A Front-Office Automation Checklist for athenahealth Practices

A front-office automation checklist is worth more before you sign than after, because almost every number you would want to compare against gets harder to reconstruct the moment the rollout starts. Most practices find this out at the renewal conversation, when nobody can say what changed.

The pattern repeats across group practices of every size. The purchase is made on a good demo and a plausible business case. The rollout goes fine. Nine months later someone asks whether it worked, and the answer is a set of impressions rather than a comparison.

This is not carelessness. Front-office performance is genuinely hard to see. The phone system holds one piece, the practice management system holds another, and the queues that matter most are the ones nobody reports on because they have no owner.

There is also a timing trap. Once automation is answering calls, the pre-automation baseline no longer exists anywhere except in whatever you captured on purpose. You cannot go back for it.

So the checklist below is ordered the way the work happens: what to capture before anything changes, what to require of the vendor in writing, and what to review at 90 days. It is meant to be taken into the room rather than read once.

Capture the baseline before anything changes

Six numbers, all obtainable in a week, and all impossible to reconstruct after go-live.

Start with the phones: call volume by hour of day, abandonment rate, and average time to answer. Then the work behind the phones, which is the part practices skip. Depth of the open order and follow-up task queue. Count of unworked referrals waiting on intake. Count of open records requests. Those four queue counts are the ones that predict whether automation will feel like relief or like a new dashboard.

It helps to know where the time actually goes before you decide what to automate. Practice leaders pointed to a familiar ranking of time-intensive phone tasks for their staff: eligibility and prior authorization at 45%, scheduling at 31%, intake at 9%, and prescription refills at 6%. If your own distribution is different, that difference is the most valuable thing on your baseline sheet, because it tells you which capability to weight in the evaluation.

Capture all of it by department rather than for the group. A group average is the number most likely to hide the one site that is failing.

Count the backlog in dollars, not in items

Queue depth is the honest measure of a front office, and it converts to money more directly than call metrics do.

An unworked follow-up order is a visit that a clinician already decided should happen. An unprocessed referral is a new patient who is currently calling somebody else. Both are revenue that has been earned in every sense except collection, and counting them in items understates what they are.

Do the arithmetic once, with your own figures, and put it on the page where the decision gets made. Suppose a group carries 1,400 open follow-up orders, converts 30% of them when someone actually calls, and books an average visit value of $180. That is 1400 x 0.30 x $180, or $75,600 sitting in a queue.

Illustrative model. Figures are assumptions, not measured results. Use your own queue depth, conversion rate, and visit value.

The reason to run this before the evaluation rather than after is that it changes what you shop for. A practice that discovers most of its frozen revenue is in unworked orders should be weighting outbound queue work heavily, and should be unimpressed by a vendor whose entire pitch is inbound call handling.

Require per-call review, not a satisfaction number

Quality measurement is where front-office automation contracts are weakest, and it is the easiest thing to fix in advance because you are writing the requirement.

Ask for per-call scoring against criteria your practice defines, with a review console your own staff can open. Ask what sample rate is reviewed, who reviews it, and how a failed call becomes a change in behavior. A vendor that reports only aggregate containment or satisfaction is reporting on themselves.

Patient-reported access is worth measuring on a standard instrument rather than a bespoke survey, so the number means something outside your building. The CAHPS Clinician and Group Survey measures include getting timely appointments, care, and information, which maps closely to what front-office automation is supposed to move. Using an established measure also protects you from the temptation to grade on a survey written to produce a good result.

One practical requirement to add: transcripts and outcomes exported to somewhere you control. Contracts end, and the record of what your patients were told should not end with them.

Measure by department, because the group average lies

Multi-specialty groups fail this step in a specific and repeatable way.

Every meaningful front-office rule resolves per department. Appointment types, provider groups, check-in requirements, and the buckets cases route to are all department-level configuration, and they drift apart over years without anyone deciding they should. Two sites in the same group frequently have different names for the same appointment type and different rules about who can hold it.

So require reporting cut by department from the first week, not as a later enhancement. Booking and fill rate by location, appointment-type utilization, wait time to appointment by type and provider, and referral funnel conversion per site. When a rollout underperforms, the cause is usually one department’s configuration rather than the product, and a group-level number will never show you that.

A related check worth doing before go-live: reconcile the provider roster on your website against the roster in the practice management system. Those two disagree at most groups. Whichever one the automation reads, patients will be routed by it, and finding out which providers actually work there and take appointments is step one before any rule can be written.

The 90-day review that actually settles it

Set the review date at signature and put three questions on it. Not a demo of the dashboard. Three questions.

First, which queue is smaller. Compare the same four counts you captured at baseline, by department. If none of them moved, the automation added coverage without removing work, which is a real outcome but a different one from the business case.

Second, what did staff stop doing. This is the question that separates augmentation from replacement of a task, and it deserves an honest answer even when it is uncomfortable. Only about 26% of practice leaders say their organization has redesigned a role or adjusted staffing with the help of AI in the past year, while 68% say they have not. That is the base rate you are trying to beat, and beating it requires deciding in advance which work moves.

Third, what got turned off. Every rollout has a workflow that was enabled and then quietly disabled because it did not work. Find out which one, and why, before the vendor’s renewal deck tells you a different story.

Run the review with the baseline sheet on the table. Everything on this checklist exists to make that one meeting short and factual.

Key Takeaways

Capture six baseline numbers before go-live: call volume by hour, abandonment, time to answer, and the depth of the order, referral, and records queues.
Record every baseline by department rather than for the group, since group averages hide the site that is failing.
Convert queue depth into dollars with your own conversion rate and visit value, and label it as a model rather than a measurement.
Require per-call scoring against criteria you define, a review console your staff can open, and transcripts exported somewhere you control.
Reconcile the provider roster on your website against the one in the practice management system before any routing rule is written.
Book the 90-day review at signature and answer three questions: which queue shrank, what staff stopped doing, and what got turned off.

None of this requires an analytics team or a new system. It requires writing six numbers down before the rollout and putting a date on the calendar to look at them again. The practices that do that have a factual renewal conversation, and the ones that do not have an argument.

Sources