From Assistive to Autonomous Brokers
In digital well being, we’ve been right here earlier than. Each new startup guarantees a greater option to handle X — psychological well being, diabetes, scheduling, coverage administration. The outcome? A crowded ecosystem of single-purpose level options that always hit limitations to scale and long-term adoption.
Now, AI brokers are right here — and we’re already seeing the identical sample emerge.
CB Insights AI Agent market map highlights a quickly rising record of AI agent startups in healthcare — many focusing on only a single operate:
- Prior authorization automation
- Coding and documentation
- Scribe instruments
- Profit navigation
- Treatment reminders
Are these true AI brokers — or simply level options in disguise?
- Purchaser fatigue is actual. Payers and well being programs don’t need 50 distributors doing 50 slender duties.
- Fragmented brokers are exhausting to control, prepare, and belief — particularly in regulated workflows.
- Lack of interoperability means these brokers can’t discuss to one another, creating extra silos.
Based on a Wall Street Journal article titled Everybody’s Speaking about AI Brokers. Barely Anybody Is aware of What They Are, Tom Coshow of Gartner suggests a whole lot of what firms are calling AI brokers at present are actually simply chatbots and AI assistants.
AI brokers are programs that may take motion on behalf of people.
Within the machine studying hierarchy of descriptive → predictive → prescriptive analytics, AI brokers function within the prescriptive layer. They don’t simply analyze knowledge — they take motion on it. In healthcare, that would imply:
- Guiding a member to close by supplier choices and scheduling the go to.
- Verifying {that a} prescription was despatched and arranging its supply.
That’s the leap from assistive AI to autonomous brokers.
The chance isn’t in constructing extra siloed brokers — it’s in constructing coordinated, cross-functional brokers that function throughout workflows.
For instance, the workflows beneath characterize how brokers might orchestrate throughout three distinct options:
- Manages consumption, scheduling, and triage
- Handles scribing, documentation, and copay billing
- Opinions contractual charges, compares benchmarks, submits claims, and auto-adjudicates
The winners received’t simply automate — they’ll orchestrate, unlocking actual working leverage throughout the stack.
If there’s one a part of healthcare that would use orchestration, it’s medical trials.
Right now, affected person recruitment, consent, monitoring, and outcomes validation are dealt with by totally different distributors:
- Sponsors depend on entry to EHR knowledge (e.g., Ontada, Truveta) to determine individuals
- Web site engagement, eConsent, and distant monitoring are dealt with by firms like Medable or Science 37
- Consequence validation is one more fragmented step
No unified agent connects these workflows — which may affect medical trial recruitment and dropout charges
Now think about a dynamic orchestration layer:
- An agent identifies eligible sufferers in actual time from EHR knowledge
- Reaches out to sufferers via most popular channels
- Gathers eConsent, verifies ID, and manages permissions
- Coordinates affected person monitoring primarily based on protocol
This isn’t simply assistive automation — it’s clever sequencing throughout the worth chain.
In fintech, APIs are desk stakes. Stripe, for instance, can orchestrate throughout KYC, funds, and fraud detection — as a result of APIs are mature, dependable, and anticipated.
In healthcare? It’s nonetheless developmental.
FHIR — the interoperability normal — varies throughout programs and sometimes solely helps fundamental knowledge pulls. And even so, the permissions are normally restricted primarily based on organizational entry factors and in some instances group/companion innovation maturity or workflows. With out sturdy, dependable interoperability, most AI brokers stay boxed into slender use instances. Till that adjustments, orchestration in healthcare will lag.
And even then, some duties like protocol adherence or medical adjudication might require human oversight — not as a result of AI can’t do the work, however as a result of regulatory, moral, and medical requirements at the moment demand a ultimate human examine.
Autonomous brokers will thrive in healthcare — as soon as builders can ship options that join the total worth chain throughout clinicians, sufferers, and directors.
