When I first talked to Dr. Marcus Webb, he was seeing 22 patients per day and finishing his notes at 10pm. His front desk manager, Linda, was fielding 80+ calls per day with two staff members, missing roughly a third of them. Their no-show rate was 17%. "We're not a bad practice," he told me. "We just spend most of our time on things that aren't patient care."
Meet the Practice
Westside Family Medicine is a 2-provider primary care practice in suburban Denver, Colorado. Dr. Marcus Webb and Dr. Priya Nair have practiced together for six years. 1,850 active patients. Three administrative staff. Annual collections just under $600,000.
By any measure, they were running a competent, patient-centered practice. Dr. Webb had a 4.8-star Google rating from 43 reviews. Dr. Nair was voted "Best Family Doctor" in a local community poll two years running. Patients stayed. Referrals came in.
The operational reality underneath those numbers was less comfortable. Dr. Webb was spending 2.5 hours per day on documentation. Dr. Nair was spending 2 hours. Their 17% no-show rate meant approximately 15 empty slots per week — $2,250 in weekly lost revenue at their average visit value of $150. Linda's team was so consumed by inbound call volume that recall campaigns and preventive care outreach simply didn't happen.
When we talked in February 2026, Dr. Webb wasn't looking for AI tools. He was looking for a way to hire a third administrative staff member. I asked if we could look at the data first.
The Two-Week Audit
Before recommending anything, I tracked how the practice was actually spending its time for two weeks.
Front desk time breakdown:
- Appointment scheduling: 28% of call volume
- Appointment confirmation and reminders: 22% (all manual phone calls)
- Patient questions (hours, directions, insurance, prescription refills): 31%
- Clinical calls requiring provider involvement: 19%
The insight: 81% of front desk call volume was routine — schedulable, confirmable, answerable without clinical judgment. The remaining 19% genuinely required human attention.
Provider documentation time:
- Dr. Webb: average 6.8 minutes per note, completed mostly after clinic hours
- Dr. Nair: average 7.2 minutes per note, completed during clinic with patient wait time extended
No-show analysis: Of the 17% monthly no-show rate:
- 58% had received no reminder (scheduling system wasn't configured for automated reminders)
- 31% had received a voicemail reminder they hadn't responded to
- 11% were genuine no-shows despite confirmed attendance
The diagnosis was clear: the majority of no-shows were preventable with better reminder automation. The majority of front desk call volume was handleable by a patient communication platform. The majority of documentation was being done after hours because there was no ambient scribing tool.
None of these required a new hire. They required three tools.
The Tools Chosen
| Tool | Monthly cost | Problem it addresses |
|---|---|---|
| Luma Health | ~$275/month (custom quote, 2 providers) | No-shows, reminder automation, waitlist management |
| Freed AI | $149/month (2 providers at $99 + $50) | Documentation — ambient scribing |
| ChatGPT Plus | $20/month | GBP content, review responses, patient education |
| Total | ~$444/month |
What we didn't add: A new staff member ($45,000+ annual salary + benefits). The tool cost of $444/month annualizes to $5,328 — roughly 12% of the cost of the hire being considered.
HIPAA verification process before go-live: Before deploying Luma Health, Linda reviewed the BAA with their healthcare compliance consultant. The process took one week — confirming BAA scope covered SMS appointment data, verifying data encryption and storage policies, and confirming their EHR (athenahealth) had a certified integration with Luma. This week of compliance verification is the step most implementation guides skip. At Westside Family Medicine, it was non-negotiable.
Month 1: The Setup Reality
Luma Health (Weeks 1–3)
The athenahealth integration was the critical path. Luma's integration with athenahealth is certified and bidirectional — confirmed appointments write back to the schedule in real time. Setup took 8 business days from signed contract to go-live: 3 days for integration configuration, 2 days for reminder sequence setup, 3 days of parallel testing with the existing workflow.
What Linda configured:
- Booking confirmation: immediate text upon scheduling, with appointment details and a preparation reminder
- 72-hour reminder: "Your appointment with Dr. Webb/Dr. Nair is in 3 days. Reply YES to confirm or call [number] to reschedule."
- 24-hour reminder: Final reminder with office address and parking instructions
- Waitlist: Any patient who called to schedule during a full week was added to the waitlist for their preferred provider and time window
What didn't work in week 1: 14% of patients had outdated or invalid phone numbers in athenahealth — the reminders simply couldn't reach them. Linda spent part of week 2 updating contact information for their highest-frequency patients. This is the data quality problem that every patient communication platform implementation encounters — plan for it.
Month 1 no-show rate: 12.3% — down from 17% in the previous month. The 4.7-percentage-point reduction was almost entirely from the 58% of no-shows who had previously received no reminder at all.
Freed AI (Week 1)
Both providers started Freed simultaneously. Dr. Nair adapted faster — she's more comfortable with ambient technology and her consultation style (structured, question-driven) produced cleaner initial output. Dr. Webb required 10 days before he stopped narrating to the microphone rather than talking naturally to his patients.
Freed setup: 45 minutes total. Both providers were taking ambient notes by day 1. Neither required IT support.
Month 1 documentation outcomes:
- Dr. Webb: Average note completion time dropped from 6.8 minutes to 4.1 minutes. Notes completed during clinic hours increased from 40% to 71%.
