Pretty Good AI vs ReferralPoint

Two different referral problems

Referral management is one phrase covering two jobs that barely touch each other, and most evaluations go sideways because nobody says which one they are buying.

The first job is steering. A primary care doctor decides a patient needs cardiology. Which cardiologist, in which network, at what cost, with what quality record, how far from the patient, speaking which language. Get that wrong and the patient goes out of network, the total cost of care goes up, and the organization holding the risk absorbs it. This is a claims and analytics problem, and it is ReferralPoint’s origin story. They came out of Lightbeam Health, a population health company, and they state plainly that every match and score is grounded in claims data.

The second job is completion. A referral arrives at a specialty practice, usually by fax, often missing the imaging or the authorization. Somebody has to read it, match it to a patient, open a chart, work out what is missing, call the sending office, call the patient, and book the visit. Nobody has time, so a stack of them sits at day nine.

Both are real. Only one of them is usually the reason your schedule has holes in it.

Where the money actually leaks

Ask your referral coordinator how many referrals from last month never reached a first visit. Most practices cannot answer, which is the finding.

Then ask how long the average one waited. In a busy specialty practice the answer is usually somewhere between a week and never, and the reasons are boring: the fax was hard to read, the patient did not answer the one call anybody made, the authorization had not come back, or it was charted correctly and then nobody followed up.

None of that is a routing failure. The referral was routed to you. It just did not turn into a visit.

That work has no glamour and no license requirement, which is exactly why it stays undone when the front desk is short two people. Everyone triages what is in front of them, and a stack of referrals is never in front of anyone.

What we actually run

We are the AI operations platform for athenaOne practices, and the four pillars run on one integration and one patient memory.

Referral management. Inbound faxes are read and charted automatically, roughly 80% of them including handwritten ones, each with a confidence score, patient matching and deduplication, and a hold-for-review step so staff see the ones the system is unsure about. Referring providers can submit through a web form built to your specifications and hosted on your site instead of faxing. Patients without a referring provider can self-refer through the same review workflow. Campaigns get their own landing page and their own link so you can see which effort produced scheduled patients. Everything lands in one dashboard showing status, how long each referral has waited, and which ones have stalled on missing information, no patient response, or staff review. After review, the agent calls the patient and books.

Patient engagement. We answer the phone day and night, and we handle secure two-way text on the same integration. That is the part that makes the referral work finish, because the referral that has sat nine days needs a person on the phone, not a fourth text message.

Task automation. Cases, orders, requests and the outreach attached to them, run to your rules, with staff working from an athenaOne inbox rather than a separate queue.

Revenue cycle. Prior-auth prep customized by payer, status pulled from payer portals and written back into athenaOne, balances worked, and payment taken by secure link during the call.

Then the intelligence layer. We combine your athenaOne data with the result of every call, message and task we handle, and use it to see where patients get stuck and where provider time goes unused.

Say the honest thing about steering

Outbound referral steering is not a workflow we run today, and we do not hold a claims dataset.

Scoring specialists by network, cost and quality requires claims at a scale we do not hold, and ReferralPoint holds it through its parent. If you are a primary care group, an independent physician association, or anyone carrying downside risk on total cost of care, network leakage is a real and expensive problem, and steering it is not a workflow we run today.

The question to settle before you buy either of us is which side of the referral you sit on. A specialty practice does not get paid for steering. It gets paid when the patient shows up, and the gap between a referral received and a visit completed is where its revenue goes missing.

Depth is why the work finishes

A lot of tools connect to athenaOne. Far fewer finish work inside it.

Creating a chart or pushing a task reaches a handful of endpoints. Charting a fax with the right patient match, chasing the missing authorization, pulling status back from a payer portal, booking into the right appointment type with the right provider and writing the outcome back touches far more of the system, and every gap turns into a task for a human.

We hold production read and write access to 820+ athenaOne APIs, with no middleware in between, and our app runs inside athenaOne with no extra login. A platform built to close the loop across six EHRs has to build to what those six share. We took the other bet.

What this looks like in a real practice

Picture a four-site orthopedic group, twenty-two providers, and two referral coordinators handling everything that comes in by fax.

On a normal week about two hundred referrals arrive. Most are legible. Some are not. Roughly a third are missing something, usually imaging or an authorization that has not come back yet. The coordinators chart what they can, call the patients they have time to call, and the rest go in the pile.

By Friday the pile is sixty deep. Some of those patients booked with the group across town, and nobody will ever know which ones.

We start at the fax. Referrals get charted as they arrive with a confidence score, the ambiguous ones held for a human, the duplicates caught before they become two charts. The ones missing imaging get flagged the same day rather than the following week. Then the agent calls the patient, and calls again on a different day at a different hour if nobody picks up, and books the visit into the right appointment type. The coordinators stop doing data entry and start working the exceptions, which is the only part that needed them.

The referral dashboard shows the manager on Monday morning what is stalled and for how long. That number goes down every week, and it is a number the practice has never had before.

How to decide in a week

Three questions.

Which side of the referral are you on? Sending, and carrying risk on where they go, means the network problem is real and steering is not a workflow we run today. Receiving, and paid when the patient arrives, means completion is the problem and steering will not help.

How deep does the loop actually close? A platform that only builds for athenaOne goes further into it than one designed to close loops across six systems, because the six-system version can only use what all six have in common. Ask both vendors to show you the write-back, in the chart, on your own referral types.

What else is broken that day? Referrals are rarely the only thing the same staff are failing to get to. If the phone is also rolling to voicemail, the confirmation calls are not happening, and the balances are aging, a referral point solution fixes one queue and leaves the rest.

Before your next leadership meeting, count two things: referrals from last month that never reached a first visit, and the average days each one waited. Put a revenue figure on the first number. That figure usually settles the order of operations, whatever else you buy this year.