Ask a care manager where the week goes and the answer is rarely "with members." It goes to reading assessments, reconciling medication lists, scanning hospital records and translating all of it into a care plan that satisfies clinical standards, regulatory requirements and auditors. A comprehensive plan for a complex member can take hours to write. Multiply that across a caseload and care planning becomes the bottleneck that limits how many members a program can actually engage.
AI-assisted care planning changes where that time is spent. Instead of composing from a blank page, the clinician starts from a structured draft built from the member's own data, then reviews, edits and finalizes it. Done well, this does not take judgment out of care management. It moves judgment to the part of the work where it matters most.
Why Care Plans Became the Bottleneck
Care plans have grown more demanding for good reasons. For special needs plans, CMS regulations codified in the 2026 Medicare Advantage final rule require an individualized care plan within 90 days of the initial health risk assessment or 90 days after enrollment, whichever is later. The plan must be person-centered, developed by an interdisciplinary team with the member's active participation where feasible, include measurable goals and specific services, and be updated when the member's health status changes or a care transition occurs. Commercial, Medicaid and value-based care programs carry similar expectations through accreditation standards and contracts.
Meeting those standards requires pulling together information that lives in different places: the health risk assessment, claims history, lab values, pharmacy fills, prior care plans, hospital discharge summaries and the notes from every conversation with the member. Much of a care manager's time goes to assembly rather than to clinical reasoning.
What an AI-Drafted Care Plan Should Contain
A useful draft is not a summary. It is a structured, editable plan that mirrors how an experienced clinician would build one:
- Problems grounded in specific findings, such as an elevated A1c trend, a positive depression screen, a fall reported in the assessment or a gap in medication fills.
- SMART goals for each problem: specific, measurable, achievable, relevant and time-bound. "Improve diabetes control" becomes "reduce A1c below 8 percent within six months, with a repeat lab at three months."
- Interventions assigned to the right role, whether a care manager, pharmacist, social worker or primary care provider, with a target date.
- Justifications that cite the source data behind each recommendation, so the reviewer can see why a problem was proposed and verify it quickly.
The justification layer is what separates a trustworthy draft from a plausible one. A clinician should never have to guess where a recommendation came from.
The Inputs That Matter
Drafts are only as good as the data behind them. The strongest results come from combining structured and unstructured sources: completed assessments and validated screenings, diagnoses and problem lists, medications with actual fill history, recent utilization, and transcripts from calls with the member. Call transcripts are often the most underused input. A member may mention in passing that they stopped a medication because of cost, that they lack a ride to dialysis or that their caregiver is overwhelmed. Those details rarely make it into structured fields, yet they are exactly what an individualized plan should address.
Clinicians Stay in Control
The operating model matters as much as the technology. AI drafts; clinicians decide. Every problem, goal and intervention should be reviewable, editable and removable, and nothing should reach the member record until a licensed or credentialed reviewer finalizes it. The member's own priorities still have to shape the plan, which is why the draft is a starting point for the care conversation rather than a substitute for it. Organizations should define in policy who may approve plans, which elements require clinical sign-off and how edits are captured for quality review.
The Capacity Math
At a regional health plan with more than one million covered lives, Curitics customers saw average care plan generation time fall from roughly three hours to under 15 minutes, a 92 percent reduction. The operational meaning is straightforward. Consider an illustrative care manager who completes eight new or substantially revised plans a week. At three hours each, that is a full 24 hours, more than half the work week. At 15 minutes each, it is two hours. The recovered time can go to outreach, member education, transitions of care and the follow-up calls that actually move clinical outcomes.
There are financial implications as well. Programs can absorb enrollment growth without proportional hiring, meet regulatory timelines for new members more reliably and reduce overtime and backlog during peak enrollment months.
Consistency Is the Underrated Benefit
Speed gets attention, but consistency may be the larger gain. Hand-written care plans vary widely by author, from detailed and measurable to vague and generic. Variation creates audit exposure, makes performance hard to compare across teams and leaves some members with weaker plans simply because of who was assigned to them. A structured drafting process applies the same clinical logic to every member, so the baseline quality of plans rises and reviewers can focus on clinical nuance rather than missing elements.
A Governance Checklist for AI-Assisted Care Planning
- Define the review standard. Specify who finalizes plans and what they must verify before sign-off.
- Require traceability. Every proposed problem and intervention should link to the source data that supports it.
- Measure acceptance and edits. Track how often drafted elements are accepted, modified or rejected, and use the patterns to refine templates and training.
- Audit a sample. Periodically compare finalized plans against source records and regulatory elements, including measurable goals and member participation.
- Monitor outcomes, not just time. Track engagement rates, goal attainment, readmissions and member experience alongside time saved.
- Protect the data. Confirm that any platform handling assessments and call transcripts meets rigorous security standards and that access is role-based.
How a Platform Approach Helps
Care planning works best when the plan is built in the same place the data lives. In Curitics, AI-assisted care plan drafting pulls from assessments, clinical data and call transcripts captured through connected telephony, proposes problems, SMART goals and interventions with justifications, and routes the draft to clinicians for review and finalization. Medication reconciliation and HEDIS gap information sit alongside the plan, so interventions reflect what the member actually fills and what quality gaps remain open.
The goal is not to automate care management. It is to give care managers back the hours that documentation has quietly consumed, and to make every member's plan as strong as the best plan your team would write on its best day.