- Dr. Nair: Average note completion time dropped from 7.2 minutes to 3.9 minutes. Patient wait time between appointments decreased by an average of 4 minutes per slot.
What required editing: Complex encounters — multi-problem visits with more than 4 active issues, patients with communication barriers, and mental health discussions where sensitive content required careful documentation — consistently needed more editing than straightforward acute or chronic disease management visits. Both providers learned to flag these encounter types for more thorough note review.
Month 2: The System Working
No-Show Reduction — The Waitlist Effect
Month 2 introduced the waitlist automation that wasn't configured in month 1. When a patient cancelled — via text reply to a reminder — Luma automatically contacted the first waitlisted patient for that provider and time slot.
The waitlist filled 67% of cancelled slots in month 2. At an average visit value of $150, 22 slots filled by waitlist patients = $3,300 in revenue that would previously have been lost.
Month 2 no-show rate: 9.4%. Down from 17% at baseline, 12.3% in month 1.
Front Desk Call Volume
Luma's two-way messaging handled routine scheduling interactions that previously required phone calls. Patients confirmed, rescheduled, and asked basic questions via text. Linda's team tracked inbound call volume for the month.
Month 2 inbound calls: 1,847 — down from 2,340 in the comparable month from the prior year (a 21% reduction). The reduction concentrated in confirmation calls and basic scheduling questions. Clinical calls requiring provider involvement were unchanged.
The front desk wasn't spending less time — the hours freed from routine call management went to recall outreach (patients overdue for preventive care), insurance verification, and the administrative backlog that had accumulated during high-volume periods. Linda described it as "finally having time to do the things we knew we should be doing but couldn't get to."
Documentation and Provider Wellbeing
By month 2, both providers were completing notes during clinic hours on more than 80% of encounters. Dr. Webb had not taken work home on a weekday in three weeks.
"I've been practicing for 14 years," he said. "I've never finished a Tuesday without chart notes to complete. It happened three times this month."
Dr. Nair reported a different effect: "I'm actually listening differently. I'm not trying to remember everything to write down later. I'm just present with the patient."
The Honest Numbers at Month 3
| Metric | Baseline (Feb 2026) | Month 3 (May 2026) | Change |
|---|---|---|---|
| Monthly no-show rate | 17% | 9.3% | −45% |
| Inbound call volume | ~2,340/month | ~1,820/month | −22% |
| Dr. Webb avg. note time | 6.8 min | 3.8 min | −44% |
| Dr. Nair avg. note time | 7.2 min | 3.6 min | −50% |
| Notes completed during clinic hours | 40% (Webb) / 55% (Nair) | 83% (both) | +28–43pp |
| Cancelled slots filled by waitlist | 0% | 67% | — |
| Google reviews (total) | 43 | 61 | +42% |
Revenue impact: Monthly no-show rate: 17% → 9.3% on approximately 440 monthly appointments = 34 recovered appointments × $150 average visit value = $5,100/month in recovered revenue.
Waitlist fill rate: 67% of 25 monthly cancellations filled = 17 additional appointments × $150 = $2,550/month.
Total recovered revenue: ~$7,650/month Tool cost: $444/month ROI: 17×
The Google review increase came from a systematic review request process Linda implemented in month 2 — using the ChatGPT prompt template from this guide for response drafting and a simplified text-based review request process through Luma. 18 new reviews in 3 months at a 4.9 average rating.
What the Practice Would Do Differently
At the 3-month mark, I asked both Dr. Webb and Linda what they'd change.
Dr. Webb: "I would have started Freed a year earlier. The time I spent documenting after hours over the past three years — if I had that back, I would have spent it differently. I'm not sure I fully understood how much the documentation burden was affecting everything else until it wasn't there."
Linda: "The phone number cleanup before going live with Luma. We should have run a data quality audit on contact information before the first reminder went out. We lost probably two weeks of full effectiveness because a significant portion of our patient database had outdated numbers."
Dr. Nair: "I'd have been more patient with Freed in the first two weeks. I almost stopped using it after day 5 because the notes on complex encounters needed so much editing. By week 3 it had calibrated to my style and the editing dropped dramatically. I didn't trust the process enough initially."
The one thing they got right: Starting with HIPAA compliance verification before going live with Luma, not after. "Our compliance consultant found two things in the BAA that needed clarification before we were comfortable. If we'd rushed the launch, we might have been running a compliant platform with a non-compliant configuration."
What You Can Take From This
Westside Family Medicine's situation — competent practice, genuine patient loyalty, operational inefficiency consuming clinical time — is the most common profile I encounter in independent primary care.
The three tools in this case study cost $444/month. The recovered revenue in month 3 was $7,650 — a 17× return. The recovered provider time from documentation reduction is harder to monetize but arguably more valuable: Dr. Webb and Dr. Nair are finishing their clinical days without taking work home.
The new administrative hire that was being considered — at $45,000+ annually — was never needed. The existing team now has capacity for the work they previously couldn't get to.
The pattern is consistent with every practice I've watched implement these tools: compliance verification first, one tool at a time, realistic expectations about the first 30 days, and patience with the calibration period.
The paperwork isn't the job. But it was eating the job. That's fixable.
