Healthcare is the most heavily constrained vertical we work in. Patient data is sensitive. Interoperability between systems is poor. Regulatory frameworks (HIPAA in the US, GDPR in the EU, equivalents elsewhere) are non-negotiable. That combination has historically made healthcare a slow adopter of new technology — but it’s exactly those constraints that make it the most interesting place to watch where AI and blockchain are starting to converge.
Why healthcare needs both, not either
AI’s value in healthcare is obvious — diagnostic assistance, clinical triage, admin automation, faster literature search, conversational front-ends for patient intake. But AI only works if the data it operates on is trustworthy, consented, auditable, and (often) doesn’t leave the organisation. That’s where blockchain primitives start earning their place.
We’ve previously written about web3 healthcare use cases and DLT solutions, traceability of records and instructions in healthcare, and the use of IPFS for distributed storage of patient data. The patterns we sketched in 2020–2022 are now the patterns that are actually shipping in 2024.
The patterns starting to ship
- Consent registries on-chain. Patient consent for data use is recorded immutably, with a verifiable audit trail. AI systems consuming the data check the registry before each access.
- Provenance and authenticity for AI-generated content. When an AI system drafts clinical notes or summarises a patient history, the output can be cryptographically signed and anchored, so the downstream reader knows what model produced it, when, and on what inputs.
- Federated learning with on-chain coordination. Multiple hospitals contribute to training a model without their raw data ever leaving the organisation. The blockchain handles attribution, incentives, and audit.
- Patient-controlled identity. Self-sovereign identity primitives let patients (or their guardians) grant, revoke, and audit access to records across organisations — without a central authority holding the master copy.
The chat layer is the underrated piece
What we’ve found from our customer base: the layer that matters most in practice is the messaging substrate. Clinicians collaborate in chat. Patients interact with care teams in chat. AI agents — whether they’re scheduling appointments, summarising lab results, or surfacing relevant guidelines — show up inside the same chat surfaces that humans already use.
That makes the messaging layer load-bearing for both the AI integration and the compliance requirements. It needs to be HIPAA-aware out of the box, support self-hosted or dedicated-server deployment, and give the operator full control over data flow. Generic consumer chat APIs don’t meet that bar; neither do most general-purpose AI agent platforms.
Where it gets interesting in 2025
The next wave is AI agents that act on a patient’s behalf — booking, requesting records, summarising specialists’ notes, even handling micro-payments for second opinions or imaging — while operating inside a chat substrate the patient already trusts. That requires the chat layer, the AI agent layer, the identity layer, and the payment layer to be designed as a single stack.
That’s the integration work we think will define healthcare technology buying decisions over the next two to three years.
Related from Ethora: Ethora healthcare chat SDK — HIPAA-aware chat and AI agent SDK for healthcare.