Guides
Switching AI Vendors at a Medical Practice
Switching AI vendors means moving phone routing, scheduling rules, athenaOne access and patient data. The cutover order that keeps calls answered throughout.
Switching AI vendors starts with everything the old one quietly owns. Practices usually decide to switch for one reason: the agent answers the phone but does not finish the work. It takes a message, the message lands in a queue, and staff end up doing the task anyway. By the time a practice admits that, the old system is wired into the phone routing, the scheduling rules and the patient record.
The real risk is the two weeks where neither system is fully in charge and calls fall through the gap.
What actually has to move
Before talking to any replacement, write down what the current vendor touches. There are usually four things.
Phone routing. Where do calls go today, and who controls that? If the vendor’s number sits behind your main line through call forwarding, switching is a forwarding change. If the vendor ported your number, getting it back is a carrier process with its own lead time. Know which one you have before you set a date.
The rules. Every working deployment encodes practice rules: which appointment types each provider takes, which plans a provider is not enrolled in, minimum lead times for new patients, procedures that need a clearance visit first. Some of that lives in the old vendor’s configuration and nowhere else. Export it, or reconstruct it from your own scheduling data before access ends.
athenaOne access. The old vendor has its own authorization into athenaOne. The new one needs its own, and that approval chain can take weeks when a management company or a weekly approval meeting sits in the path. Start it first. Then make sure the old vendor’s access is actually turned off at the end, not left running in the background.
Patient data. Call recordings, transcripts and logs contain protected health information. Your business associate agreement should already say what happens to it: HIPAA requires the contract to provide that at termination the business associate returns or destroys the PHI it still holds, if feasible, and keeps no copies (45 CFR 164.504). Decide whether you need the history first. Then get destruction confirmed in writing.
The cutover order that keeps calls answered
Most failed switches try to flip everything on one morning. A safer order:
- Inventory and data. List every queue, number and workflow the old vendor touches. Pull 90 days of booking history from athenaOne by department, provider and appointment type. That history, not the written rule sheet, is the ground truth, and practices routinely find that part of the logic they would have handed over is out of date.
- New access approved. Get the replacement’s athenaOne authorization through before anything else moves.
- Supervised live blocks. Run the new agent on a slice of real calls, a half day at a time, with staff listening and filing change requests. Start with one queue (after-hours, or a single location) rather than the whole practice.
- Move routing queue by queue. Shift forwarding for each queue only after its live block comes back clean.
- Close out the old vendor. Revoke athenaOne access, confirm data return or destruction under the BAA, and remove the old numbers from every patient-facing place they appear: reminders, portal messages, the website, voicemail greetings.
One community health center asked to switch its phone menu off entirely during a test block. Their worry was that patients would navigate a menu expecting a person, reach an AI anyway, and that the frustration would skew the results. Small decisions like that one decide whether the first week reads as a success.
What to verify in the replacement before you sign
The reason you are switching is the first thing to test. If the old agent took messages instead of finishing work, ask the new one to finish work, live, in your configuration.
- Does it write back into athenaOne? Booking into the correct appointment type and slot, creating the patient case, updating the record. Pretty Good AI works across 820+ athenaOne APIs, which is what lets a call end with the task done rather than a note for staff.
- How does it handle a rule your old vendor got wrong? Bring a real one. A vascular group pairs most visits as an ultrasound plus a provider visit with fixed spacing, and nothing in the template enforces it. Ask how the new agent books both legs, and where it hands off when it cannot.
- What does a human see on a handoff? The transfer should arrive with the caller’s identity, the request and what was already tried, so staff do not start over.
- What are the commercial terms? Ours are short: No setup fees. First 30 days live free. Month to month after. No annual lock-in. A vendor that needs a long contract to make the economics work is telling you something about how confident it is in the first month.
- How long until it is live? Agree the launch sequence during scoping, including configuration and access approval. {{TAILOR_FIT_LINE}}
Where switching does not fix the problem
A new vendor does not fix stale data. If the chart’s primary-provider field is wrong, or the provider roster on your website disagrees with the one in athenaOne, any agent will route on bad information. Reconciling those is staff work, and it should happen during the inventory step, not after go-live.
Phones are also not the whole front office. In a March 2026 MGMA Stat poll of 294 practice leaders, 45% named eligibility and prior authorization as their staff’s most time-consuming phone task and 31% named scheduling (MGMA). If the replacement only answers calls and cannot touch the eligibility and scheduling work behind them, you are buying the same limitation twice.
For the questions that separate a vendor writing into athenaOne from one reading a calendar, see how to tell whether an athenaOne add-on can do the real work. For the numbers to pull before any change, see the front-office metrics to pull before a platform change.
Key Takeaways
- Before switching AI vendors, list what the old one owns: phone routing, practice rules, athenaOne access and patient data.
- Start the new vendor’s athenaOne authorization first; multi-layer approvals can take weeks.
- Use 90 days of booking history as the source of truth for rules, since written rule sheets drift out of date.
- Move routing one queue at a time, after a supervised live block on real calls comes back clean.
- Close out properly: revoke the old vendor’s athenaOne access and get return or destruction of PHI confirmed in writing, as your BAA requires.
The best switch is one patients never notice. Plan it as a sequence of small, reversible moves and the first week will look like any other.
Sources
Ready to See It in Action?
See how Pretty Good AI runs a supervised cutover inside athenaOne
Schedule a Demo →Written by Kevin Henrikson