A queue, not a juggling act

Walk into any ASC business office on a Monday morning and you'll find the same scene: a dozen browser tabs, three EMR windows, a stack of payer portal logins on a sticky note, an inbox full of denials, and a queue of cases that need something — eligibility, prior auth, peer-to-peer, missing clearance, payment plan — before they can go on Thursday.

The work itself isn't the problem. It's the orchestration. Each case touches multiple systems. Each task has a different deadline, different payer rules, different documentation requirements. And the staff doing it are often the same people fielding the inbound patient calls, calling the payer for the previous claim, and training the new coder. The ASC News 2025 Industry Outlook survey found 25% of ASC leaders rank staffing constraints as their single greatest operational challenge — a number that doesn't go down when the business office is buried in administrative triage.¹

The CAQH Index, the industry standard for measuring administrative transaction costs, finds that manual eligibility checks consume about 16 minutes per medical transaction and manual claim status inquiries about 24 minutes.² MGMA's 2025 prior authorization research reports 35% of practices spend upwards of 35 minutes on average per PA request, with at least three employees typically involved in completing a single request.³ The 2024 AMA Prior Authorization Physician Survey puts the per-physician burden at 13 hours per week across 39 authorizations.⁴

Underneath all of that is task switching. Every time a coordinator stops working a denial to pick up an inbound call, moves to a different patient's prior auth, or logs into a different payer portal, the cost isn't just the minutes spent — it's the cognitive overhead of reloading context.

Lily's Task Management agent doesn't replace the people doing the work. It replaces the queue they have to maintain in their heads.

Every task — an eligibility check that's pending, an auth that needs a clinical document, a patient who hasn't responded to a pre-op call — surfaces in a single prioritized queue, ordered by case date, payer turnaround time, and risk. For each task, Lily does the upstream work, then presents a recommended next action. The human reviewer sees the full context on one screen, approves with one click, modifies with another, or rejects and tells Lily why.

The time savings are substantial but consistent rather than dramatic. We see, on average, about 30 seconds saved per task — not by skipping steps, but by eliminating the search, login, and copy-paste work that fills the gaps between decisions. At a center processing several hundred tasks a day, that compounds quickly.

This is intentional. The published research on human-in-the-loop AI in healthcare consistently lands in the same place: AI should automate the assembly and the recommendation; the human stays accountable for the decision.⁵ That model is what makes the system both safe and auditable.

A nurse approving a pre-op clearance flag, a biller approving a denial appeal, a coordinator approving an outbound patient call — every action is captured, every decision is logged, every modification trains the system. Lily handles the work that doesn't require judgment. The staff handles the work that does. Neither one is doing the other's job.