Healthcare
-
April 28, 2025
Somewhere between the conference booth and your front desk, dental AI lost the plot.
The demos are impressive. The decks promise eliminated denials, automated verification, and a front desk that finally has time to focus on patients. Then you go live, and six weeks later your biller is still spending three hours a day on the phone with payers, your denial rate looks the same, and you're left wondering if "dental AI" was just a buzzword the whole time.
Here's what most vendors won't tell you: a surprising number of companies marketing themselves as dental AI aren't running AI at all. Behind the dashboard, there's a team of offshore or outsourced workers doing the same insurance calls your front desk was already making — just cheaper, with a software interface layered on top to make it look automated. You're not buying intelligence. You're buying arbitrage.
The practices that have figured this out are asking harder questions. Not "does it use AI?" but "what is the AI actually doing, inside my practice management system, without a human in the loop?" Most vendors can't answer that.
This guide covers what dental RCM actually involves, where genuine AI transforms it, how to spot the tools that are faking it, and what purpose-built dental AI looks like when it goes all the way down to the data layer.
"AI" is being used to describe things that are fundamentally different from each other. That's a big part of why so many dental AI tools disappoint.
Generative AI is what most people think of first — ChatGPT, writing assistants, image generators. Tools that create content based on patterns learned from massive datasets. Generative AI is impressive for drafting and summarizing. It is not designed to log into a payer portal, read structured benefit data, and write it back into a database. That's simply not what it does.
Robotic Process Automation (RPA) is an older category that automates repetitive tasks by mimicking mouse clicks and keystrokes — recording what a human does on a screen and replaying it. RPA works for simple, predictable workflows and breaks the moment something changes. A payer portal updates its interface. A new field appears. A login times out differently. RPA fails silently, and nobody knows until the data is wrong or missing.
AI agents are different from both. An AI agent perceives its environment, makes decisions, and takes action to accomplish a goal. It navigates a payer portal the way a human does — adapting when something changes, reading what it finds, acting on that information in a connected system. It doesn't need a perfectly predictable workflow. It handles variability the way a trained person would.
The question isn't "does your tool use AI?" It's: what kind, and what is it actually doing?
For dental RCM — where every payer portal is slightly different, every plan has its own rules, and data needs to land in a specific place in your practice management system — the only AI category that can do the job end-to-end is an agent. Everything else is a workaround.
Revenue cycle management sounds like a finance term. In a dental practice, it's everything that happens between a patient booking an appointment and the practice collecting payment — and in dentistry, that process is more layered and more payer-specific than most people outside the industry realize.
Here's what it looks like in practice.
A patient schedules a crown. Before they sit down, someone needs to verify their insurance — not just confirm they're covered, but understand exactly what their plan covers for that procedure on that tooth, whether they've hit their annual maximum, whether a waiting period applies, and what their out-of-pocket will be. Get it wrong and you're either surprising the patient with a bill they didn't expect or submitting a claim that comes back denied.
Then the claim goes out. A dental claim isn't a simple form — it's a CDT-coded document that accounts for the specific tooth, surface, procedure, and the payer's rules around coverage. Delta Dental has different rules than Cigna. Cigna has different rules than MetLife. Every plan within every payer can have its own frequency limitations, missing tooth clauses, and coordination of benefits logic for dual-coverage patients.
If the claim gets denied — industry-wide, about 7% do on first submission — it has to be identified, categorized, corrected, and resubmitted before the filing deadline. Miss that window and the revenue is gone.
Then there's the patient balance. Whatever insurance didn't cover needs to be billed and collected — with enough persistence to actually get it, and enough care not to alienate a patient you want back in six months.
Every step is connected. A mistake at verification creates a problem at claims. A problem at claims creates a problem at collections. Most practices run this entire cycle on staff expertise, phone calls, and institutional knowledge that lives in one person's head.
AI fits dental RCM not just because it's fast, but because the work is rules-based, data-intensive, and repetitive in exactly the ways humans struggle and AI excels.
