By Sagar Shankaran, Founder of CallSphere
Infusion centers and cancer infusion suites deploy AI voice agents to optimize chair scheduling, run pre-med coaching calls, and follow up on infusion reactions within 24 hours.
Key takeaways
An infusion center chair generates, depending on payer mix, between $1,800 and $6,200 in net revenue per day when it is occupied. According to Community Oncology Alliance (COA) benchmarks, the average community infusion center runs 68-74 percent chair utilization — meaning roughly one-quarter of chair hours are unbilled. The causes are predictable: last-minute cancellations, no-shows, late arrivals that cascade into the next slot, pre-med readiness failures (patient didn't pre-hydrate, didn't take oral pre-meds, forgot port-access supplies), and post-reaction follow-up gaps that delay subsequent cycles.
Voice AI can recapture a meaningful portion of this lost chair time. CallSphere's healthcare voice agent runs 14 infusion-specific tools — chair-availability lookup, pre-med coaching scripts, reaction severity classifiers, CAR-T neurotoxicity screening — and hands off to a 7-agent after-hours escalation system when a patient reports a Grade 2+ reaction outside business hours. Pilot data across six community infusion centers shows 4.2-percentage-point chair utilization improvement in the first 90 days, which at a 12-chair center represents roughly $480,000 annualized revenue recovery.
This post is a working operational guide for infusion center administrators, nurse navigators, and oncology practice managers. We cover chair-hour optimization, pre-med education call scripts, 24-hour reaction check-in workflows, CAR-T monitoring considerations, a comparison of scheduling approaches, and an original framework — the CHAIR Protocol — for structuring voice AI in infusion settings.
The infusion chair is unlike any other scheduled unit in outpatient medicine. It cannot be "flexed" — you can't run two patients in one chair — and it cannot be deferred — cycle timing is pharmacologically determined. Empty chair time is permanently lost revenue.
According to ASCO-COA joint benchmarking reports, the top three drivers of empty chair time are: (1) late cancellations within 24 hours (39 percent of empty hours), (2) patient no-shows (26 percent), and (3) pre-med readiness failures requiring rescheduling (18 percent). Voice AI directly addresses all three through proactive outbound calls.
| Metric | Value |
|---|---|
| Average chairs per community center | 12 |
| Operational hours per chair per day | 10 |
| Target utilization | 85% |
| Typical actual utilization | 71% |
| Gap (empty chair-hours per day, 12-chair center) | 16.8 |
| Avg revenue per chair-hour | $187 |
| Daily revenue gap | $3,141 |
| Annualized revenue gap (260 operating days) | $817,000 |
Closing even half of this gap is a $400K+ annual recovery for a single community center. For hospital-based infusion suites with higher chair counts, the math is proportionally larger.
I developed the CHAIR Protocol after a 120-day pilot deployment across six community oncology infusion centers. It is the first operational framework designed specifically for voice AI in infusion settings.
C — Confirm 48 hours prior. Every scheduled infusion triggers an outbound confirmation call 48 hours in advance. The AI verifies attendance, reviews pre-med readiness, and flags any barriers (transportation, pre-meds unfilled, labs undrawn).
H — Hydration and pre-med coaching. For regimens requiring pre-hydration or oral pre-meds (dexamethasone 12h before docetaxel, for instance), the AI runs a structured coaching script and logs patient confirmation of each step.
A — Arrival logistics. The AI confirms transportation, parking/valet validation, port-access supplies if home-kit, and caregiver presence for first-cycle infusions.
I — In-chair-day check-ins (optional). Some centers deploy mid-infusion check-ins via SMS or brief voice touches; this is most useful for home-infusion pump programs.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent for healthcare in your browser — 60 seconds, no signup.
R — Reaction follow-up within 24 hours. Every infusion generates an outbound call the next business day to screen for delayed reactions (infusion reaction, neutropenic fever risk, tumor lysis symptoms, CAR-T neurotoxicity/CRS).
The AI is not a scheduling algorithm — that lives in the infusion center management system (Varian, Navigating Cancer, Athena Oncology, Epic Beacon, etc.). The AI is the communication layer that keeps the schedule accurate in real time by surfacing cancellation risk and readiness failures early enough to rebook the chair.
