Stories from real work inside a primary care practice, anonymized. Each one says what was built, what was hard, what the logs measured, and what is not measured yet. No practice, staff, patient or health plan is named.
RPA bot means robotic process automation: software that clicks through screens the way a person does. It is not AI; it is the right tool for clicking, reading and matching.
Tracking HEDIS and Star measures, prioritizing outreach and preparing chart reviews.
1,405 of 1,409
patients matched to their EHR chart
Source: Build notes and the bot's run logs, Oct 2026
Before each chart review, medical assistants need to know what is already on file. An RPA bot opens each upcoming patient in the EHR, reads history notes and recent labs, and matches them to the quality measures. The hard part wasn't AI. It was getting every patient to the right chart.
29 of 29
test cases passed, including shifted, tilted, faded and already-filled scans
Source: Form-filling bench run, 14 Oct 2026
Many conditions found in outside claims are real, but nothing can be done about them without the records. An RPA bot takes the patient's signed release, fills in the facility and fax number, and sends the request with a cover page. It passed every case on its test bench, and it is built so it cannot fax twice by accident.
Surfacing documentation gaps with the evidence behind every suggestion.
357 → 2,154
conditions confirmed as captured for one health plan, once the bot read the practice's own claims
Source: The bot's run logs and build notes, Oct 2026
Plan claims files run weeks to months behind and barely show the practice's own visits. So an RPA bot reads the claims the practice actually sent, and a gap engine marks each risk-adjusting condition captured or open, with the evidence for every suggestion. A practice manager's audit of real rows became the rules that keep the list honest.
30 / 20 / 50
weights for surveillance flags, spending trend and value-based balance
Source: Build notes, Jul 2026
The roster's first risk column was a placeholder with no data behind it, and its replacement disagreed with the patient page. The fix was a single 0 to 100 score that blends surveillance flags, spending trend and the value-based balance, with the breakdown shown in every patient's header.
Dashboards for membership, utilization, spending trends and contract performance.
103
releases since July 2026, each tested on a staging site first
Source: Commit history, Oct 2026
Three plans, three file formats, and the surplus-and-deficit picture in a workbook: there was no single place to see one member's story. The finance dashboard puts claims, hospital and ER use, risk, value-based balance and upcoming visits on one page per member. It was built on de-identified data first and tested before any real record went in.
507 → 277
apparent admissions became true admissions
Source: Commit history and notes, Jul 2026
Counting claim lines overstated ER visits about nine times and nearly doubled hospital admissions. The utilization engine merges day-by-day claim lines into real stays and visits, with length of stay and 30-day readmission flags.
Insurance verification, pre-registration, referral routing and patient queues.
137
patients pre-registered on the first clinic day, with zero duplicates
Source: The bot's run logs and build notes, Sept–Oct 2026
A patient used to appear on the front desk queue only when they arrived. An RPA bot now exports the schedule every 15 minutes and pre-registers the next two weeks of appointments, routed to the right desk, with insurance and copay filled in before the patient arrives.
281
cards created on the first run
Source: Build notes, Oct 2026
Follow-up calls for missed and cancelled appointments were reported as a number typed into a daily checklist. Now every missed or cancelled provider appointment becomes a card on a shared board, with phone and insurance filled in, and the card closes itself when the patient rebooks.
Claim tracking, rejections, unbilled visits and re-bill worklists.
2,727
claims captured in the first history load, across 72 biller-days
Source: Build notes, Oct 2026
Billers used to type 'charts billed' and 'rejections' into the dashboard by hand. An RPA bot now reads the claims each biller sent, in every status, plus the clearinghouse's claim-status reports, every 30 minutes, and links each rejection back to its claim and its biller.
15 s → under 1 s
year-to-date view after one query fix (a few seconds today, with features added since)
Source: Build notes, Oct 2026
There was no automatic list of visits that had been checked out but had no claim. The dashboard now compares checked-out insured visits with the claims the billing bot has seen and lists the gap, per biller and insurance.
Live dashboards and daily performance reporting for every department.
4 + 47
on-site RPA bots and scheduled server jobs keeping the dashboard current
Source: Build notes, Oct 2026
In the first version, departments typed their daily numbers in by hand and the entries went stale. One by one they were replaced with live data from RPA bots, the EHR, the pharmacy system and the phone system, until the leadership tiles came straight from the work itself.
248 → 13
prescriptions open past 14 days, after the cleanup
Source: Build notes, Sept 2026
The pharmacy system's feed reports some events and not others, so reversed and picked-up prescriptions stayed open on the dashboard and showed up as falsely overdue. A live connection with automatic cleanup cut the open-past-14-days list from 248 to 13.
Bring one process that eats your team’s day. We’ll talk through what to measure first, and whether automation or AI fits it at all.
