AI Agent for Pulmonology Clinics: The 2026 US Front-Desk Guide
Pulmonology schedules carry three fragile pieces: a sleep-lab queue where every unfilled bed is $800 to $1,200 in lost contribution margin, a PFT room that ties up a tech for 45 to 60 minutes per patient, and a consult book flooded with PCP referrals for shortness of breath and suspected OSA. This guide shows how a US pulmonology practice puts an AI agent on phone and WhatsApp to protect sleep and PFT slots, triage referrals the same day, and keep patients prepped for the test that got ordered.
Why pulmonology schedules break in a specific way
A pulmonology front desk juggles four workflows at once, each with different lead times, prep requirements, and escalation rules.
- New consults from PCP referrals: dyspnea, chronic cough, abnormal chest imaging, post-COVID symptoms, suspected OSA. Each has a different urgency, but the fax or e-referral lands in a shared inbox that nobody processes until midweek.
- PFT and 6-minute walk visits: often pre-op clearance for a surgeon who needs results by a specific date. A missed PFT holds up the surgery date too.
- Sleep studies, HSAT and in-lab PSG: high-margin, prep-heavy, and the most expensive no-show in the clinic.
- Follow-up and chronic management: COPD titration, asthma step-up, OSA CPAP compliance review, pulmonary hypertension surveillance, ILD monitoring. These live on recall and get dropped first when the phones are loud.
The benchmarks are ugly. A 2026 MGMA Stat poll found 32% of US medical groups reported higher no-show rates year-to-date versus 2025, with rising patient out-of-pocket costs cited as the driver. A March 2026 Medical Economics brief put the national no-show rate around 18% across specialties. An athenahealth case study of Gwinnett Pulmonary & Sleep reports the practice ran at 88% daily completion and a 3.6% no-show rate after adopting scheduling automation, which is what a well-run pulmonology schedule looks like when the front desk is not the bottleneck.
The three most expensive problems in a pulmonology front desk, an empty sleep-lab bed, a pre-op PFT rescheduled past the surgery date, and a PCP referral that took nine days to call back, are scheduling and communication problems. They respond to the same solution.
What a pulmonology AI agent actually does
Treat the agent as a scheduler that never stops picking up, protects sleep and PFT slots with real deposits and confirmations, and writes every action back into the EHR. Five workflows carry most of the weight.
1. Referral triage with specialty-specific urgency rules
When a PCP office sends an e-referral or fax, the agent parses the clinical trigger and applies the practice's urgency rules. Hemoptysis or a chest CT suspicious for malignancy routes to the next urgent new-patient slot with a same-day telehealth bridge. Epworth Sleepiness Scale above 15 with witnessed apneas routes to the sleep queue. Routine chronic cough gets the standard 2 to 4 week target. The agent texts the patient the earliest three options and books the one they pick. The scheduler reviews a worklist of exceptions in the morning instead of answering the phone all day.
2. Pre-test prep for sleep studies, PFTs, and bronchoscopies
This is the single biggest driver of day-of cancellations when it is done poorly. The agent sends a scheduled sequence 7 days, 48 hours, and 3 hours before the appointment, with prep rules per test type.
- In-lab PSG: no caffeine after 2 pm, no napping, bring medications in original bottles, arrival at 8 pm, parking location.
- HSAT: pickup window, device return window, charging instructions, what to do if the oximeter falls off overnight.
- PFT: hold short-acting bronchodilators for 6 hours and long-acting for 12 to 24 hours (per the practice's own list), no heavy meal within 2 hours, no smoking day-of.
- Bronchoscopy: NPO window, anticoagulant hold, driver required, pickup location.
Patients can reply "confirm," "reschedule," or ask a question that routes to a nurse. Every confirmation posts back to the encounter in the EHR, so the morning huddle sees a clean prep-status column instead of a stack of voicemails.
3. Sleep-study deposit-at-booking and waitlist fill
In-lab PSG is a $150 to $250 no-show economically before the tech's hours. The agent offers two paths at booking: a card on file with a disclosed no-show fee, or a pre-authorization released after the study runs. When a bed opens inside 72 hours, the agent broadcasts to a pre-built waitlist and books the first confirmed reply.