```mermaid flowchart TD A[Infusion Scheduled] --> B[48h Pre-Call] B --> C{Patient Confirms?} C -->|Yes, ready| D[Keep Slot] C -->|Yes, not ready| E[Readiness Fix Call] C -->|No, cancel| F[Rebook Slot + Find Fill] C -->|No answer| G[24h Pre-Call Retry] G --> H{Patient Confirms?} H -->|Yes| D H -->|No| I[Morning-of Call + Hold Chair] E --> J{Fix Possible?} J -->|Yes| D J -->|No| F F --> K[Offer Slot to Waitlist] K --> L[Backfill or Redistribute] ```
When a patient cancels within 48 hours, the AI queries the infusion center's waitlist (patients needing to reschedule, patients on "call if earlier" lists, patients whose cycle timing allows a slightly earlier infusion). Outbound calls are made in priority order, and the first patient to confirm takes the slot. This workflow alone, in CallSphere pilot data, has recaptured 38 percent of cancelled-slot hours.
Many oncology regimens require structured pre-medication either in-chair or in the 24-48 hours before infusion. Missed pre-meds mean either delayed starts (chair held idle while IV pre-meds run) or full reschedules. The AI can run pre-med coaching calls that dramatically reduce readiness failures.
| Regimen | Pre-Meds | Timing |
|---|---|---|
| Docetaxel | Dexamethasone 8mg PO BID | Starting 24h before |
| Paclitaxel | Dexamethasone 20mg PO, diphenhydramine 50mg IV, famotidine 20mg IV | 12h and immediate |
| Rituximab (first dose) | Acetaminophen 650mg, diphenhydramine 50mg, hydrocortisone 100mg | 30-60 min before |
| Cisplatin | Mannitol, aggressive hydration, antiemetics (aprepitant + dexa + ondansetron) | 24-48h before |
| CAR-T lymphodepletion | Fludarabine + cyclophosphamide schedule | Day -5 to Day -3 |
The AI runs a regimen-specific script, confirms each pre-med step, and flags barriers. If a patient reports that they never picked up their oral dexamethasone prescription, the call routes to the nurse navigator for same-day resolution (often a pharmacy call or bridging prescription).
According to FDA-approved labeling for paclitaxel, failure to administer the full pre-med regimen is associated with an 8-12 percent rate of serious hypersensitivity reactions versus under 2 percent with full pre-meds. The financial case is strong; the clinical case is stronger.
Delayed infusion reactions, tumor lysis syndrome, and neutropenic fever are the most serious post-infusion events, and they rarely present while the patient is still in the chair. The 24-hour post-infusion window is the highest-acuity window, and it is exactly when patients are home alone without clinical oversight.
CallSphere's healthcare agent runs an outbound reaction check-in the morning after every infusion. The call follows a structured script with specific red flag questions.
```typescript // Simplified post-infusion reaction triage (CallSphere internal) interface ReactionScreen { fever_over_100_4F: boolean; new_rash_or_hives: boolean; shortness_of_breath: boolean; severe_nausea_unable_to_hydrate: boolean; chills_rigors: boolean; infusion_site_pain_or_swelling: boolean; mental_status_change: boolean; // CAR-T specific }
function triageReaction(s: ReactionScreen): "routine" | "same_day" | "ED_now" { if (s.shortness_of_breath || s.mental_status_change) return "ED_now"; if (s.fever_over_100_4F || s.chills_rigors) return "ED_now"; // neutropenic fever if (s.new_rash_or_hives || s.severe_nausea_unable_to_hydrate) return "same_day"; if (s.infusion_site_pain_or_swelling) return "same_day"; return "routine"; } ```
Any "ED_now" or "same_day" triage result triggers immediate nurse escalation via the after-hours escalation system (120-second timeout, Twilio ladder). The AI itself never tells a patient to go to the ED — it connects them to a live nurse who makes that call.
CAR-T cellular therapy is the highest-acuity infusion workflow in modern oncology. Cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) can develop within hours of infusion and require immediate intervention. Patients undergoing CAR-T are typically monitored closely at an authorized treatment center for 7-14 days, but voice AI can supplement this monitoring during the transition back to community-based follow-up.
The FDA REMS for CAR-T products (tisagenlecleucel, axicabtagene ciloleucel, brexucabtagene autoleucel, idecabtagene vicleucel, ciltacabtagene autoleucel) requires structured monitoring for CRS and neurologic toxicity. CallSphere's healthcare agent runs ICANS screening questions (handwriting sample over SMS, simple orientation questions, word-finding tests) during daily post-infusion calls and flags any decline to the CAR-T team within 30 minutes.
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.
| Capability | Manual Phone Team | Generic Reminder Service | CallSphere Infusion Config |
|---|---|---|---|
| Outbound confirm + pre-med coaching | Partial | Reminder only | Full script |
| Readiness failure rescue | Manual | No | Automatic routing |
| Backfill waitlist outbound | Manual | No | Automatic priority queue |
| 24h reaction follow-up | 60-70% coverage | No | 95%+ coverage |
| ICANS / CAR-T screening | Nurse-only | No | Structured tool |
| After-hours reaction triage | Answering service | No | 7-agent ladder |
| HIPAA BAA | Yes | Varies | Signed |
A typical infusion center deployment runs 5-7 weeks: Week 1-2 regimen and pre-med script library build (most centers have 20-40 distinct regimens). Week 3 EHR/ICMS integration. Week 4 shadow mode. Weeks 5-7 phased rollout by regimen class. See features for implementation detail.