Verifying insurance for 30 patients requires checking 30 different plans across dozens of payers, against hundreds of procedure combinations, and getting results into your system before the morning huddle. A good coordinator can do this. It takes hours, it's exhausting, and errors happen. AI does it overnight — without fatigue, without variation, without anyone picking up a phone.
But here's where real AI and fake AI separate.
The payer portal problem
Most dental AI tools verify insurance the way a clearinghouse does: ping an eligibility API, get back a basic response. Active or inactive. Sometimes a coverage percentage. Rarely the full picture.
Payer portals contain far more than any API returns. Remaining maximums, remaining deductibles, frequency limitations, missing tooth clauses, waiting periods, coordination of benefits for dual-coverage patients — this is the information your coordinator gets when she logs into Delta Dental's portal, navigates to the member, and reads the plan. It's what determines whether a treatment plan is financially viable before the patient commits.
Real AI doesn't take the shortcut. It follows the human workflow — logging into payer portals exactly as your coordinator does, navigating the interface, reading the full benefit details — and then does something your coordinator can't: writes every piece of that structured data back into Open Dental automatically, into the correct fields, before your morning huddle.
Not a note. Not a PDF. Actual structured data — remaining max, remaining deductible, frequencies, limitations, coinsurance — populated in Open Dental and ready to use.
What this means on the ground
Your team walks in already knowing what every patient's plan covers for today's treatment. No scrambling. No surprises at the chair. The conversation shifts from "let me check on that" to "here's exactly what your plan covers and here's your out-of-pocket."
Claims status — the same approach, applied further
The payer portal advantage doesn't stop at verification. Claims status is the same problem with the same solution. Instead of your team splitting their morning across five different portals chasing outstanding claims, AI does the navigating — checking status directly at the source and updating Open Dental automatically. The information is there when your team needs it, without the manual effort.
Insurance verification and claims status are two pain points in a revenue cycle that has many — eligibility errors, claim denials, aging AR, patient balance follow-up, recall gaps. Every one is a place where admin work accumulates, revenue leaks, and staff bandwidth gets consumed.
AI can address all of them. But how deeply depends almost entirely on the practice management system it's working with.
Not all PMS platforms are built the same. Some are closed — they control what data can be accessed, what can be written back, and what third-party tools are allowed to do. Integrations are surface-level by design. You get what the vendor exposes and nothing more.
Open Dental is different. As an open-source platform, it allows direct access to the underlying data architecture in ways most PMS systems don't permit. That openness is what makes genuine automation possible — not just reading basic patient data, but writing structured benefit information back into the correct fields, updating claim statuses in real time, and building AI workflows that operate at the same level a trained human would inside the system.
This is why the depth of a tool's PMS integration matters as much as the AI itself. A sophisticated AI on top of a limited integration is a Ferrari engine with no wheels. The capability is there — the infrastructure to use it isn't.
For practices on Open Dental, the infrastructure is there. The ceiling for what AI can automate across the full revenue cycle is significantly higher than most practice owners realize.
This is what nobody in the dental AI market wants to talk about. But if you've bought a tool that promised to transform your revenue cycle and delivered a glorified reminder system, you deserve an honest explanation.
There are four failure modes. Most underperforming dental AI tools hit at least two.
They're not actually AI
The most uncomfortable truth in the market right now: a significant number of "AI-powered" dental solutions are running human labor behind the interface — offshore or outsourced teams manually doing the same portal lookups and insurance calls your front desk was already doing, with a dashboard layered on top to make it look automated.
The economics work for the vendor. Building a dashboard is faster than building real AI. Offshore labor is cheap. Most practice owners don't look behind the curtain.
But you're not buying automation. You're buying a more expensive version of the same manual process, with worse accountability and a monthly subscription on top.
Ask every vendor: what happens technically when my verification runs? Who or what accesses the payer portal? Is there a human in that process? If the answer is vague, you have your answer.
