68% of practices have not changed staffing because of AI

"AI will replace the front desk." If you manage a practice you have heard some version of that from a vendor, from an owner doing sums on the back of an envelope, or from a receptionist who has quietly stopped asking about next year's rota. Before you build a business case on it, look at what practice leaders told the Medical Group Management Association (MGMA) in two of its own polls. Both contain the number 68%, and they point in opposite directions.

One number, two meanings

The first poll, an MGMA Stat poll of 30 September 2025 with 351 applicable responses, found that 68% of medical groups added or expanded AI tools in 2025. Another 31% did not, and 2% were unsure. MGMA reports that leaders who added AI overwhelmingly targeted clinical work, especially ambient documentation, scribing and dictation, provider note-taking and charting.

The second, an MGMA Stat poll of 2 June 2026 with 260 applicable responses, asked a different question: has your organization redesigned a role or adjusted staffing with the help of AI in the past year? Sixty-eight percent said no. Twenty-six percent said yes, and 5% were unsure.

Put side by side, those two results say that AI is now widespread in medical groups and that staffing has barely moved. Both are voluntary polls of MGMA's audience, so read them as a signal about your peers, not a national rate. Even so, anyone claiming the front desk is being swept away should be able to name the practices where it happened. The polling suggests they are a minority.

So what for you: if the replacement story reaches your owner or your team, you can answer it with a dated source rather than an opinion, and that lowers the temperature before you have proposed anything.

The kernel of truth: 26% did change something

Dismissing the myth outright would be wrong, because a quarter of leaders did change roles. A Physicians Practice write-up of the June 2026 MGMA data reports that the changes were mostly small and concentrated in scheduling, registration, prior authorization, billing and call-center work. MGMA's own article gives examples that include cutting or reassigning full-time equivalents, putting AI agents on the phones, and not backfilling scribes who leave.

Read the list as a description of tasks being absorbed and roles being rewritten, with attrition and reassignment doing most of the work. No example in it involves a whole desk closing. The risk to your team is real, but it is a risk to specific tasks, and you control the order in which those tasks change.

So what for you: treat "which of our tasks are exposed" as the question, and "do we still need a front desk" as the wrong one.

What the phones actually consist of

The strongest evidence against the replacement story is what the phone work contains. In an MGMA Stat poll of 10 March 2026, practice leaders were asked which phone task takes the most time. Eligibility and prior authorization came first at 45%, scheduling second at 31%, intake 9%, prescription refills 6% and other 9%. MGMA's write-up ties the top answer to insurance verification, coordination-of-benefits checks, prior authorization submissions, status tracking and denials.

Note that this is a count of leaders, not a share of call minutes. It does tell you is that the task most often named as the heaviest is insurance work that depends on payer systems and judgement calls, not the greeting and booking that most vendor demos show. Scheduling is the second-largest slice and is the natural place to start. It is also the part a tool can plausibly handle without a human, which is a smaller prize than the myth implies.

Phones are a pressure point but not the only one. In an MGMA Stat poll of 9 December 2025 with 236 applicable responses, leaders named their top patient access focus for 2026: no-shows 27%, online scheduling 24%, phone access 22% and wait times 21%. Your front desk is being asked to fix four problems at once, and AI touches each of them differently.

So what for you: before you price a tool, find out what share of your own call time goes to scheduling and what share goes to insurance, because the answer sets the ceiling on any saving.

A business case built on minutes

Here is the arithmetic, using a wage figure you can check. The Bureau of Labor Statistics (BLS) Occupational Outlook Handbook lists a May 2025 median wage of $45,930 for medical secretaries and administrative assistants. On a 2,080-hour year that is about $22.08 an hour, before benefits and payroll costs, so treat it as a floor for your own loaded rate.

Now an illustration, with numbers you should replace from your own call logs. Say your desk spends 10 hours a week on scheduling calls. At $22.08 that is about $221 a week, or roughly $11,500 a year in wage cost. If a tool took over every one of those hours, you would have freed a fraction of one salary, and the eligibility and prior authorization work would still need a person. Nobody loses a job on that arithmetic. What you gain is time: hours that move to the patients standing in front of the desk, to denials follow-up, and to the calls that need judgement.

That is the honest version of the capacity argument. The case that survives scrutiny from your owner is "we recover these hours and spend them on this", and it holds up because every figure in it comes from your own logs.

So what for you: price the minutes, not the person, and put the redeployment plan in the same document as the tool cost.

What the evidence supports doing first

The evidence points toward a narrow first step. Frankly, if you are a practice manager with no tool in place today, the sensible move is a four-week test on one task, usually scheduling or reminders, with a baseline measured first.

Start by tagging one week of call logs by task and recording the minutes. Pick the single task with the clearest rules and the lowest clinical risk. Ask any vendor for a signed business associate agreement before patient call audio is processed, and confirm how a caller reaches a human. Run the test for four weeks, then compare minutes, answered calls and complaints against your baseline.

Tell your team what you are doing and why before you start. The same MGMA data that shows few practices changing staffing also shows task-level change is already under way, and a team that hears about it from you early will give you better feedback than one that hears it from a vendor. Our earlier comparison of AI phone agents for physiotherapy scores three vendors against the same dimensions, and the piece on dental voicemail as a safety net covers what happens to calls nobody answers.

If you would rather have an independent read on which of your front-desk tasks are exposed and in what order to change them, the AI Opportunity and Growth Assessment scores this against your own workflows, and you can start with a free 20-minute discovery call.

So what for you: you do not need to decide whether AI replaces your desk. You need to decide which task changes first, and a measured four-week test lets you set that pace instead of having it set for you.

Frequently asked questions

Will AI replace the front desk at a medical practice? The polling says not yet. In an MGMA Stat poll of 2 June 2026, 68% of practice leaders said their organization had not redesigned a role or adjusted staffing with AI in the past year, and 26% had. Most changes were small and sat in scheduling, registration, prior authorization, billing and call-center work.

What phone tasks take the most front desk time? In an MGMA Stat poll of 10 March 2026, 45% of practice leaders named eligibility and prior authorization as their most time-intensive phone task, 31% named scheduling, 9% intake and 6% prescription refills.

How do I build a business case for AI at the front desk? Build it on hours returned, not headcount removed. Pull a week of call logs, tag each call by task, price the minutes at your own loaded hourly cost, and test one task for four weeks before you change any role.

Is it safe to put an AI phone agent on patient calls? Only after the basics are in place. Ask for a signed business associate agreement before any patient call audio is processed, confirm a human can take over any call, and keep eligibility and prior authorization work with staff until the tool has proved itself on scheduling.

Sources: MGMA Stat, "Document, schedule, communicate: Where ambulatory care has added or expanded AI in 2025", poll of 30 September 2025 (351 applicable responses); MGMA Stat, "AI is slowly redesigning work in medical practices rather than replacing workers", poll of 2 June 2026 (260 applicable responses); Physicians Practice write-up of the same MGMA data, 2026; MGMA Stat, "Phones are still a bottleneck costing medical practices time they can't afford", poll of 10 March 2026; MGMA Stat, "Patient access priorities for 2026", poll of 9 December 2025 (236 applicable responses); U.S. Bureau of Labor Statistics, Occupational Outlook Handbook, medical secretaries and administrative assistants, May 2025 median wage. MGMA Stat polls are voluntary and self-selected. The hours and dollar figures in the worked example are an illustration, not a finding. This article is general information, not legal or financial advice.

See also: the dental hygienist shortage, a staffing gap AI cannot fill.

Related: AI phone agents for physiotherapy compared · The dental voicemail safety net · The dental hygienist shortage