4. CPAP compliance outreach and OSA recall
OSA compliance drops hardest at days 30, 90, and 180. The agent reads the AirView or Encore compliance report, flags patients below 4 hours per night on 70% of nights, and reaches out with a nurse-queued offer for a mask refit, pressure review, or follow-up visit. OSA recall is where the real downstream revenue is in a pulmonology practice, and nobody runs it manually anymore.
5. Overflow and after-hours call handling
The agent sits in front of the main line as overflow or after hours. It handles appointment changes, refill requests routed to the right pool, location and hours questions, and billing routing. Clinical questions it does not answer. It takes a structured message with onset, severity, and SpO2 if the patient has a home pulse ox, and routes to the nurse queue. Worsening dyspnea with low SpO2 escalates to a scripted "call 911 or go to the ED" message.
Ninety-day case study: 3-pulmonologist practice on Epic with Natus sleep lab
Southeast practice, two offices plus a 6-bed in-lab sleep center, 8,400 active patients, mixed general pulmonology, sleep medicine, and ILD. Pre-deployment baseline: 20 business days of Avaya call logs, Epic Cadence data, and a Natus SleepWorks utilization report. Deployment: 13 business days to go-live.
Swipe to see the full table →
| Metric | Baseline | Day 90 | Change |
|---|---|---|---|
| Call abandonment rate | 21% | 4% | -81% |
| Median referral-to-contact time | 58 hours | 19 minutes | -99% |
| New-consult no-show rate | 17% | 9% | -47% |
| In-lab PSG no-show rate | 22% | 8% | -64% |
| HSAT return rate (14-day) | 71% | 92% | +30% |
| PFT prep compliance | 54% | 88% | +63% |
| Sleep-lab bed utilization | 68% | 86% | +26% |
| CPAP 90-day compliance review | 31% | 64% | +106% |
| After-hours messages handled | 0 | 538 | n/a |
The practice averaged $280 in contribution margin on a new consult, $950 on an in-lab PSG net of tech and consumables, and $410 on a PFT with interpretation. The no-show, utilization, and recall deltas translated to about $112,400 in incremental revenue over 90 days against an all-in cost of $9,400 for setup plus three months of platform fees.
Two things went less well. The agent initially over-routed shortness-of-breath callbacks to the nurse pool, fixed in week 2 with a rule that reads pulse-ox and onset timing first. And WhatsApp opt-in among patients over 70 was 38%, so SMS stayed the default for that cohort.
EHR and sleep-lab integration: what to ask for
Every pulmonology practice sits on one of a handful of EHRs plus a sleep-lab system. Ask vendors to answer against this table before you sign.
Swipe to see the full table →
| System | Integration path | What works today |
|---|---|---|
| Epic (Cadence + MyChart) | App Orchard, FHIR R4, HL7 v2 | Read schedule and patient demographics, write appointments to open slots, post documents and encounter notes to the chart. Recall triggers via Cadence worklists are mature. |
| athenahealth (athenaOne) | Marketplace API, FHIR | Full scheduling read and write, patient-communication logs back to the chart, task creation in the correct clinical pool. Strong for a mid-size practice. |
| eClinicalWorks | HL7 v2, eCW API with partner sponsorship | Reliable scheduling read and write once the HL7 feed is in place. Document posting is supported, patient-comms writeback is partial. |
| NextGen | NextGen Share, HL7 | Scheduling read and write, messaging writeback via Share. Watch for slot-type naming drift across offices. |
| Natus SleepWorks / Nox Medical | HL7 orders and results, SFTP for study files | Bed-level utilization read, order and result routing to the Epic or athena chart, waitlist-fill triggers. |
| ResMed AirView / Philips Encore | Vendor API with DME partner consent | CPAP compliance read, usage flag reads, outreach triggers. Confirm the data-sharing agreement your DME partner holds, since the pulmonology practice is often downstream. |
Confirm beyond the headline: how slot types are named per office (PFT_30, PFT_45, PFT_PRE_OP all exist in the wild), whether the agent can read provider-preference and insurance-eligibility fields, and whether documents post back to the correct encounter rather than a loose documents folder.
HIPAA, BAA, and the parts people skip
A pulmonology AI agent handles PHI the moment it collects an Epworth score or a CPAP compliance percentage. A clean compliance posture has five pieces:
- Signed BAA with the AI vendor and every subprocessor that touches PHI (voice carrier, transcription, LLM, messaging gateway, DME API).
- Encryption in transit and at rest, TLS 1.2 or higher, with explicit scope for call recordings and transcripts.
