Pretty Good AI vs Assort Health

One EHR, all of it

Pretty Good AI builds for athenaOne and nothing else. Assort Health sells a patient access agent across several EHRs, and its leadership has been public about moving upmarket into health system accounts. Both of us will be on the floor at athenahealth Thrive 2026 as Platinum sponsors, so this is the year an athenaOne practice can put the two products next to each other and ask the same questions of each.

The multi-EHR path is a bigger market and a harder engineering problem, because a workflow written once has to hold up across systems that model appointments, orders and eligibility differently. Assort has written about that difficulty itself, arguing that an EHR integration is defined by the write rather than the read: reads are commoditized, while every write has to be validated against a specific practice’s scheduling and referral rules before it can be trusted. We agree. It is the whole argument for what we do.

That validation cost does not get paid once. It gets paid per EHR, per workflow, per rule, and then maintained as each platform changes. Our integration work stays focused on athenaOne, with production access to 820+ athenaOne APIs. The useful comparison is the configured workflow and the record it leaves behind, not the number of EHRs a vendor supports.

Broader than the phone

Assort’s athenahealth guide describes voice, outbound outreach, web scheduling, digital intake, referral processing, payment resolution and analytics. Those are published vendor claims, not a head-to-head test of the two products. Compare the workflow included in each proposal, the athenaOne records it changes and the work left for staff.

Pretty Good AI brings Patient Engagement, Task Automation, Referral Management and Revenue Cycle workflows together on one athenaOne integration. That includes fax, portal and web referral intake, tickler and order outreach, waitlist backfill, eligibility and prior-auth preparation, balance outreach under your billing team’s rules, post-visit outcome calls, forms and KPI reporting. We map the workflow with your team and maintain it as your rules change. New workflows are priced when activated.

Our exclusive athenaOne focus is a commitment to where we build and maintain that work. It is not proof that another vendor lacks a native integration or a particular capability.

Why writeback matters to containment

Integration depth is one part of whether a call can finish. The agreed workflow, available information, practice rules and the need for human judgment also matter.

The agent needs the relevant live athenaOne records: open slots against the scheduling template, the appointment type the provider group uses, the department the patient belongs to, insurance information and any orders or ticklers behind the visit. Evaluate whether the agreed steps finish correctly and whether work requiring a person arrives with context. A documented human exception is different from an incomplete task silently left behind.

The detail that decides it

Practices do not build schedules out of tidy named appointment types. They build them out of generic slots, an “Any 15” or an “Any 30”, and the system will hand back a generic slot when the agent searches for a specific visit. An “Any 15” might genuinely need to be an hour for a new patient. Some appointment types can only be booked if a companion type can also be booked. Sorting that out means knowing which specific appointment types each generic slot is actually eligible for, per provider, per department, and that mapping is different at every practice we onboard.

There is no clever conversation design that gets around it. Either the agent resolves the mapping against the live athenaOne configuration and books correctly, or it books into the wrong slot and the front desk finds out at check-in. Where a request matches nothing, the right answer is to stop and route the exception to a named person rather than guess. That kind of thing never appears on a feature grid, and it is most of the work. It is also the work a multi-EHR vendor has to redo for every system it supports, and the work we do for one.

The test that settles it

Ask both vendors the same three questions at Thrive, or on a call, and compare the answers side by side.

1. Which athenaOne records change when the AI completes a task, and what does it write back before a handoff? “It creates a task” and “it books the appointment, updates the chart and closes the tickler” are very different products.
2. What happens to the work it cannot finish, who sees the exception, and who maintains the workflow when our rules change?
3. What does the first 30 days live cost, and does it run on live patient calls in our own athenaOne environment or in a demo tenant?

Then run us on your own call volume for thirty days. The first 30 days live are free and the transcripts will tell you more than either vendor’s website will.