They use clearinghouses instead of payer portals
Tools running real AI often take a shortcut: they ping a clearinghouse API for a basic eligibility response instead of logging into the payer portal directly. It's fast and cheap — and incomplete.
Clearinghouse responses don't include the full benefit breakdown. No remaining maximums, no frequency limitations, no missing tooth clauses, no dual-coverage logic. That missing information is what causes surprises at the chair, rejected treatment plans, and claims denied for reasons that should have been caught before the appointment.
They can't write back to your PMS
Even tools that access the right data often stop short of the most important step: getting it into your practice management system in a usable form.
The workaround is a PDF. Or a note. Or a separate portal to log into. Which means someone still has to read it, interpret it, and manually enter it into the PMS. That's not automation — that's adding a step.
Real integration means structured data goes directly into the correct PMS fields automatically, with no manual hand-off.
They're built horizontally, not for dental
A lot of what's marketed as dental AI started as general-purpose automation — scheduling software, communications tools, healthcare AI — with a dental layer applied. That layer is thin. The tool handles simple workflows and falls apart when things get complex.
Dental insurance is complex by design. Payer-specific rules, plan-level exceptions, dual coverage calculations, frequency limitation tracking — this is not a general healthcare problem. It requires dental-specific logic built into the AI from the ground up, not bolted on after the fact.
The dental AI market isn't getting less crowded. The best defense is a better set of questions. These five cut through the noise — bring them to every demo.
1. Is a human involved anywhere in the verification process?
Ask it directly. A real AI system can describe exactly what happens technically when a verification runs — what accesses the portal, what data it pulls, how it gets into your PMS. If the answer includes "our team," "our specialists," or "our network," you're looking at an outsourced labor model. Move on.
2. Are you accessing payer portals directly, or through a clearinghouse?
A clearinghouse gives you active or inactive. A payer portal gives you everything. Ask which one — and if the answer is a clearinghouse, ask specifically whether remaining maximum, remaining deductible, frequency limitations, and missing tooth clauses are included. If they're not, you'll still be making manual calls for the cases that matter most.
3. How does the data get into my practice management system?
"We provide a report" means someone has to read it. "We send a PDF" means someone has to open it. "We have a portal for results" means another login. None of these are automation. The right answer: structured data writes directly into the correct PMS fields, automatically, with no manual step required.
4. Was this built specifically for dental, or adapted from something else?
Ask directly. Was it built on dental billing logic — CDT codes, ADA claim forms, payer-specific dental rules — or is it a broader platform with a dental layer? If it's the latter, probe hard on how it handles dual coverage, frequency exceptions, and plan-specific limitations. Those are daily realities in dental, not edge cases.
5. How deep is your integration with my PMS?
"We integrate with Open Dental" can mean anything from a full bidirectional data connection to a once-daily CSV sync. Ask: do you read from and write to Open Dental in real time? Which fields do you write back to? How long after a verification runs does it appear in the system? Specific answers are a good sign. Vague answers about "connectivity" are not.
There's a version of your practice where your team walks in every morning already knowing. Already knowing which patients are covered, what their plan allows, what they'll owe. Already knowing which claims cleared and which need attention. Already having done — overnight, automatically, without anyone staying late — the work that currently consumes the first two hours of every day.
That's not a fantasy. It's what happens when AI is built the right way, integrated at the data layer, and connected to your practice management system rather than bolted on top of it.
The practices that get there first aren't the biggest or the most tech-forward. They're the ones that asked the right questions before signing a contract. The ones that looked behind the dashboard and demanded a real answer.
Foji was built to be that answer — a purpose-built AI agent that goes directly to payer portals the way a human would, retrieves the full picture, and writes it back into Open Dental automatically. No humans in the loop. No PDFs. No portal-juggling before the morning huddle.
If that's what you've been looking for, we'd like to show you what it looks like inside your practice.
Transform Your Business with AI – Start Building Smarter Solutions Today!
The Framework for Operationalizing Joint Intelligence
