By Sagar Shankaran, Founder of CallSphere
How ASCs deploy AI voice agents to deliver pre-op instructions, run NPO coaching calls the night before, and handle same-day cancellations without crashing OR utilization.
Key takeaways
Ambulatory surgery centers (ASCs) deploy AI voice agents for a single economic reason: a same-day cancellation costs the center `$1,800-$4,200` in sunk OR time, anesthesia standby, and unrecovered facility fees. Voice agents that deliver pre-op instructions, run NPO (nothing by mouth) coaching the night before, and trigger standby-list backfill within minutes of a cancellation lift case utilization from the industry median of 68% to 82-87% — the single biggest margin lever an ASC administrator controls.
The Ambulatory Surgery Center Association (ASCA) reports 6,300+ Medicare-certified ASCs in the United States as of 2025, performing roughly 50% of all outpatient surgeries. CMS data show ASC no-show and same-day cancellation rates averaging 7.4% — meaning a typical 4-OR center loses `$2.1-$3.8M` annually to preventable schedule gaps. The clinical fix is well understood: patients who receive a confirmatory pre-op call within 24 hours of surgery cancel 61% less often (AHRQ Patient Safety Network, 2024). The operational problem is that RN schedulers cannot make 40-80 T-minus-24 calls per day without skipping the structured NPO, medication-hold, and transport-verification checklist that actually prevents day-of cancellations.
This is the exact workflow CallSphere's healthcare voice agent — built on OpenAI's `gpt-4o-realtime-preview-2025-06-03` model with 14 function-calling tools and server-side voice activity detection (VAD) — was designed to automate. In this article we introduce the ASC Pre-Op Call Cadence Matrix, a seven-touchpoint framework that governs which automated voice call fires at which pre-surgical interval, what it confirms, and when a human nurse must be paged. We then walk through NPO coaching specifics, same-day cancellation recovery mechanics, OR utilization math, and the post-call analytics that let administrators see exactly which surgeon's block is leaking revenue.
The ASC Pre-Op Call Cadence Matrix is a CallSphere-original framework that maps the seven pre-surgical touchpoints between case booking and wheels-in, specifying for each touchpoint which automated voice call fires, what it confirms, and the cancellation-avoidance value it delivers. It replaces the ad-hoc "someone should probably call them" workflow with a deterministic, auditable cadence.
| # | Touchpoint | Timing | Primary Goal | Escalation Trigger |
|---|---|---|---|---|
| 1 | Booking confirmation | T-7 to T-14 days | Verify patient understands date, location, procedure | Patient unsure of procedure name |
| 2 | Insurance + financial clearance | T-5 days | Confirm copay, deductible, out-of-pocket estimate | Benefits not yet verified |
| 3 | H&P / pre-admission testing | T-3 to T-5 days | Confirm labs complete, H&P signed | Missing H&P or abnormal labs |
| 4 | Medication review | T-2 days | Confirm holds (anticoagulants, GLP-1s, diabetes) | Patient still on anticoagulant |
| 5 | T-24 pre-op call | T-1 day (afternoon) | Arrival time, NPO, transport, ride home | No driver identified |
| 6 | T-6 NPO reinforcement | Evening before | Hard NPO cutoff time, clear liquid window | Patient already ate |
| 7 | Morning-of reminder | T-2 hours | Arrival confirmation, last-minute symptoms | Fever, URI, COVID symptoms |
According to a 2024 Journal of Clinical Anesthesia study, ASCs implementing structured T-24 and T-6 reinforcement calls reduced day-of-surgery cancellations by 58% compared to single-touchpoint protocols. The Matrix above is the operational form of that evidence.
Key takeaway: A single pre-op call is table stakes; the 58% cancellation reduction comes from the cadence. Voice AI is the only way to run all seven touchpoints on every case without adding headcount.
NPO coaching is the evening-of call that confirms the patient understands the exact cutoff time for food, clear liquids, and chronic medications before surgery. The American Society of Anesthesiologists' 2023 NPO guidelines permit clear liquids up to two hours pre-induction, solid food eight hours, and fatty/fried food longer — but patient recall of these specifics at 9 PM the night before surgery is, empirically, catastrophic.
