AI Won't Fix a Broken System. Here's What Might.

Last month we made you a promise. Once you see the administrative load clearly, the fixes get a lot less mysterious. We also promised no tool-of-the-week roundup, and no pretending technology solves what is fundamentally a systems problem.

We meant both. Let's start where the problem actually lives.

The system isn't only yours

Payer bureaucracy, prior authorization requirements, and documentation mandates come from outside your practice. No dashboard removes a fax-only prior auth process. No chatbot changes what an insurer requires before it approves an MRI. That burden is systemic, and it deserves the anger it generates.

But inside that larger system sits a smaller one: your practice. And that smaller system runs on decisions you make, one at a time, about what you do yourself, what someone else does, and what a tool does. That's the part worth examining before you buy anything.

Three buckets, not one tool

Every recurring task in your practice sorts into one of three places.

Stop. Some tasks exist because nobody questioned them. The daily report no one reads. The triple-confirmed appointment reminder your no-show rate doesn't justify. Before you automate a task, ask whether the task needs to exist at all. Automating a task that shouldn't happen just makes it happen faster.

Hand to a person. Some work needs judgment, relationship, or a human voice on the phone: a distressed patient, a nuanced insurance appeal, a scheduling conflict that calls for real negotiation. Automation here doesn't save time. It creates a worse version of the problem and pushes the real work back onto you when the automation fails.

Automate. This is the bucket AI and software genuinely fit: repeatable, well-defined, low-judgment tasks with a clear rule. Intake form collection. Inbox sorting by category. Appointment reminders. Basic documentation drafts a clinician still reviews and signs. These tasks have one right answer most of the time, and a tool applies that answer faster and more consistently than a tired person at 9pm.

Notice what's missing from that third bucket: anything involving clinical judgment, a hard conversation, or the kind of decision that defines whether a patient trusts your practice. Keep those out of the automation column, no matter how confidently a vendor promises otherwise.

Ask the pain point question before the tool question

Here's where most practices go wrong. They see a tool, like the tool, and try to find a use for it. That's backward.

The order that actually works: name the specific hour that hurts, name what's true about that hour (is it repetitive, does it require judgment, does a rule already exist for it), then ask which bucket it belongs to. Only after that do you look for a tool, and only for the tasks that landed in the automate bucket.

This is slower than downloading an app. It's also the only method that doesn't leave you paying monthly for software that solves a problem you didn't actually have.

The non-tech fixes still count

Not every fix has a login page. A written protocol that tells your front desk exactly how to handle a specific insurer's prior auth requirements removes ambiguity without any software at all. Saying no to a payer contract that costs you more in staff hours than it pays out removes an entire category of administrative load. Redesigning your intake process so a human collects the right information once, instead of three people asking for it three times, fixes more than most automation platforms do.

The question was never "AI or no AI." The question is which fix, tech or not, actually matches the problem in front of you.

One question to sit with

Pick the task that ate the most of your week last week. Before you look for software, ask which bucket it belongs in: stop, hand to a person, or automate. If you're not sure, that uncertainty is worth more of your attention than any tool on the market.

Tracy Cherpeski International works exclusively with independent healthcare practice owners on exactly this kind of sorting, the kind that happens before any tool gets chosen. If the three buckets raised more questions than they answered, that's the whole point of our upcoming roundtable, AI, Automation & the Smart Practice on Tuesday, August 18. You'll complete a live diagnostic that shows you where your own practice loses ground, and leave with a clear framework instead of a list of apps. Space is limited, and it's free to attend. Register here.

About the Author

Miranda Dorta is a writer and operations strategist who brings creative instinct to the work of building and sustaining a brand. A graduate of the Savannah College of Art and Design, she holds a B.F.A. in Writing with concentrations in Creative Writing and Fashion Journalism -- a foundation that shapes how she approaches everything from content development to client communications.

Her background spans public relations, publishing, retail management, and social strategy. At Tracy Cherpeski International, she serves as Manager of Operations and PR, where she oversees the day-to-day infrastructure of the brand while leading content production, podcast operations, email marketing, and communications strategy. She is the connective tissue between vision and execution.

Miranda resides in the outskirts of Raleigh, with her husband, daughter, dog and cat.