The Problem: 15 Minutes of Work After Every Consultation
In veterinary telemedicine, the consultation itself is only part of the work. After each session — whether it takes place over video, voice, or chat — the veterinary expert must write a structured consultation report: what the owner described, what was observed, the anamnesis, and in many cases a diagnosis with recommended next steps.
To do this accurately, vets go back through the full transcript. They review the conversation from beginning to end, extract the relevant clinical details, and structure them into a report. On average, this process was taking up to 15 minutes per consultation — tedious, repetitive work that added significantly to the post-consultation workload without contributing any additional clinical value.
For platforms running dozens of consultations per day, that adds up to hours of avoidable administrative time.
Our Solution: AI-Generated Transcripts and Structured Summaries
We integrated AI-powered processing into our veterinary telemedicine platform to automate the most time-consuming parts of post-consultation reporting. Regardless of consultation type — video call, voice call, or text chat — veterinary experts now receive:
- Full transcript — a complete, readable record of the consultation, automatically generated from video, voice, or chat input
- Structured summary — key points extracted and organized, so the vet can review rather than reconstruct
- Anamnesis section — the reported problem, symptoms, and history, pre-structured and ready for review
- Diagnosis and recommended actions — for consultation types where this is appropriate, the AI proposes a structured starting point that the expert reviews, adjusts, and confirms
The veterinary expert remains fully in control. The AI generates a structured draft — the expert reviews, edits, and approves. The goal is to eliminate the blank-page problem and the manual transcript review, not to replace clinical judgment.
Data Protection: Anonymization Before AI Processing
In any healthcare application, data protection is not optional — and telemedicine platforms handle particularly sensitive information: patient identities, owner details, clinical histories. We gave this significant engineering attention.
Before any content is sent to an external AI service, it is fully anonymized. Personal identifiers — names, contact details, any data that could link the content to a specific individual — are obfuscated. What reaches the AI service is clinical content only, stripped of anything that could be considered personal data under GDPR.
This anonymization layer operates automatically, is configurable per deployment, and leaves a full audit trail. Clinics using the system can be confident that their client and patient data does not leave the system in identifiable form.
The Technical Approach: Factory Pattern for AI Provider Flexibility
One of the key architectural decisions we made was not to hard-code a single AI provider. Different clinics operate under different constraints — some have existing agreements with specific cloud vendors, some have data residency requirements, and some simply prefer one AI service over another.
We implemented a Factory Pattern for AI provider selection. At the settings level, a clinic can configure which AI service powers the summarization and report generation: ChatGPT, Claude, Azure AI, or another compatible service. The underlying processing logic remains identical — only the provider is swapped.
This approach, made possible by our deep experience with modern .NET Core backend patterns, means the feature is fully configurable without any code changes. It also means the platform remains independent of any single AI vendor — an important consideration as the AI landscape continues to evolve rapidly.
Transcription: From Video and Voice to Text
For chat-based consultations, the transcript is already available as structured text. For video and voice consultations, we integrated a third-party transcription service that processes the audio stream directly.
This works seamlessly with our existing video infrastructure. Our platform is built on Twilio Video with WebRTC and video rooms — a stack we have worked with extensively and understand deeply. The transcription service connects to the session output and produces a full text record, which then feeds into the same AI summarization pipeline as chat consultations.
The result is a consistent post-consultation experience regardless of how the session was conducted.
The Result: 56% Reduction in Report Writing Time
After deploying AI-assisted reporting across the platform, post-consultation report writing time dropped from an average of approximately 12 minutes to under 6 minutes per consultation — a reduction of 56%.
For a clinic running 20 consultations per day, that is roughly two hours of saved expert time, every day, redirected from administrative work back to clinical activity.
Fully Configurable, White-Label, Multi-Tenant
The AI reporting feature is entirely optional. Every clinic using our platform can choose whether to enable it, which AI provider to use, and which consultation types it applies to. For clinics that prefer to keep the process fully manual, nothing changes.
This configurability reflects our broader platform philosophy. Our veterinary telemedicine solution is a white-label, multi-tenant SaaS platform — each clinic operates in its own isolated environment, with its own branding, settings, and feature configuration. What works for a large urban practice is not necessarily right for a single-vet rural clinic, and the platform is built to accommodate both.