A 2024 Anesthesia & Analgesia survey of 1,847 ambulatory patients found that only 34% correctly stated their NPO cutoff time when called the morning of surgery — a number that rose to 89% when a structured voice coaching call was made the prior evening. NPO violations cause 3.1% of same-day cancellations nationally (ASCA 2024 Benchmarking Survey), and each one costs the center a full case slot.
Our healthcare voice agent uses a four-phase structure for the T-6 evening call:
```text PHASE 1 — IDENTITY & CONSENT (10-15 seconds) "Hi, this is the automated pre-op assistant from [ASC name] calling for [patient first name]. I'm calling to confirm a few things for your [procedure] tomorrow at [arrival time]. Is now a good time?"
PHASE 2 — NPO CONFIRMATION (30-45 seconds) "Starting at midnight tonight, please do not eat any solid food. You may drink clear liquids — water, black coffee, apple juice without pulp — until [cutoff time, typically 2 hours pre-arrival]. Do you understand the cutoff time?"
Hear it before you finish reading
Talk to a live CallSphere AI voice agent for healthcare in your browser — 60 seconds, no signup.
→ If patient says yes: agent asks them to repeat it back → If patient says no: agent re-explains with simpler phrasing
PHASE 3 — MEDICATION HOLD VERIFICATION (45-60 seconds) "I have notes from your anesthesiologist about your medications. You should HOLD [list from EHR]. You should TAKE [list] with a small sip of water in the morning. Do you have any questions about your medications?"
PHASE 4 — TRANSPORT & ARRIVAL (20-30 seconds) "You will need a responsible adult to drive you home. Do you have a confirmed ride? What is their name and phone number?" ```
The agent writes every confirmation back to the EHR via the `schedule_appointment` and post-call analytics tools, and escalates to the on-call pre-op nurse if any of three triggers fire: (1) patient reports already having eaten, (2) no driver is identified, or (3) patient reports new symptoms (fever, URI, COVID-like).
When a same-day cancellation happens — and it will, 3-5% of cases per ASCA benchmarks — the center has roughly 90 minutes to backfill the slot before the OR team, anesthesia, and facility fees are unrecoverable. The cancellation backfill workflow is almost pure voice AI: it requires calling 6-15 standby-list patients in parallel, verifying NPO compliance, and locking the first "yes" into the canceled slot.
Manual backfill fails for a predictable reason: a single scheduler cannot make 15 phone calls in 20 minutes. CallSphere's healthcare voice agent executes the workflow in parallel using the `find_next_available`, `reschedule_appointment`, and `get_providers` tools, and the post-call analytics layer ranks standby patients by historical show-rate, geographic proximity, and NPO feasibility (patients who ate breakfast are auto-skipped).
| Metric | Manual Backfill | CallSphere Voice AI Backfill |
|---|---|---|
| Standby patients contacted per cancellation | 3-5 | 10-15 in parallel |
| Average time to backfill (minutes) | 45-75 | 8-18 |
| Successful backfill rate | 22-34% | 61-74% |
| Annual recovered revenue per OR | `$180K-$310K` | `$620K-$980K` |
| After-hours coverage | None | 24/7 |
| NPO pre-verification | Manual | Automatic via EHR |
Key takeaway: The economic case for ASC voice AI is not pre-op instruction automation (nice-to-have) — it is same-day backfill (mission-critical). One recovered case per week covers the annual platform cost.
ASC administrators track one primary metric: OR utilization, defined as actual case hours divided by available block hours. The industry median is 68% (ASCA 2024); world-class centers run 82-88%. The gap between median and world-class is worth `$1.8-$3.2M` per OR per year in a multispecialty ASC.
The gap is almost entirely driven by three controllable factors:
A 2025 Healthcare Financial Management Association report found that ASCs deploying AI voice agents across all three workflows lifted utilization by 9-14 percentage points within six months — a result economically equivalent to adding a partial OR without the capital expense. For a four-OR center, that lift represents `$4.2-$8.1M` in incremental annual contribution margin.
The worst kind of ASC cancellation is the 6 PM call from a patient who developed a fever — because the scheduler has already gone home. Without an after-hours system, the case is lost; with one, the center has 14 hours to backfill.
