Every practice manager I’ve talked to knows the problem. The phones are ringing, the front desk is swamped, and patients are sitting on hold listening to the same loop of hold music. Some of them hang up. Some of them don’t call back.

That’s not just a patient experience problem. That’s a revenue problem — and most practices have no idea how large it is.

The data is sobering: the average hold time at a medical practice runs close to two minutes, and most patients won’t wait longer than one. More than 60% abandon the call if they’re held for over a minute. Across practices of all sizes, around 23% of incoming calls go completely unanswered. If your practice handles 100 calls a day, you are almost certainly losing 20 to 25 patient interactions every single day.

The lifetime value of an established patient is estimated at $12,000. A new patient generating four visits a year is worth $800 annually from day one. Do the math on what 20 missed calls a day costs over a year. For most practices, we’re talking six figures.

The good news is that this is one of the most solvable operational problems in healthcare. The front desk isn’t losing patients because the team is bad at their jobs. They’re losing patients because they’re spending most of their time answering the same ten questions over and over — appointment times, insurance acceptance, directions, prescription refill status, parking — while the actual patients with complex needs wait.

AI can handle the ten common questions. That’s where I’d start.

Why the Hold Time Problem Is Structural

Before we talk about solutions, it’s worth being honest about why hold times are so persistent.

Front desks are staffed for average demand, not peak demand. Monday morning, right after a weekend. Friday afternoon before a holiday. The 8:00 AM rush when the practice opens. During those windows, your team is physically incapable of answering every call immediately — not because they’re slow, but because 30 calls arrive in 15 minutes and there are two people at the desk.

Adding staff is the obvious answer, and plenty of practices do it. But you’re hiring people to handle volume peaks that last two or three hours a day. The rest of the time, those same people are idle or handling administrative tasks. That’s an expensive solution to what is fundamentally an availability and routing problem.

The structural fix is to separate the calls that need a human from the calls that don’t. Most practices never do this, which means their best clinical coordinators spend half the day giving directions and confirming appointment times.

The Five Calls That Don’t Need a Human

In every medical practice I’ve seen, the majority of inbound calls fall into a handful of categories:

Appointment confirmations and reminders. Patients calling to confirm their upcoming appointment, ask what time they need to arrive, or check whether they need to bring insurance cards or identification. These are pure information lookups that require zero clinical judgment.

Prescription refill requests. “Can I get a refill on my metformin?” The request still needs to go to the provider, but the intake process — capturing the medication name, pharmacy location, and patient details — is fully automatable.

General directions and parking. You might be surprised how many calls this accounts for, especially in practices with multiple locations, shared buildings, or limited parking.

Insurance questions. “Do you take Blue Cross?” “Is my copay the same as last time?” Again, these are lookups, not clinical interactions.

Lab result status. “I had bloodwork done last week. Are my results back yet?” In many practices, these calls land at the front desk, who then has to transfer to a nurse, who may or may not be available. The initial intake step is completely automatable.

Together, these five categories represent a large portion of inbound call volume at most practices. If an AI voice agent handles those calls — answering instantly, collecting information where needed, logging everything in the EHR or practice management system — your human staff is suddenly available for everything else.

What AI Call Routing Actually Looks Like

I want to be practical here because there’s a lot of vague language in this space about “AI” doing things that are really just basic phone trees.

A genuine AI voice agent for a medical practice does a few specific things:

It understands natural language. Patients don’t say “option three for prescriptions.” They say “I need to refill my blood pressure medication.” The AI needs to understand that and route accordingly — or, better, handle it directly.

It answers questions from practice-specific knowledge. Your hours, your locations, your providers, your insurance panel, your parking situation, your new patient intake process. The AI is trained on the specifics of your practice, not generic healthcare responses.

It collects and logs information. When a patient calls about a prescription refill, the agent captures the medication, the pharmacy, and the patient’s date of birth for verification, then creates a task in your EHR. Your staff sees a structured request waiting for them instead of a pink slip from a voicemail.

