Number 01
42.5%
Payer rows that needed no outreach
13,831 of 32,548 payer rows needed no outreach.
Denominator: all 32,548 payer rows in the measured file.
The FQHC Operational AI Playbook
Start with your numbers, not our story.
A self-assessment pack for health center operations leaders: five numbers to pull from your own operation, six questions to answer before any vendor call, a vendor scorecard, a pilot charter, and a full nine-play case reference. Free, and usable whatever tools you choose.
Case reference: Aunt Martha's Health & Wellness. Production measures cover June 1 to July 28, 2026, with populations and denominators stated alongside the result.
Part one
Your five numbers
Measure your own operation before you talk to any vendor, including us. The case measures below show what a well-defined answer looks like. Every blank you cannot fill is a visibility gap, and that is a finding too.
Number 02
72.8%
Patients contacted on two or more channels
A single-channel workflow would have missed the shape of the work.
Denominator: patients with recorded contact activity.
Number 03
85.1%
Referrals with first outreach within 24 hours
Median first outreach was 4.6 hours. For care gaps, 72.5% received first outreach within 24 hours, with a 9.5-hour median.
Populations: measured referral orders and care-gap items, reported separately.
Number 04
4,740
Staff actions across 2,749 referrals
Routine handling and staff judgement were measured separately.
Denominator: 2,749 referrals with recorded staff action.
Number 05
24.9%
Referrals escalated
80% of those escalations were due to internal routing or no available specialist.
Denominators: all measured referrals, then all escalated referrals.
Evidence window: June 1 to July 28, 2026. These are observed, case-specific measures, not projections. Use the same discipline on your own baseline: define the population, formula, date field, and exclusions before interpreting a result.
Part two
The six-question worksheet
If AI has a role, the harder part is operating it well. Work through these six questions on one of your own workflows before taking a vendor call.
Name the outcome you want for patients and staff
Confirm no staff replacement, no system switch
Define what success looks like on day one
Give every metric one definition and one denominator
Instrument before you scale
Plan the change curve and the hypercare
Part three
The vendor scorecard
Make any vendor, including us, show you these six things before you trust their numbers.
A living metric dictionary
One definition per metric: population, formula, date field, and exclusions. When two views disagree, the dictionary decides.
Denominator rules
Every percentage travels with its base and population. Two percentages can both be true while describing different groups.
A zero must mean zero
Separate zero from not tracked, not wired, and field empty. Ambiguous zeros cannot support decisions.
Complete audit trails
Give every action an actor and a timestamp. Complete attribution changed one coordinator’s recorded resolutions from 22 to 61.
Screen equals export
What leadership sees on screen must match what finance downloads. Require parity checks.
Label grain and time basis
State whether a measure is per patient or per record, and whether it uses order date or action date.
Part four
The phase-gated pilot charter
Start narrow and earn the next step. The scope will be yours; the gate discipline is the part to keep.
Pilot
Prove one bounded workflow
- Referrals and care gaps
- One payer
- Eleven named metrics
- Definitions and denominators set before measurement
Gate
Verify before expansion
- Closure definition agreed
- Screen and export reconciled
- Staff time baselined
- Adoption and workflow risks named
Phase I
Expand with evidence
- Seventeen named metrics
- All five MCOs
- All specialty referrals
- Continued hypercare and adoption review
The expansion rule
Move to the next phase only when the measurement, workflow, and people-change gates are all met.
Part five
The full case playbook, as reference
How Aunt Martha's approached the work: nine operating plays in sequence, followed by the four lessons we would carry into the next implementation.
Name the work honestly
Closing care gaps and coordinating referrals was a capacity gap, not a technology gap. The backlog identified work that volume had pushed beyond the team’s available time.
Start with the work that never gets done consistently. The backlog is the brief.
Set guardrails before tools
Extend capacity, keep the people, and keep athenahealth. Protect the patient experience with approved scripts, operator-set outreach windows, and an opt-out model reviewed by legal.
Automate routine work. Keep judgement and relationships with the team.
Measure so good work gets seen
Measurement protects the patient and makes previously invisible work visible. It creates accountability without turning the dashboard into surveillance.
With complete attribution, one coordinator’s recorded resolutions moved from 22 to 61.
Lead the change curve on purpose
Plan for go-live optimism, the visibility dip, the turn, and team ownership. Name the discomfort, set fair targets, stay visible, and celebrate improvement.
The team should run the dashboard, not the other way around.
Hypercare after go-live
Use daily drop-in help, a shared weekly tracker, explicit escalation paths, item-by-item verification, rapid iteration, training, and staff communication.
In one hypercare review, 19 of 25 delivered items were confirmed after item-by-item verification.
Earn trust in the numbers
Apply the six scorecard rules to every operating report. Trust is earned when the metric dictionary, denominator, audit trail, screen, and export agree.
A useful automation knows its limits and makes those limits visible.
Instrument before you scale
Track the referral ladder from order ingested to loop closed: validated, specialist matched, packet faxed, patient reached, scheduled, seen, consult note returned, and closed.
The case used nine referral stages and ten operational reports.
Phase-gate the expansion
Move from a bounded pilot to broader operations only after the agreed measurement, workflow, and adoption gates are met.
Pilot: 11 metrics and one payer. Phase I: 17 metrics, five MCOs, and all specialty referrals.
What we learned
What we’d tell the next operator
Define closed on day one
Two reasonable definitions of the same result produce two different numbers. Pick one before you measure.
State the denominator
A percentage without its base invites the wrong conclusion. Put both on the same slide.
Measure staff time from the start
Baseline it before go-live so you can show the gain with confidence afterwards.
Lead the people change yourself
A memo is not change management. Set the expectation, stay visible through the first weeks, and make it safe for the team to rely on the new way of working.
Three lines to carry forward
The operating principles
Buy back capacity. Keep your people.
What gets measured finally gets done.
Go-live is halftime. Adoption is the win.
Keep going
Put the playbook to work
Continue with practical worksheets, buyer guidance, operating guides, and the complete case study.