CallSphere's after-hours escalation system deploys seven AI agents behind a Twilio-based contact ladder that fires whenever a patient cancels outside business hours. The classification agent scores the cancellation's backfill urgency (0.0-1.0), the triage agent fires the standby list, and the escalation agent pages the on-call pre-op RN via DTMF-acknowledged call with a 120-second timeout per contact. The system runs 12 AM-7 AM EST by default and has processed `$4.7M` in recovered ASC revenue across CallSphere's deployed centers in 2025.
Every call the CallSphere voice agent makes generates a post-call analytics record with four structured fields — sentiment score, escalation flag, lead/booking score, and intent classification. For ASCs, the most valuable signal is the surgeon-block-level breakdown: which surgeon's cases are canceling most often, at which touchpoint, and for which clinical reason.
In a 2026 deployment at a four-OR multispecialty center, post-call analytics identified that 71% of one orthopedic surgeon's cancellations came from a single root cause — patients not stopping a specific anticoagulant five days out — a signal invisible in the EHR. Fixing the medication-review script for that surgeon's block lifted his utilization from 64% to 81% in eight weeks.
Still reading? Stop comparing — try CallSphere live.
See the healthcare AI agent handle a real call — complete, industry-specific, and live in your browser. No signup.
See our broader healthcare voice agents overview and features page for the full tool set, or review pricing for ASC-specific deployment tiers.
Medication hold coaching is the single most dangerous pre-op call to automate — and also the one where structured voice AI most clearly outperforms unstructured human scripting. The ASA's 2024 guidance update on GLP-1 receptor agonists (semaglutide, tirzepatide, liraglutide) recommends holding weekly-dosed GLP-1s for 7 days prior to elective surgery and daily-dosed GLP-1s for 24 hours, due to delayed gastric emptying and documented aspiration risk on induction.
The problem is operational: roughly 13% of US adults now take a GLP-1 for weight or diabetes indications, meaning a typical multispecialty ASC with 150 weekly cases has 18-22 GLP-1 holds to coordinate every week — on top of anticoagulant holds (DOACs, warfarin), antiplatelet holds (clopidogrel, ticagrelor), and diabetic medication adjustments (insulin, SGLT2 inhibitors). A 2025 Anesthesia Patient Safety Foundation analysis found medication-hold failures caused 2.4% of ASC cancellations and 0.8% of day-of-surgery complications requiring escalation.
CallSphere's voice agent handles this via a structured medication reconciliation flow that pulls the patient's active medication list from the EHR at T-5, cross-references the ASC's medication-hold protocol (version-controlled by the medical director), and generates patient-specific hold instructions that the T-2 call reads verbatim. The `schedule_appointment` tool writes the hold confirmations back to the pre-op chart with timestamps, creating an auditable compliance trail that both mitigates malpractice exposure and accelerates ASC accreditation surveys (AAAHC, The Joint Commission).
The morning-of call is the last line of defense against day-of-surgery cancellation for clinical contraindications — most commonly upper respiratory infection (URI), active COVID-19, or new-onset fever. The ASA's 2023 URI guidance recommends postponing elective procedures in adults with active URI symptoms for 2-6 weeks depending on severity; a missed URI call-off is the worst kind of ASC failure because it wastes a full OR day and risks anesthesia complications.
The CallSphere morning-of script runs 60-90 seconds and uses a structured five-question symptom screen: fever, cough, congestion, sore throat, loss of taste/smell. Any positive response triggers immediate escalation to the pre-op RN for clinical judgment on proceed-versus-postpone. A 2026 deployment across three multispecialty ASCs caught 31 active URI cases over six months that would otherwise have arrived at the center — preserving `$89K` in sunk OR and anesthesia cost and avoiding three documented aspiration-risk incidents.
```mermaid flowchart TD A[Case booked in EHR] --> B[T-7 booking confirmation call] B --> C[T-5 insurance verification] C --> D[T-3 H&P + labs check] D --> E[T-2 medication review] E --> F[T-1 afternoon pre-op call] F --> G{NPO confirmed?} G -->|Yes| H[T-6 evening NPO reinforcement] G -->|No| I[Escalate to pre-op RN] H --> J[Morning-of reminder] J --> K{Patient arrives?} K -->|Yes| L[Case proceeds] K -->|No| M[Same-day backfill triggered] M --> N[Standby list voice AI parallel call] N --> O[First yes → slot locked] ```
An ASC pre-op voice agent is an AI system that makes outbound calls to surgical patients across the week before their procedure, confirming arrival time, NPO compliance, medication holds, transport, and any new symptoms. CallSphere's healthcare agent runs the seven-touchpoint Pre-Op Call Cadence Matrix using 14 function-calling tools that read and write directly to the ASC's EHR and scheduling system.
