Telehealth and AI are often discussed separately. But they’re natural complements: telehealth creates data streams that AI can analyse, and AI can address some of telehealth’s inherent limitations.

Since COVID accelerated telehealth adoption, I’ve been watching where AI adds genuine value to remote care. Here’s what I’ve observed.

Where AI Enhances Telehealth Now

Pre-consultation preparation. AI can review patient records before a telehealth consultation and generate a concise summary for the clinician. Relevant history, recent results, current medications, outstanding care gaps—all available when the call begins.

This is particularly valuable when telehealth providers may not know the patient well. A specialist doing a video consultation can start with context they wouldn’t otherwise have.

Real-time transcription and documentation. AI transcription during telehealth consultations, followed by generative AI drafting consultation notes, is one of the most immediately useful applications.

The workflow: consultation proceeds normally, AI generates draft documentation, clinician reviews and approves (or edits) after the call. Documentation overhead drops significantly.

Patient symptom assessment before consultation. AI chatbots that guide patients through symptom assessment before their telehealth consultation can improve efficiency. The consultation starts with structured information rather than open-ended history-taking.

These aren’t diagnosis systems—they’re structured data collection. The clinician still assesses and decides. But they have better information to start with.

Waiting room triage. When telehealth services have queues, AI can assess waiting patients and prioritise those with concerning symptoms. This mirrors emergency department triage adapted for virtual care.

What’s Not Ready Yet

Automated diagnosis. AI that diagnoses patients via telehealth without clinician involvement isn’t ready for deployment and probably shouldn’t be, even when technology improves. Diagnosis requires contextual judgment that current AI lacks.

Physical examination augmentation. Despite exciting research, AI that guides patients to perform physical examinations or interprets patient-captured images reliably isn’t ready for clinical use outside specific narrow applications.

Mental health assessment. AI analysis of speech patterns, facial expressions, or language to assess mental health state is being researched but isn’t validated for clinical use. The risks of getting this wrong are significant.

Chronic disease management automation. AI systems that independently manage chronic conditions (adjusting medications, ordering tests, scheduling follow-up) require human oversight. Augmentation, not automation.

Technical Integration Considerations

Implementing AI in telehealth isn’t just about the AI—it’s about integration:

Video platform integration. Your AI needs to work with your telehealth platform. If they’re separate systems, clinicians end up with multiple windows and workflows. Native integration is better than bolt-on.

EMR integration. AI that can read from and write to your EMR is far more useful than AI that operates in isolation. Integration standards (FHIR) help here but aren’t universal.

Audio quality requirements. AI transcription depends on audio quality. Telehealth connections vary. Test how AI performs with typical (not optimal) connection quality.

Consent management. Recording telehealth consultations for AI analysis requires patient consent. This needs to be managed within the workflow, not as an afterthought.

Patient Experience Considerations

AI in telehealth affects patient experience:

Disclosure. Patients should know when AI is involved in their telehealth care. A brief explanation builds trust; undisclosed AI creates suspicion if patients discover it.

Chatbot interactions. Pre-consultation AI chatbots need to be well-designed. Poor chatbot experiences frustrate patients before they even see a clinician.

Accessibility. AI interfaces need to work for diverse patients, including those with limited digital literacy, language barriers, or disabilities. Accessibility testing matters.

Data concerns. Patients may have questions about how their telehealth data is used by AI systems. Clear, honest answers are essential.

Implementation Approach

If you’re integrating AI into telehealth:

Start with clinician-facing applications. AI that helps clinicians (documentation, summarisation, decision support) is lower-risk than AI that directly interacts with patients. Build experience before expanding.

Pilot with willing clinicians. Find telehealth providers who are genuinely interested in AI augmentation. Their feedback will be more useful than feedback from reluctant participants.

Measure time savings honestly. The promise of AI is efficiency. Measure actual time impacts—consultation length, documentation time, clinician satisfaction. Don’t assume benefits; verify them.

Monitor quality. AI transcription and documentation have error rates. Monitor accuracy through regular audits. Errors in medical records have consequences.

Plan for failure modes. When AI doesn’t work (network issues, poor audio, system errors), clinicians need fallback workflows. Don’t create dependency on systems that might fail.

The Telehealth + AI Value Proposition

When I think about why AI matters specifically for telehealth, three points stand out:

Compensating for limited examination. Telehealth’s biggest limitation is inability to perform physical examination. AI can’t fully compensate, but better data aggregation, symptom analysis, and patient history review partially offset this limitation.

Scalability. AI-augmented telehealth can handle more patients with the same clinician resources. As telehealth demand grows, this scalability matters.

Consistency. AI can help ensure telehealth consultations consistently capture relevant information, follow appropriate protocols, and generate complete documentation. Human performance varies; AI augmentation can reduce harmful variation.

These benefits aren’t automatic. They require thoughtful implementation. But the potential is real.

What’s Coming

Near-term developments I’m watching:

Ambient documentation improving. AI that can generate quality clinical notes from natural telehealth conversation is getting better. This will become standard for telehealth within two to three years.

Integrated symptom checkers. Pre-consultation AI assessment will become more sophisticated and better integrated with telehealth platforms.

Remote monitoring integration. AI analysis of patient-collected data (wearables, home monitoring devices) combined with telehealth consultations will create more comprehensive remote care models.

Specialist decision support. AI that helps generalist telehealth providers manage conditions that would otherwise require specialist referral could expand telehealth scope.

Telehealth is here to stay. AI will increasingly be part of how it’s delivered. The organisations building this integration thoughtfully now will have advantages as the field matures.

---

Dr. Rebecca Liu is a health informatics specialist and former Chief Clinical Information Officer. She advises healthcare organisations on clinical AI strategy and implementation.