Pretty Good AI vs Yosi Health

Two ways to make the phone stop ringing

Yosi Health publishes a clear claim about phone volume: direct messaging cuts it by 70 percent. Move patients to self-scheduling, digital check-in and messaging, and most of them stop calling.

That works. It is the right strategy for a large share of practices, and Yosi is unusually good at it. Best in KLAS 2024 for Patient Intake Management, 306 practice connections on the athenahealth Marketplace, 120 reviews at a 4.98 average. Those are the strongest numbers in the digital front door category and they were earned over nearly a decade.

We built the opposite thing. Our agent answers the phone.

Both approaches reduce the same cost. They just make different bets about your patients.

The bet each product is making

A deflection product is betting your patients will self-serve if you give them a good enough digital option.

An answering product is betting that some meaningful share of them will call no matter what you build.

Neither bet is right in general. Both are testable in your own data in about twenty minutes.

Pull your last full month of inbound calls and sort them by reason. Then ask a harder question than the obvious one: not “could this have been a form”, but “would this particular patient have filled one in”. A 34-year-old rescheduling a physical, yes. A patient calling because their insurance changed on the first of the month and they do not know if you still take it, less often. Someone calling about their mother’s appointment, almost never.

What tends to come back is that deflection removes the simple calls first. That is genuinely valuable, and it is also why the calls left over are harder than the ones that went away. Your average handle time goes up even as your call count goes down, and the staff time you saved is smaller than the volume drop suggests.

That is not an argument against deflection. It is an argument for knowing what is left.

What we do not build

We do not collect clinical screening information. No review of systems, no history of present illness, no symptom questionnaires. Yosi publishes all of that as core product, filed into athenaOne as structured data.

That is a deliberate line rather than a roadmap item. Our agent handles front-office logistics: booking the visit, confirming the coverage, chasing the paperwork, routing the request to the right person. When a conversation turns into anything a clinician should be reading, it goes to a human. We are not going to end up doing screening later, and a practice that wants pre-visit clinical information should buy a product built for it. Yosi is a strong one.

There is a certification difference worth naming too. Yosi Health is listed as HITRUST certified on its marketplace record. Our own HITRUST i1 is in audit rather than held, alongside HIPAA with BAAs, SOC 2 Type II, ISO 27001 and GDPR that we do hold today. If HITRUST is a hard gate in your procurement process right now, that is a real difference and you should weigh it.

Where the calls actually survive

A multi-site pediatric group we work with had already deployed digital intake and had good completion rates on it. Their form completion was not the problem. Their phone line still was.

The reason turned out to be structural rather than a failure of the intake product. A large share of their remaining calls involved someone who was not the patient and was not straightforwardly the account holder either. A parent calling about one child while another child’s appointment was the one in the system. A grandparent doing pickup who needed the visit moved. Custody arrangements where the calling parent was not the guarantor on the account.

None of that is a form problem. The information needed to resolve those calls sits across the patient record, the guarantor record and the appointment, and the person calling frequently cannot supply the identifier that ties them together.

The approach that worked read the existing athenaOne structures rather than inventing new ones: patient and guarantor relationships, appointment types and their scheduling rules, and the department the appointment sat in. The agent could resolve the common shapes, including moving a sibling’s appointment once it had confirmed the caller against the account.

It handed off in one specific case, deliberately. When the caller was not on the account at all, the agent stopped, captured what it had, and routed it to a staff member. Deciding whether an adult who is not on a minor’s account may reschedule that minor’s visit is a judgment about that family, and it belongs to a person who can make it and document why.

That handoff is the point. The measure of a front-office agent is not how many calls it finishes. It is whether it knows which calls it should not finish.

Running both

Most practices that buy both end up with a clean division: Yosi owns the digital front door and everything a patient does before they arrive, our agent owns the inbound phone line and the requests that come through it.

The friction is not architectural. It is that both platforms can send appointment reminders, and a patient who gets two of them for one visit assumes somebody made a mistake.

Decide before go-live which system owns each outbound message type, write it down, and check it again after the first month. That single decision prevents nearly all of the double-messaging problems we see when two capable platforms are live in the same practice.

Questions worth asking both of us

Ask each vendor to handle your hardest recurring case live, in an athenaOne environment rather than a sandbox. For Yosi, that means the intake flow for your most complicated appointment type. For us, it means the call your front desk least wants to take.

Then ask what happens when the system cannot complete the request, because the handoff is where front-office automation actually breaks.

Then ask how many athenahealth API endpoints each product connects, and what its marketplace connection count and review history look like. Yosi’s answer on that last question is better than ours. Ours is better on the first.