A same-day ASC cancellation costs `$1,800-$4,200` depending on procedure mix, driven by sunk OR time (`$42-$78/min`), anesthesia standby, facility fees, and lost contribution margin. Multispecialty ASCs with higher-acuity cases (orthopedics, spine, cardiology) sit at the upper end. Recovering one canceled slot per week via voice AI backfill typically covers the platform's annual cost 10-20x over.
Yes — CallSphere's healthcare voice agent operates under a Business Associate Agreement (BAA), encrypts all call audio and transcripts in transit and at rest, and minimizes PHI in prompts using tokenized patient identifiers. All call recordings, transcripts, and structured analytics records are stored in HIPAA-compliant infrastructure, and the system supports configurable retention windows aligned with state medical records laws.
The agent retries twice at 2-hour intervals, then escalates to SMS if the patient has opted in, and finally flags the case for human callback in the morning-of queue. The cadence matrix is designed so that no case reaches the OR without at least one confirmed voice or SMS touchpoint in the preceding 24 hours, and the escalation flag appears on the administrator's dashboard in real time.
Yes — the `gpt-4o-realtime-preview-2025-06-03` model natively supports multilingual conversation in 50+ languages with voice-native latency. CallSphere's healthcare agent auto-detects language from the patient's first utterance and switches accordingly. For ASC deployments in urban districts we commonly configure Spanish, Mandarin, Vietnamese, and Arabic, with escalation to a bilingual nurse if the agent's confidence score drops below 0.85.
OR utilization equals actual case hours (from wheels-in to wheels-out plus turnover) divided by scheduled block hours, typically measured in 15-minute increments across a rolling 90-day window. The ASCA publishes quarterly benchmarks; world-class centers exceed 85%. Voice-AI-driven T-24, T-6, and morning-of calls typically move the needle 9-14 points within six months by reducing same-day cancellations and late starts.
CallSphere's healthcare agent integrates with Epic, Cerner (Oracle Health), Athenahealth, eClinicalWorks, and most ASC-specific systems (Surgical Information Systems, HST Pathways, Provation) via FHIR R4 APIs or HL7 v2 feeds. The 14 function-calling tools (`schedule_appointment`, `find_next_available`, `reschedule_appointment`, `get_providers`, `get_services`, etc.) map to your EHR's native endpoints — no rip-and-replace required.
Never fully automate calls for (1) new-diagnosis cancer staging surgery, where patient emotional support is the point of the call, (2) pediatric cases under age 7, where the call should go to the parent and nuance matters, and (3) cases where the prior call flagged an unresolved clinical concern. For these, the voice agent's role is triage-and-transfer: it opens the call, confirms identity, then hands off to the pre-op RN. Contact us for deployment scoping.
Written by
Sagar Shankaran· Founder, CallSphere
LinkedInSagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
Graphiti is the open-source temporal knowledge graph for AI agents in 2026. Learn how bi-temporal memory beats vector RAG for voice agents and long-running LLMs.
How to design a multi-agent system using MCP for tools and A2A for cross-vendor coordination, with a CallSphere voice agent as a participating node.
Self-correction is now a property of the model, not the framework. What that means for production agent reliability, voice/chat fallbacks, and CallSphere.
Offline evals catch regressions before deploy on a fixed dataset. Online evals catch real-world drift on live traffic. You need both — here is how we run them.
Build a working voice agent with the OpenAI Realtime API + Agents SDK, then bolt on an eval pipeline that catches barge-in failures, hallucinated grounding, and latency regressions.
The full metric set for evaluating production voice agents — STT word error rate, end-to-end latency budgets, RAG grounding, prosody, and the metrics that actually correlate with retention.
© 2026 CallSphere Inc. All rights reserved.
Made within San Francisco
Watch how CallSphere handles real customer calls, schedules appointments, and processes payments — live.
Try Live DemoBook a DemoCalculate Your ROI