On episode 274 of DEO's Growth Secrets podcast, released 16 July 2026, host Kendall talked with Kaylie's co-founders, Louis Chan and Vere Whittome, about the walls between insurance verification, treatment coordination and claims. Their argument, in one line: most of what goes wrong on a claim was decided at verification, so verify earlier and in more depth, bring every EOB's lesson back to the next verification, and quote the patient a conservative estimate. The episode runs 46 minutes; the main points follow. Quotes are transcribed from the episode's audio, with filler words removed.
Why the one-star reviews are about billing, not the dentist
Louis ran a group of oral surgery practices in the San Francisco Bay Area before Kaylie, and the reviews there followed a pattern every operator knows: the patient praises the surgeon, then turns on the front office when a balance bill arrives. Vere recalled one that Louis had shared, edited four months after the visit: "The doctor saved my life. The billing team is a scam."
Louis's explanation is the time lag. The office verifies, gives a good-faith estimate and submits the claim, and then "maybe three, four rounds fighting the insurance company" pass before the balance bill goes out, and "that's the only thing that the patient sees." In a specialty practice that runs on referrals, one angry patient who complains to the referring dentist can cost the practice every future referral from that office.
Oral surgery makes verification harder. Louis put it at about 60 procedure codes to check, where a pediatric office might need 18, and said a category percentage is not enough: "I can't just tell them, hey, oral surgery is going to be covered at 80%." The bone graft may pay nothing, and the implant may hit a missing-tooth or replacement clause.
Every EOB should teach the next verification
The idea both founders came back to was closing the loop between claims and verification. An experienced insurance coordinator carries dozens of carrier rules in their head, Louis said, and each one is "a scar of an insurance claim that got denied." His example: at some carriers, taking out a single third molar under sedation because the patient is anxious does not qualify, but two wisdom teeth in different quadrants do. Every oral surgery treatment coordinator knows that one. Fewer know that a plan's rider excludes congenitally missing teeth, and that rule usually sits in a box of old EOBs rather than in front of the person building the treatment plan.
The second example was the out-of-network fee. A breakdown that says "covered at 80%" leaves the office asking "80% of what?" The first paid claim answers it, because the EOB states the allowed amount against the office's charge. Louis's advice was to quote from that allowed amount, not from the office's own fee, and to lean conservative: patients "will rather get a refund rather than getting a bill."
Vere's closing line on the episode: "every EOB is a lesson, whether it's your fees, whether it's denials, limitations, all of that, make it a lesson, bring it back into your insurance verification."
Verify earlier than feels necessary
Most practices, Louis said, run the revenue cycle in a reactive mode. Some offices verify 21 days before the appointment, which surfaces the inactive plan, the medical card handed over instead of the dental one, and the wrong subscriber while there is still time to ask the patient. His own practices used what he called a 7-2 system: verify everything seven days out, then re-verify two days before the visit.
Cost decides how deep an office can go. Louis put a front-office wage in the Bay Area at $35 to $50 an hour and a Medicaid prophylaxis allowed at $50 to $60; Kendall put the cost of a verification at $3 to $8 a patient, which high-volume pediatric and oral surgery offices could not afford on every patient. Louis's practices also left any outstanding claim under $100 alone, because two hours of a $35 coordinator's time costs $70 before the claim is won.
Where AI still falls short
Asked where AI fails, Vere named context first. AI does not fix the structure the insurers created, and it works well only when the office's own setup is organised: which networks each provider is credentialed in, and whether a plan is contracted directly or through a leased network. Louis added consistency: ask a language model the same hard question three times and the answers differ slightly, which is not acceptable for a deductible. Vere added cost: at high volume and low margin, AI has to be used only where it pays for itself.
Neither founder claimed full automation. Louis described the goal as 80 to 85% of verification work automated, with the team spending its time on the rest.
What to change on Monday
Louis's advice depended on the size of the organisation:
- A single practice: automate some of the verification work, and before trusting any tool or person, check how accurate the current process really is.
- Groups in the tens of locations: stop verifying separately at each office. Build one team, standardise the verification process and give that team the tools. When the portal says out of network and the office manager disagrees, call the carrier and settle it.
- Twenty or more locations: put the EOBs and denials into structured data, sort denials into categories, and watch for variation inside one group number. A mother and child on the same group can carry different maximums and different preventive, basic and major percentages.
His last point was about priorities: before fixing a problem, count how many patients and plans it touches, because "you might be chasing a 5% problem with 80% of your time."
Related questions
- Why out-of-network estimates are wrong: MAC vs UCR
- Is general anesthesia covered for wisdom teeth removal?
- Does the missing tooth clause apply to implants?
Figures in this article are the speakers' own, as said on the episode released 16 July 2026.