- Role-based access on the admin console, with audit logs the practice can export during an OCR audit.
- Minimum necessary disclosure rules on what the agent reads from the EHR and sends over SMS or WhatsApp. If a patient has not opted in to SMS content, the agent sends a neutral "please contact the office" message instead of clinical detail.
- Retention and purge policies on call recordings, transcripts, and messaging history, stated in writing.
Ask for the BAA and the subprocessor list before the demo, not after. Any vendor that cannot produce both in a day is not someone you want near your sleep-study results.
2026 pricing bands for a US pulmonology practice
Swipe to see the full table →
| Tier | Monthly | What is included | Best fit |
|---|---|---|---|
| Starter | $349 to $649 | Phone agent on overflow or after-hours, basic SMS, 1,500 to 3,000 patient interactions, 1 EHR integration, 1 sleep-lab feed. | Solo pulmonologist, 1 office, no in-lab sleep center. |
| Practice | $899 to $1,990 | Phone and WhatsApp, pre-test sequences for sleep and PFT, CPAP compliance outreach, 5,000 to 12,000 interactions, 2 EHR integrations, HIPAA admin console, waitlist-fill engine. | 2 to 5 providers, 1 to 2 offices, in-lab sleep center. |
| Group | $2,400 to $4,900+ | Multi-office routing, procedure prep flows per site, analytics, SSO, 20,000+ interactions, custom EHR work, DME API integration, ILD registry recall. | 6+ providers, multiple offices, hospital-affiliated sleep lab. |
Add a one-time setup fee of $1,500 to $5,500 for EHR work and call-flow design. Per-minute voice costs sit around $0.07 to $0.15. A HIPAA BAA should never be a paid upgrade.
Two-week rollout plan
- Days 1 to 3, scope and data pull: baseline call abandonment, no-show, referral-to-contact, sleep-lab utilization, and CPAP compliance review rates. Decide pilot scope and the single procedure flow to launch with, usually in-lab PSG or pre-op PFT.
- Days 4 to 7, build: stand up the agent on a test number, connect the EHR and sleep-lab feed, write urgency rules for hemoptysis, suspicious imaging, and Epworth over 15, build pre-test sequences, and record the 8 to 12 intents that cover 90% of traffic. Dry-run with 20 real patient calls you forward on purpose.
- Days 8 to 10, pilot: route overflow and after-hours to the agent. Keep the main line on humans. Daily 15-minute review with the practice manager.
- Days 11 to 14, full rollout: move the main line behind the agent with a human fallback. Turn on CPAP compliance outreach and the waitlist engine. Publish the KPI dashboard.
Two weeks is realistic when the EHR integration exists off the shelf. Add a week for a custom HL7 feed or a sleep-lab vendor that needs a bespoke SFTP pipeline.
Five KPIs to track from week one
- Call abandonment rate at the main line and after hours. Target: below 5% after week 2.
- Median referral-to-contact time from fax or e-referral received to the first patient outreach. Target: under 30 minutes in business hours.
- No-show rate split by new consult, PFT, in-lab PSG, and HSAT. Target: 5 points better than baseline within 60 days on every split.
- Sleep-lab bed utilization, with the waitlist-fill mechanism turned on. Target: above 80% weekly utilization by day 60.
- CPAP 90-day compliance review completion in the eligible cohort. Target: above 55% by day 90.
If the vendor cannot show these five numbers in a weekly report, they are selling you a demo, not a platform.
What a pulmonology AI agent is not
It does not practice pulmonology. It does not read a sleep study and tell a patient their AHI. It does not titrate a CPAP. It does not interpret a PFT. It does not replace the nurse triage line for an acute dyspnea call. The job is operational: pick up the calls your front desk cannot, book into the right slot the first time, keep every PFT and PSG patient prepped and confirmed, keep every OSA patient on the compliance cycle, and give the schedulers back the hours they are losing to voicemail.
Done right, the sleep lab runs closer to its bed count, schedulers work on exceptions instead of hours-and-location calls, and patients get a response in minutes instead of a voicemail.
Ready to put an AI agent in front of your pulmonology practice?
At ZENIA we deploy voice and WhatsApp agents for US pulmonology practices in two weeks, with EHR writeback, sleep-lab integration, HIPAA posture, and the five KPIs on a weekly report. Diagnosis, build, pilot, rollout.
Talk to us on WhatsApp Book a 30-min call