The next wave of AI isn't chatbots that answer questions. It's agents that take actions.

In healthcare, that means AI systems that don't just provide information but actively participate in care: scheduling appointments, coordinating referrals, monitoring conditions, managing medications, conducting research. These agents will need something most current AI systems lack: access to real-time patient medical records.

The HTI-5 proposed rule, released in December 2025, explicitly creates pathways for this future. If you're building AI for healthcare, this is the moment to pay attention.

From Assistants to Agents

Most AI healthcare applications today are assistive. Chatbots that answer health questions, tools that summarize medical literature, systems that help with documentation, analyzers that interpret test results. These applications provide information. Humans take the actions.

AI agents are different. They operate with greater autonomy: an agent that monitors a patient's vital signs and adjusts care recommendations based on trends, an agent that coordinates care across providers and ensures handoffs happen, an agent that screens patients for clinical trial eligibility and initiates enrollment workflows, an agent that manages medication refills by checking interactions and coordinating with pharmacies.

Agency requires data access. To act on behalf of patients, AI systems need access to the information those actions depend on.

What HTI-5 Actually Says

The HTI-5 proposed rule explicitly addresses AI systems accessing health data, proposing updates to the definitions of "access" and "use" to include autonomous AI systems. It articulates a goal of advancing "AI-enabled interoperability solutions through modernized standards and certification."

That language carries real weight. AI data access is being treated as a legitimate use case to enable, not a gray area to tolerate. When AI systems request patient-authorized data, the same information blocking rules apply as for any other request. Providers and EHR vendors cannot treat AI-initiated requests as second-class. And the rule's focus on FHIR APIs and reduced certification burden creates clear incentives for infrastructure that supports AI applications.

If HTI-5 is finalized as proposed, it provides regulatory cover that many AI healthcare builders have been waiting for.

What AI Agents Actually Need

Effective AI healthcare agents require more than occasional data access. The requirements are specific:

Real-time data. Agents that monitor and respond cannot work with stale information. They need APIs that surface current patient status: latest vitals, current medications, recent labs.

Comprehensive data. Agents coordinating care need visibility across the patient's full health picture. Records from one provider are not sufficient.

Patient-authorized access. Agents act on behalf of patients, which requires patient authorization. The consent model must support AI systems holding and using data access tokens over time.

Structured data. Agents that take actions need data they can process programmatically. FHIR resources are agent-friendly; PDFs and unstructured documents are not.

Continuous access. Agents operating over time need persistent authorization, not one-time data pulls.

The Infrastructure Problem

Building AI agents for healthcare means solving the data access problem before you can solve anything else. That means integrations with major EHR systems, OAuth implementations that support long-lived authorization, real-time APIs, normalization across data sources, and security controls appropriate for autonomous systems.

This is infrastructure work. Important, but not what differentiates an AI company whose value lies in the intelligence it provides.

This is what Consolidate Health offers: the data layer that AI agents need. Multi-EHR coverage across Epic, Cerner, athena, eClinicalWorks, NextGen, Flatiron, and Modernizing Medicine through a single integration. Patient-directed authorization flows. Real-time APIs that return current patient data, not historical snapshots. Clean, normalized data regardless of source EHR.

You build the intelligence. We handle the data access.

The Opportunity Window

AI healthcare is competitive. Models improve constantly and features get commoditized quickly. Sustainable advantage comes from data access (better data means better AI), patient trust, and clinical integration that actually fits real healthcare workflows.

HTI-5 opens a window for AI companies to establish data access infrastructure while the regulatory path is clear and before the market crowds. Companies that build for agentic AI now, with robust data access and production-ready infrastructure, will have real advantages as the category matures. Companies that wait may find themselves building data infrastructure under competitive pressure, when speed matters even more.

The Trust Imperative

AI agents operating in healthcare carry significant trust obligations. Patients who authorize AI access to their medical records are extending substantial trust, and that trust has to be earned and maintained.

That means being transparent about what the AI does with data and how. It means requesting only the data needed for the agent's actual function. It means robust security, easy mechanisms for patients to view activity and revoke access, and AI that delivers genuine value in return for the access it's been granted.

Building trustworthy AI agents is not just an ethical obligation. It is a strategic one. Patients share data with systems they trust and revoke access from systems that violate it.

What We're Building Toward

The vision is AI agents that genuinely help patients manage their health: coordinating care across fragmented systems, catching problems before they become crises, extending the reach of healthcare beyond what humans alone can provide.

Getting there requires data access infrastructure, regulatory frameworks that permit AI agency, and patient trust in AI systems. HTI-5 addresses the regulatory piece. Consolidate Health addresses the infrastructure piece. AI companies building today need to address the trust piece.

The components are coming together. The question is who builds the agents that patients actually trust with their data and their health.