The rules connected each incoming item to a responsible person.
We built the routing on the practice’s server. It read device-data exports and alert messages, matched items to structured patient records, and applied physician-approved rules. The chart connection read the information needed for routing. It didn’t write an interpretation into the clinical record.
Flagged items and cases the rules couldn’t classify went to the physician. Routine clinical items went to the nurse practitioners, while connectivity and scheduling work went to the coordinator. The receiving clinician reviewed the supporting report and signed the interpretation. Software didn’t interpret raw heart-signal tracings or sign clinical work.
For monthly device reports, a billing check matched signed reports to claims and flagged missing claims for the billing manager. This connected completed clinical work to a billing review rather than assuming that a signed report had already been billed.
Past decisions supplied a test the team could repeat.
We captured the physician’s criteria from signed reports and explanations, then built a reference set of 1,100 past dispositions. This provided examples against which to compare the routing before expanding it across the device-data sources.
The acceptance checks addressed routing agreement, missed urgent escalations and the share of items referred back because the rules couldn’t classify them. Those checks tested the routing task. They didn’t establish that software could make independent clinical decisions.
After rollout, the physician audited a sample of cleared items weekly. The reference set ran nightly, and a changed result stopped rule changes until the clinical lead reviewed it. The team could therefore inspect both changes to the rules and how cleared work was being handled.
The export check addressed the failure we found in rollout.
A changed vendor column caused 14 flagged transmissions to reach the wrong queue. Our log review found the error, and we notified the practice that morning. The physician reviewed every affected item within a day and found that none required action.
We added a check of the expected export fields that halted routing and alerted the coordinator when the format changed. We reran the reference tests and documented the correction at no charge. This guarded the data entering the routing, which the earlier decision tests alone hadn’t protected.
The practice retained clinical approval and operating ownership.
The routing and patient detail stayed on the practice’s server. We used role-specific access, logged queue changes, and documented backup, restore and manual fallback procedures. Administrative and remote access used multi-factor authentication, and backups were encrypted.
The clinical lead owned the rules, with the physician approving changes. The coordinator owned the queues and exports, and the IT partner maintained the server. We trained backup operators so support didn’t depend on one person.
The practice’s separately purchased scribe was a different data path from the local routing. Our role included the comparison tests and consent-script design. The scribe processed visit recordings through its vendor’s cloud service, while device routing stayed local.
The operating comparisons and financial result answer different questions.
The baseline study paired electronic health record activity with a two-week diary and device and referral logs. The operating follow-up used those same types of records: device logs for the share cleared without the physician, referral records for median appointment waits, and diaries for the physician’s working week.
The long-term operating comparisons were 0% to 88% of device items cleared without the physician’s personal review, 6.5 to 3.1 weeks median new-patient wait, and 66 to ~58 working hours per week. The 88% describes who reviewed the work, not an accuracy score or the share of patients treated automatically.
In a later conversation with Already Built’s founder, the practice owner confirmed a $2.6 million increase in annual collections and attributed the full increase to our work. That is the owner’s reported business result. The operating measurements explain how work changed. They aren’t converted into a separate dollar benefit or added to the collections gain.
A new rule demonstrated that their team could extend the system.
We gave the clinical lead, coordinator and IT partner operating instructions and taught them to change the rules and rerun the reference tests. After handover, the lead nurse practitioner and coordinator added a new alert rule, obtained physician approval, and tested it without us.
The practice could maintain and extend the routing while retaining clinical approval and repeatable checks. For technical and operating leaders, the full case shows how that capability changed the physician’s workload and patient access.