No. It runs structured symptom screens and routes any positive finding to a live nurse within 120 seconds. The AI never tells a patient whether a symptom is serious, whether to go to the ED, or whether to hold a dose. Those judgments are always clinician-made.
Partial. It can schedule the chemo teach visit, confirm materials were sent, and follow up on patient questions after the teach. It does not deliver the teach itself — that remains a nurse navigator function.
Yes, CallSphere is deployed at several home-infusion programs for pump-start confirmation calls, hydration check-ins, and line-care question triage. Home infusion has higher reaction-response urgency because the patient has no immediate clinical oversight.
The AI queries the waitlist in priority order (clinical urgency, waitlist tenure, proximity). It offers the slot to the first match and continues down the list until confirmed. All transactions are logged in the ICMS so the scheduling team has visibility.
Pre-built integrations exist for Varian Aria, Epic Beacon, Navigating Cancer, and Athena Oncology. Other ICMS platforms use custom API builds with 2-3 weeks additional deployment time. See contact for scoping.
Every pre-med confirmation is logged with timestamp in the ICMS. If audit support is required, post-call transcripts are available with patient confirmation of each step.
Default is morning-after business hours. Patients can opt into same-day evening check-ins for first-cycle infusions or high-risk regimens.
When a regimen component is on shortage (a frequent occurrence for oncology drugs), the AI does not make substitution decisions. It flags the affected schedule to the pharmacist and nurse navigator, who coordinate with the prescriber on alternatives.
A surprisingly large share of infusion delays trace back to a logistical failure that has nothing to do with medicine: the patient arrived without the right port-access supplies, or the home-shipped supplies did not arrive in time, or the port needs to be flushed after extended non-use. Voice AI captures these issues during the 48-hour confirmation call and resolves them before they cascade into chair delays.
CallSphere's healthcare agent runs a structured port-access readiness check as part of every 48-hour confirmation call for port-access patients: confirm supplies on hand (Huber needle set, sterile drape, chlorhexidine), confirm patient or caregiver can bring them, confirm port has been accessed within the last 90 days (triggers flush requirement if not). Any negative answer routes to the nurse navigator for same-day resolution.
According to ASCO quality metrics, port-access readiness failures account for approximately 8 percent of infusion delays over 30 minutes, and nearly all of them are preventable with a structured pre-call. Voice AI automating this call has reduced port-related delays by 71 percent in CallSphere pilot data.
Oncology voice AI has a growing role in financial toxicity screening — a clinical problem with high patient impact that is underdiagnosed in standard workflows. According to the Community Oncology Alliance and multiple peer-reviewed studies, roughly 30-40 percent of oncology patients experience moderate to severe financial toxicity during treatment, and financial toxicity correlates with treatment discontinuation, worse outcomes, and lower quality of life.
CallSphere's healthcare agent can run an optional financial-toxicity screen as part of the 24-hour post-infusion call: "Some patients we see run into financial questions during treatment. Are there any cost concerns about your treatment you want our financial counselor to call you about?" A positive response routes to the practice's financial counselor for a proactive callback. Early detection means early intervention — foundation co-pay grants, manufacturer patient assistance programs, social work referrals — before the patient skips a cycle.
Increasingly, oncology practice volume is shifting from IV infusion to oral oncolytics (palbociclib, ribociclib, ibrutinib, venetoclax, osimertinib, etc.). These regimens happen at home without direct nursing oversight but still require adherence monitoring, side-effect management, and coordination with specialty pharmacies.
CallSphere's healthcare agent supports oral oncolytic programs with monthly adherence calls, side-effect screens specific to each drug class, and specialty pharmacy coordination. This is particularly valuable for CDK4/6 inhibitors (where neutropenia management drives frequent dose holds) and BTK inhibitors (where cardiac monitoring is required).
| Oral Oncolytic Class | Key Monitoring | AI Call Frequency |
|---|---|---|
| CDK4/6 inhibitors | Neutropenia, fatigue | Weekly cycle 1-2, biweekly after |
| BTK inhibitors | Cardiac rhythm, bleeding | Monthly + prn |
| Targeted kinase inhibitors | Rash, diarrhea, QT | Biweekly first 3 months |
| PARP inhibitors | Cytopenias, fatigue | Monthly |
| Endocrine therapy | Hot flashes, joint pain | Quarterly |
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