Pretty Good AI vs Infinx

A services company and a software company are not the same purchase

Infinx has been doing revenue cycle work since 2012. Prior authorization, eligibility, estimates, insurance discovery, coding, credentialing, charge capture, billing, payment posting. It sells to physician practices, hospitals, imaging centers and laboratories, and it does that work with a combination of automation and a large specialist workforce. Its own site puts headcount at 7,200+ across the United States, India and the Philippines. KKR announced an investment on 21 May 2024, and Infinx says it counts over 400 healthcare organizations as customers.

That scale is the product. When a practice buys Infinx it is buying capacity it does not have to hire, across a wider span of the revenue cycle than any front-office platform covers.

Pretty Good AI is a different purchase entirely. We are software that runs inside athenaOne, with production access to 820+ athenaOne APIs, and we are built around the patient-facing half of the practice: the call, the text, the schedule, the referral, the form, and the revenue work that touches all four.

Practices confuse the two because both say prior authorization. It is worth separating them properly before anyone builds a comparison grid.

What the marketplace record says, and what it does not

Infinx has been listed on the athenahealth Marketplace since January 2022. As of 19 September 2026 the listing showed 8 practice connections, concentrated in orthopedic surgery, and it has not been updated since December 2022.

That is not a verdict on the company. A business of that size does not depend on one marketplace listing, and its customers skew toward hospitals, imaging and laboratory organizations rather than athenaOne practices.

It does tell an athenaOne practice something useful. Depth in a specific EHR is built, not assumed, and it comes from doing the work repeatedly against that system. When athenaOne is the system your whole day runs in, ask any vendor how many athenaOne customers it runs today, what it reads and writes in the record, and whether you can watch it do so.

Augmented intelligence is a precise description, and it matters

Infinx describes its own prior authorization product as augmented intelligence: AI and robotic process automation, with exception handling by experienced specialists. It publishes up to 70 percent reduction in manual effort, a 98 percent plus accuracy rate and 99.5 percent adherence to turnaround times.

That is a careful claim and it is worth taking at face value in both directions. It is not a claim of full autonomy, and it is not manual outsourcing with a dashboard. It is a pipeline where software handles the common path and people handle the rest.

The practical consequences of that model are real and they cut both ways. Anything can be absorbed, because there are people behind it, and turnaround is governed by a service level rather than by a system call. The work happens next to your system rather than inside it.

Our model puts the decision and the action in the same step. The agent holds the conversation and writes to athenaOne while it is happening: booking the slot, routing the order, attaching the document, updating the authorization status. Containment is not a function of voice quality, it is a function of integration depth. An agent can finish a task end to end only because it can read and write the live record.

The complication that decides the boundary

The following is an illustrative workflow, not a customer story.

An orthopedic group with six sites outsources authorization work and keeps its own front desk. A patient is scheduled for an MRI three weeks out. The authorization request goes to the outsourced team. Two days before the appointment the payer asks for additional clinical documentation.

The outsourced team logs the request correctly. The note lands in a shared queue. Nobody at the practice sees it before Monday, because the person who watches that queue is covering the phones.

On the day, the patient arrives for a scan that is not authorized.

Nothing failed technically. The gap is that the system holding the authorization status could not act on the schedule or reach the patient, and the system that could was not looking.

What closes it is not a better vendor on either side. It is one system that can see the authorization is at risk, move the appointment inside the practice’s own rules for notice and provider matching, call the patient to explain, and leave the case in the staff queue with the reason attached. That is the work we do, and it only works because the same platform holds the status, the schedule and the phone line.

One case still stays with a person. A patient who arrives anyway and is upset is a service recovery conversation with real judgment in it. Route it to a human with authority to make it right, and stop automated contact on that account until they have.

The regulatory clock underneath both

Authorization workflows are being reshaped by rule as well as by software. Under the CMS Interoperability and Prior Authorization Final Rule, impacted payers were required to implement certain prior authorization provisions from 1 January 2026, with the API requirements landing primarily from 1 January 2027.

Payer status and decision data is going to get more structured and more available. The question for a practice is what reads it. A feed that produces a better-informed queue still produces a queue. A system that can read the status and then move the appointment, call the patient and close the loop turns the same data into a completed task.

Who each one is for

If the pressure is across the whole revenue cycle, including coding, credentialing and claims, across more than one system and more than one facility type, that is a scope no athenaOne front-office platform covers, and Infinx is built for exactly that span.

If the pressure is that athenaOne practices are drowning in patient contact, that calls go unanswered while staff sit in payer portals, that referrals age in a fax queue, that forms come back late and authorizations are incomplete on the day, the answer is depth in the system your practice actually runs on.

Many groups have both pressures, and running both is normal. The line worth agreeing before anything goes live is which system contacts the patient about money, and when.

Why we can shape this around how you already work

Production access to 820+ athenaOne APIs is what makes an unusual workflow buildable at all. Most marketplace vendors pick a narrower wedge and build deep inside it; we carry significant depth across the whole of athenaOne.

So we map the process your authorization and billing teams already run, in the order they already run it, and we automate that. You do not re-cut your payer rules, your queue conventions or your escalation path to fit a product’s opinion. Scoping a new workflow is measured in days and weeks, inside an onboarding window that is already weeks long, and it is refitted as your payer mix and volumes change.

Tell us how your front office and your authorization queue run today, and we will tell you what we would build against them.

Clinical decisions stay with your care team.