Informaticists convert intent into safe work.
Clinicians and operators describe outcomes; agents need constraints. Informaticists translate requests into workflow logic, data boundaries, exception rules, training needs, and acceptance criteria.
The next healthcare AI leader sits between clinical reality and machine execution
Executives will own strategy and risk. IT will own platforms, identity, integration, and security. But the people who can turn agentic AI into safe operational capacity are the informaticists who understand what clinical work actually means inside the system.
They manage agents like operational workers.
Agents will prepare drafts, route issues, assemble context, and recommend next actions. Informaticists define what good looks like, review failures, tune playbooks, and decide when automation needs a human stop.
They design the approval layer.
The winning pattern is prepare, propose, approve, execute, audit. Informaticists help decide which workflows can be automated, which must be approval-gated, and which should remain off-limits.
Why executives should fund this function now
Agentic AI will appear across EHR roadmaps, Microsoft and cloud platforms, ambient documentation, contact centers, scheduling, service desk, revenue cycle, inboxes, portals, and local innovation projects. Without a mature informatics layer, every department will invent its own rules for risk, workflow, and accountability.
- The bottleneck is not model access. The bottleneck is knowing which work is safe to automate, how to review it, and how to measure whether it actually improves care operations.
- Clinical trust is the scarce asset. Agents that feel wrong, create rework, or bypass workflow reality will lose credibility fast.
- Application knowledge is not enough by itself. The new role combines EHR expertise, clinical operations, data literacy, AI supervision, and governance design.
- Shadow AI is the default path. If leaders do not build a sanctioned model, motivated teams will assemble their own.
The 6 / 12 / 24 month informatics horizon
A practical forecast for health systems where clinical applications, operations, and AI strategy are starting to converge.
Informaticists become AI translators.
The first useful work is not autonomous care. It is preparing and reviewing work around care.
- Map high-volume workflows suitable for draft-only assistance
- Write acceptance criteria for agent outputs
- Define escalation triggers, no-go zones, and human review points
- Partner with security, compliance, IT, and operations on a shared risk vocabulary
Informaticists supervise agent workflows.
Agents become assigned helpers with constrained tools, approved data, and observable work queues.
- Manage pilot playbooks for intake, scheduling, support, follow-up, and clinical-apps prep
- Review agent failures and improve prompts, policies, and workflows
- Train operational teams to delegate safely and verify outputs
- Measure throughput, exception rate, rework, user trust, and safety signals
Informaticists become AI governance operators.
The role moves from project support to standing operational capability.
- Own the bridge between enterprise AI policy and frontline workflow reality
- Help govern which agents can propose, draft, schedule, message, or execute
- Shape new analyst, informatics, and operations roles around agent supervision
- Turn local pilots into repeatable, auditable operating patterns
The leadership agenda
The executive job is not to buy a chatbot. It is to build the conditions where informaticists can safely turn agents into capacity.
1. Create an AI informatics lane
Name the function explicitly. Give it authority to define workflow suitability, acceptance criteria, approval patterns, and operational measures.
2. Start with prepared work
Prioritize summaries, routing, draft responses, intake packets, follow-up scripts, test plans, and review queues before granting execution authority.
3. Build approval architecture
Classify workflows as auto-draft, approval-gated proposal, supervised execution, or prohibited. Make the gate visible and auditable.
4. Measure capacity and trust
Track backlog movement, turnaround time, rework, escalation quality, safety signals, adoption, and clinician/operator confidence.
A safe informatics-led pilot pattern
1
Choose one workflow with clear boundariesScheduling intake, service desk, benefits questions, discharge follow-up, or clinical-apps prep are better first pilots than open-ended clinical decision support.
2
Let agents prepare, not decideHave the agent assemble context, draft outputs, flag exceptions, and recommend next steps.
3
Put informaticists in the review loopUse their workflow knowledge to evaluate whether the agent output is operationally correct, clinically sensible, and safe to scale.
4
Leave receipts and improve the playbookEvery draft, review, approval, exception, and correction should improve the next version of the workflow.
The informaticist becomes the scaler, not the blocker.
Good governance should not mean “no.” It should mean the organization knows which work can move faster, which work needs review, and which work should not be automated. Informaticists are positioned to make that distinction because they understand both the clinical stakes and the system mechanics.
The real enterprise asset is the pattern. Once a health system learns how informaticists constrain, review, audit, and improve agent work, the same model can expand across departments.
What this means for senior leaders
This is not only a CMIO or CIO issue. Agentic healthcare work crosses clinical operations, applications, security, compliance, workforce design, patient access, and financial performance.
- CIO / CTO: fund the platform, identity, integration, observability, vendor posture, and durable governance model.
- CMIO / CNIO: protect clinical workflow integrity, trust, safety, adoption, and frontline usability.
- COO / VP Operations: identify where agent-prepared work can improve access, throughput, handoffs, and administrative load.
- VP Applications / Informatics: redesign analyst work around agent supervision, approval queues, exception handling, and release governance.
- Compliance / Security / Legal: define data boundaries, audit standards, procurement expectations, and incident response before pilots spread.
Signals worth watching
The exact timeline will vary by health system, but the direction of travel is visible: EHR vendors, cloud platforms, ambient documentation, and healthcare operating models are all moving toward AI-assisted work.
Microsoft Dragon CopilotClinical workflow AI and ambient documentation moving deeper into health-system work.
Epic AIEpic's public AI direction across clinical, administrative, and workflow use cases.
Epic agent factory reportingCoverage of health systems building agents that orchestrate EHR-adjacent work.
Deloitte healthcare agentic AI surveyHealthcare leaders are increasing investment and expecting operational impact from agentic AI.
This page is directional, not clinical, legal, security, or compliance guidance. Patient-impacting actions require authorized review, validation, and local governance.