Back to Swatches

Tortus.ai - Clinical Consultation Scribe

Fabric Score 4.8
PublicHealthTortus AIGenAI modelDeveloper ×1User / Operator / Front-line Staff ×1

Workflow Diagram

Tortus.ai - Clinical Consultation Scribe workflow diagram

Fabric Score

ValueEfficacySecurityRiskExternalitiesEfficiency4.95.05.04.64.94.44.8Fabric Score

Task

During a patient–clinician consultation, the AI system generates clinical notes and a referral letter (if needed) for the clinician to review and use.

Intent

The AI system improves clinician efficiency, focus, and attentiveness during the consultation, whilst improving the comprehensiveness, accuracy, and structure of the consultation notes.

AI Workflow

Input

The consultation audio is recorded (with patient consent) via the Tortus application. The clinician can select their preferred note template and style, and (optionally) add additional context as text.

Process

The AI system transcribes the audio, incorporates additional context, and generates consultation notes. Online LLM safety evals are run to monitor note quality. If the clinician deems a referral necessary, the AI system also drafts the referral letter. Automated checks are performed to detect and correct clinical errors in the generated notes and letter (e.g. transcription inconsistencies, ambiguous drug dosing/schedules, missing allergy information). Output that does not meet automated clinical safety thresholds triggers internal audits.

Output

Clinician reviews/edits notes, then adds them to the EHR system once satisfied. If a referral letter was requested, the clinician reviews/edits the letter before sending it. The clinician is responsible for the final clinical artefacts sent to external systems.

Human Oversight Level

Human-Led with AI Assistance

Institutional Oversight Examples

  • •Medical device regulation: UK MDR 2002 Class IIa AI medical device* with software safety processes per IEC 62304 and ISO 14971
  • •Clinical safety: Full compliance with NHS DCB0129 and DCB0160 risk management, supported by Clinical Safety Case, Hazard Log, Risk Management Plan, Incident Log, and appointed Clinical Safety Officer
  • •Data Protection and Privacy: UK GDPR / Data Protection Act 2018, NHS DSP Toolkit, ISO 27001:2022
  • •Quality control: ISO 13485-compliant development and auditing

Risk

Clinicians could become reliant on the AI system, lowering their performance when it is unavailable to them.

Output Modification Telemetry

  • Other32%
  • Factual errors / hallucinations27.5%
  • Irrelevant information27.5%
  • Missing or incomplete information5.5%
  • Unfaithful outputs (policy, tone, or style misalignment)2.5%
  • Poor usability (unclear language)2.5%
  • Internal inconsistency2.5%

Transversal Metrics

Grouped by Fabric dimension.

Efficacy

Accuracy96.5 %

Efficiency

Modification Rate12 %
Modification Time1.5 min/output
Verification Time0.6 min
Rejection Rate1 %
Operational Friction2.5 hrs/wk
Implementation Overhead0.5 hrs
Governance Overhead151 hrs
Time to Launch12.5 wks

Value

Effort Reduction66.5 %

Risk

Reliability0.5 incident/month
Autonomy98.5 %