Customer case study · Trident Anesthesia
How Trident Anesthesia stopped building call by hand.
Twenty-five physicians. Two hospitals, six surgery centers. The call schedule used to take one partner’s nights and weekends, and the daily schedule took a second doctor. Today one doctor runs both in a fraction of the time, and the call schedule has not been built by hand since November 2025.
Before
Two doctors. Every month. Their nights and weekends.
Trident Anesthesia is a twenty-five-physician group covering two hospitals and six surgery centers. It is also the group PostCallMD was built for. Before any of this was a product, one of the partners, Dr. Feller, built the monthly call schedule with paper and a spreadsheet, and a second doctor built the daily schedule.
The call schedule is not a Saturday afternoon job. It is nights and weekends. A fair call schedule is a genuinely hard problem, and the inputs make it harder: email requests, last-minute changes, special cases.
The requests alone were a job. Day-off asks arrived by email and text whenever doctors thought of them, and building October meant digging back through June’s inbox to find what had been promised.
Vacation ran as a draft. So did holidays. In practice that meant a spreadsheet emailed around the group for weeks, with the scheduling doc stuck in the middle: send it to the next doctor, wait, record the pick, send it again.
And the daily schedule was paper. Print the month as a day-by-day list of names, sit down with the vacation spreadsheet, scratch out who was away or on call, and build each day by hand.
None of this was broken. It is what conscientious scheduling looks like without software, and it worked for years. It worked because two doctors quietly paid for it in nights and weekends.
The rules are specific: contract-employee minimums, post-call cardiac coverage, weekend pairings the group avoids. A spreadsheet can hold them as long as one person keeps them in their head. The risk is what happens to the schedule when that person stops being available, or wants a weekend back.
The work, before any schedule was generated
Rule capture, then the solver.
The first weeks were not about software. They were about getting the rules out of one partner’s head and onto paper. We sat down with Dr. Feller and walked through every situation the spreadsheet had been quietly handling for years: which sites need which credentials, which partner pairings get avoided on weekends, what post-call rest looks like leading up to a cardiac case, how contract employees’ committed shifts get allocated before partners absorb the rest.
Then we imported the prior months’ schedules so fairness accounting could start from history, not from zero. Then we encoded the rules into the solver as constraints. Only then did we generate the first draft schedule.
Contract minimums protected first
Contract employees get their committed shifts first. Partners absorb the remainder. Encoded as a hard constraint: the solver assigns contracts before anything else.
Post-call cardiac recovery
Cardiac call comes with specific post-call rest requirements distinct from the rest of the schedule. Modeled separately, enforced before any cardiac assignment lands.
Weekend partner pairings
Specific partner combinations the group prefers to avoid on weekends. The solver knows which pairs to keep apart and finds combinations that work.
Year-long vacation glide path
Vacation distribution is solved across the whole year. No physician hits December with an unrealistic number of target shifts because the system didn't properly account for vacation.
After
Every month since November 2025. The spreadsheet stays closed.
Since November 2025, every monthly Trident call schedule has been built through our solver and published to the group. The April 2026 schedule above is one of them.
The process has also survived its first handoff. Scheduling used to be split between two doctors, one on call and one on daily. Today a different doctor runs both jobs, on the same solver and the same rules, in a fraction of the time, from anywhere the app opens. That handoff is the proof that matters: the rules live in the system now, not in one person’s head.
Each piece of the old process got its own fix. Requests land in one queue now, from the app or plain email, with the AI turning messy notes into dated entries the scheduling doc approves or denies. October gets built without an archaeology dig through June’s inbox.
The vacation and holiday drafts run themselves. Draft Runner emails each doctor when their turn comes, shows what is still open, and records picks as they land. We built it out of necessity rather than ambition: the solver needs vacation as structured data, not a spreadsheet attachment, so the draft had to finish inside the system. The weeks-long email relay ended as a side effect.
And the daily board starts full instead of blank. Vacations, call, and capabilities are already on it, so the doctor building the day paints coverage where coverage is needed instead of scratching names off a printout.
What changed for Trident is that a partner’s weekends came back, and for the first time there is a system of record: requests, vacations, capabilities, and history in one place.
What it does not solve
The honest part.
The call schedule is not fully hands-off. The solver generates the draft, and the scheduling doc still edits it before it publishes. Some edits are judgment the solver does not have yet. Some are preference. That is by design: the draft is a starting point the group controls, and each month’s edits are feedback we tune the next cycle against.
And we have not automated the daily solve yet. The daily schedule is finalized the night before, a lot of judgment goes into it, and what the system automates today is the assembly work underneath that judgment, not the judgment itself. When we do automate it, it will be built the way the call schedule was: rules first, then the solver.
What we built it on
If you want to see the pieces in more depth.
What anesthesia call scheduling actually requires
Post-call rest, site coverage, partner pairings, fairness history: the rules every group has and most tools can’t model. The pillar page.
The AI-plus-solver walkthrough we published for free
The do-it-yourself version of the same approach. ChatGPT plus OR-Tools CP-SAT. Walks through how a constraint solver gets built and what it actually does.
How we compare to QGenda and Amion
An honest side-by-side. Where each tool is the right answer for an anesthesia group, where it isn’t.
The questions you should ask a small vendor
Continuity, data, PHI, certifications, and what leaving looks like, answered in one place. Reference calls with Trident are available to groups late in a serious evaluation.
Your group
If a partner in your group is doing Dr. Feller’s job today, send us their hardest rule.
The thing that makes the schedule take a weekend instead of an hour. We read every one and reply within one business day. We only take on two new groups per month, so the rule-capture work is done properly.