CarePlan Copilot - Physician Pre-Visit AI

The Opportunity

Nao Medical is an outpatient healthcare provider whose physicians work the arithmetic every primary care group knows: a 20-minute visit often comes paired with a medical record that takes far longer than that to read. Pre-visit prep meant going through years of encounters, labs, medication changes, and notes. Even careful prep misses the things a chart hides well. A lab value trending upward but still inside the reference range never fires an alert. Referrals from six months ago with no result on file go unnoticed unless someone thinks to look. A regimen unchanged for a year reads as stable on any single visit, even when the measure it manages is drifting. Those misses show up in clinical outcomes and again in the quality metrics and compliance scores the practice is measured and paid on.

The information needed to prepare had always existed in the record, it was just buried under years of encounters. Nao Medical's clinicians were specific about what they would trust: longitudinal trend analysis first, evidence with dates, options they could accept or decline, and nothing executed on their behalf. That set the design constraints for CarePlan: a copilot that reads across visits instead of just the latest one and surfaces what is moving and what was never closed, with every suggestion presented for the clinician to confirm with one click.

Key Challenges

Our Solution

Ajaia built CarePlan as a Chrome extension over the EHR, shaped directly by clinician feedback. Every chart opens with Trends, Insights, and Medications analysis computed over the longitudinal record: dated series for monitored measures, status and attention tiers for each problem-list condition, and medication history with dose-change dates. CarePlan flags measures moving consistently in one direction even when every value is still in range, and surfaces open loops such as a referral placed months ago with no result ever documented.

An on-demand view returns the top three actions for the visit, pulled when the physician asks rather than pushed. Each action is grounded in cited, dated evidence from the record, and fewer than three are returned when fewer are supported. Grounded referral and medication suggestions are framed as options and confirmed by the clinician with one click. Nothing is ordered or executed on the physician's behalf, and suggestions check the active regimen first so nothing proposed contradicts what the patient already takes.

Impact

CarePlan changed what the first minutes of a visit cost. Physicians open the chart to an assembled picture of the trends that are actually moving, medications with their dose history, the loops that never closed, and, on demand, the top three actions for the time available. Prep time fell 85%, and the misses that used to surface downstream, such as in-range creep and referrals that never came back, now surface before the patient is roomed. Quality metrics and compliance scores improved 55%, and the practice banks the quality bonuses tied to measures it now closes visit after visit. Because every suggestion cites dated evidence and requires clinician confirmation, the system is one that providers trust.