Top 3 bottlenecks
Ranked by monthly manual hours. Each card pairs the finding with the automation that addresses it and the hours it could return, because a bottleneck without a recommendation is just a complaint.
Manual hours per month, by task
Bar length = hours/month. Purple = share technology can absorb (automatable); gray = residual human work.
Weekly manual hours (13 weeks, all tasks)
Where the labour cost sits, by role
What-if: automation adoption
Automation never captures 100% of a task: each task has a realistic automatable share (65-90% for scheduling, intake and eligibility; about 50-60% where judgement stays human, like denial rework and billing questions). Drag adoption to model a rollout.
Adoption across automatable work: 60%
FTE = 160 hours/month. Savings use per-role hourly rates (front desk $22, billing $28, clinical $35, admin $30). Freed hours are capacity for patient-facing work, not necessarily headcount reduction. That framing decides whether practice staff adopt or resist the rollout.
Methodology
A 13-week task-event log (6,000+ daily aggregates across 8 practices, 12 task types, 4 staff roles) is loaded into SQLite and analyzed in SQL: the same instrument-then-aggregate approach I used to find 100+ hours/month of reporting effort at a previous client. Each practice's bottleneck profile differs (a GI center drowns in prior-auth; a dermatology front desk in scheduling calls), which is why the recommendation engine is per-practice, not one-size-fits-all.