Weekly signal

Healthcare AI agents moved from pilot-to-product this week as several vendor platforms and a sector governance report signaled commercial rollouts and operational governance workstreams. Vendor announcements focused on agentic care orchestration for chronic and administrative workflows (patient intake, referrals, payments), while life‑sciences leaders published research showing organizations are actively building AI governance to scale agentic systems safely. Practical deployments emphasize EHR and payer integration, clinician supervision models, and auditability.

What changed

1. Digbi Health launched ASRA, which it calls an "agentic care operating system" for metabolic, inflammatory, and hormonal care (US, employer and health‑plan channel). ASRA bundles precision‑biology data (microbiome, labs, genetics), clinician‑supervised agent automation for enrollment, coaching and medication programs (including GLP‑1 pathways), and explicit controls to prevent customer data from being used to train third‑party models. Digbi positions ASRA as an integrated, enterprise‑grade agent architecture for employers and plans.

2. Andor Health announced a new collaboration (Oct 26) expanding its ThinkAndor agentic platform to bring specialist academic access and virtual nursing capabilities into rural and community hospitals—an example of agentic orchestration layered into teleneurology/tele‑specialty workflows and hospital operations. The release highlights deployment scale and operational integration with hospital systems.

3. Evidently released "Evidently Anywhere": an agents + MCP (message/control/processing) layer that pushes clinical data intelligence and agent capabilities deeper into EHR contexts so clinicians can surface patient complexity where decisions are made. That signals momentum for embedded clinical agents (assistant + orchestration) rather than stand‑alone chat UIs.

4. ACTO published the RAIL (Responsible AI in Life Sciences) Report 2026 showing 64% of life‑sciences leaders are actively establishing AI governance for agentic systems, while a majority still view industry standards as incomplete. The study draws on roundtables and a survey of senior leaders; it introduces practical scoring and governance constructs for agentic workflows. This is a major governance signal for regulated healthcare and pharma buyers.

5. Access & workflow players expanded agent capabilities: Assort Health added digital intake + payments to its patient‑journey agents (Nov 2), and Steer Health promoted a no‑code SteerAI Studio and referral‑automation agent at athenahealth Thrive (announced Oct 29), showing continued commercial focus on access, scheduling and referrals. These are near‑term revenue cycle and access automation use cases.

What to do with it

1. Health systems and payers: prioritize pilot paths for agentic automation where data integrity and escalation rules are simple (referrals, intake, medication adherence), and require audit trails, FHIR/HL7 connectors, and explicit clinician escalation. Use vendor proofs (Digbi, Assort, Steer, Evidently) as templates but insist on role‑based access and model‑training prohibitions.

2. Life sciences/commercial teams: use the RAIL Report to benchmark governance—create or accelerate an AI agent oversight board, require RAIL‑style scoring for any agent touching safety, claims, or regulatory content, and plan operational KPIs for agent performance, escalation, and drift monitoring.

3. Builders and vendors: design agent observability, per‑agent permissioning, and native EHR embedding (SMART on FHIR/HL7) from day one; publish clear clinical ownership and audit logs. Vendors without governance stacks will struggle in enterprise procurement.

4. Security & compliance teams: insist on contractual bans on customer data exfiltration for model training, enable SOC 2/HIPAA controls, and test agents in shadow mode before writeback to EHRs.