It escalates intelligently. The AI knows when to stop and transfer to a human. Anything clinical — symptoms, urgent concerns, anything that sounds like it might need a nurse triage — routes to a human immediately with a summary of what was discussed.

It handles overflow. When your team is on other calls, the AI answers instantly. No hold time. No “please hold while we assist other callers.” The patient gets help immediately or is offered a callback if the request needs human involvement.

This is what an Omni Voice deployment looks like in a medical practice context. It’s not replacing your front desk — it’s removing the category of work that doesn’t require your front desk while ensuring patients who need real human attention get it faster.

What to Measure Before and After

The practices that implement this well start by baselining their current call metrics. Most practice management systems or phone systems can pull this data:

• Average hold time
• Call abandonment rate
• Call volume by hour of day
• First-call resolution rate (how often does the caller get what they need without a callback)
• Calls by category (how many are prescription refills vs. appointment questions vs. clinical)

That last one often requires a manual audit of a week’s worth of calls or a brief survey of your front desk staff. It’s worth doing. The category breakdown is what tells you how much volume an AI agent can actually handle and gives you a realistic ROI estimate before you commit to anything.

After implementation, you’re tracking the same metrics. In practices I’ve worked with, the results typically look like:

• 50 to 70% reduction in average hold times during peak hours
• Abandonment rate drops by half or more
• Front desk staff time spent on pure information calls drops significantly
• Overflow calls (those that would have gone to voicemail or been lost) are captured and handled

The staff impact is notable too. When your team stops spending half the day answering the same questions, they have capacity for the interactions that actually require them — complex scheduling, insurance issues, billing disputes, patients who are anxious or confused and need a real person to talk them through something.

Common Objections I Hear

“Our patients are older and won’t use AI.”

The voice channel is actually the most accessible form of AI for older patients precisely because it doesn’t require an app, a portal login, or any digital literacy. They call a number. A voice answers. If they want to speak to a person, they say so or press zero. The experience from the patient’s perspective is often indistinguishable from a human receptionist for the call types I described above.

“We tried an automated phone system and patients hated it.”

Traditional IVR — press one for this, press two for that — is genuinely frustrating, and patients have every right to hate it. Modern AI voice agents are a different category of technology. They don’t require you to navigate a menu. You just say what you need, in whatever words come naturally to you. The improvement in patient satisfaction scores we see when practices move from IVR to conversational AI is significant.

“What about HIPAA?”

This is the right question to ask of any vendor, and you should demand a clear answer. A compliant voice AI deployment encrypts audio data in transit and at rest, does not retain identifiable patient audio beyond what’s required for the specific transaction, operates under a Business Associate Agreement, and logs interactions in a way that supports audit requirements. These are standard requirements for enterprise healthcare deployments. If a vendor can’t answer these questions clearly, that tells you something.

Where to Start

If I were advising a practice manager today, I would start with a one-week call audit. Pull your call volume data and have someone manually log the category of a sample of incoming calls — even 50 to 100 calls is enough to build a reliable picture.

Then identify the single highest-volume repetitive call type. In most practices, it’s appointment confirmations or prescription refills. Build your first AI-assisted workflow around that one category. Run it alongside your existing setup so you can compare.

You don’t need to overhaul your entire phone system on day one. The goal is to demonstrate, with real data from your own practice, that patients are getting faster responses, staff load is decreasing, and the AI is escalating appropriately when human judgment is needed.

From there, you expand. Add the next call category. Then the next. Within three to six months, most practices that take this approach have transformed their phone operations from a daily source of stress into something that largely runs itself.

The front desk is still there — and in most cases, the people are doing more satisfying work, because they’re spending their time on actual patient care rather than directions and insurance confirmations. That’s a better outcome for everyone.

If you want to talk through what this would look like for your specific practice — volume, EHR system, patient demographics — that’s exactly the kind of conversation our team has with healthcare providers before any deployment. You can book time here: https://calendly.com/sam-mckay/omni-by-enterprise-dna-virtual-workshop-call

No pitch deck. Just an honest look at whether this is the right fit and what results you should realistically expect.