, investigations to order, prescriptions to write, and follow-up appointments to book. Tasks appear as a checklist that you can tick off as you complete them.

When medications are discussed during a consultation, Ceveta's AI can extract structured prescription data including:

• Medication name — the drug being prescribed
• Dose — the amount per administration
• Frequency — how often to take (e.g. twice daily)
• Duration — the treatment course length
• Route — oral, topical, intravenous, etc.
• Notes — additional instructions or context

Extracted prescriptions appear in an editable table. You can modify any field before saving. Click Save as Prescription to create a formal prescription document that can be exported as PDF.

For returning patients, Ceveta can compare the current consultation with previous ones. The AI identifies changes in symptoms, new diagnoses, medication changes, and treatment progress.

This is particularly valuable for chronic condition management, post-operative follow-ups, and any scenario where tracking change over time matters. The comparison is generated as a structured narrative highlighting what has changed, improved, or requires attention.

Every AI-generated summary can be saved as a document in Ceveta's document management system. Documents are automatically linked to the appointment and include:

• Your organisation's letterhead (header and footer)
• Clinician's digital signature
• PDF export with professional formatting
• Secure sharing via time-limited links with download limits
• Access via the patient portal

Audio recordings can be played back from the consultation view. The audio player includes playback controls and is available whenever a recording exists for the consultation.