Albert Katz

The through-line

Katz keeps returning to a single conviction: the way to modernise musculoskeletal clinics is not to buy them but to give them software that does what an acquisition would have done. He started from the opposite premise. Backed by a family office with a partner, he set out to roll up neurosurgery groups, orthopedics, physical therapy and interventional pain clinics, and built a patient triage tool as an internal instrument of that roll-up [1]. The tool turned out to be the business. Once a pilot at Rush Medical showed that "if you added a couple additional filters you could turn this uh triage tool into a procedure recommendation tool where you could actually train the algorithm to recommend which treatments patients were not only financially but also medically eligible for" [1], the acquisition plan was abandoned in favour of selling the capability to clinics directly, so that "instead of having to acquire these clinics you can essentially leverage our AI services" [1].

The economic diagnosis underneath has stayed constant. Clinics are squeezed between falling reimbursement and rising costs, and the default endgame is a sale: "reimbursements are going down turn is increasing like patients are always unsatisfied it's it's just like selling to private Equity or to a health system is you're saving grace and you're just trying to your economics up to a point where they're willing to buy you" [1]. He continues to argue that too many practices operate as though nothing has changed, framing the problem as clinics still running on a 1998 model in 2026 and asking why most healthcare technology fails [2].

On why triage was the wedge

The original design logic was time allocation, not automation: route high-risk patients up to the physician and lower-risk patients down to the nurse practitioner or advanced practitioner, so that physician hours are spent on the patients likely to need procedures while script refills are handled further down the ladder [1]. That gets "the best bang for their buck" while still protecting care quality [1]. The system learns from the clinicians it serves rather than imposing a standard: recommendations that practitioners decline feed back into the model, and "the algorithm will learn from those recommendations and uh improve to practice in the same behavior as the practitioner" [1].

He is blunt that the intelligence is the easy part. "A lot of the work we do to be honest is just turning like cluttered unstructured data into something that essentially is more structured like something that's readable" [1], and the real difficulty sits there: "the hardest part isn't building an algorithm that can recommend procedures and acting like a doctor the hardest part is making sense out of a patient note" [1]. From triage the platform widened into scheduling, virtual care management, remote therapeutic monitoring and chronic care management, with behavioral health integration slated next, amounting to what he calls "a business in the Box for musco skeletal clinics" [1][3].

On building for clinics without touching the physician

Every product is tested against three rules: it has to improve patient outcomes, it has to either increase revenue or decrease expenses, and "you cannot affect the physician workflow at all like you can't single thing you can't make them hit an additional button it needs seamless" [1]. That forces deep EHR integration and a near-zero onboarding burden: "it's 15 minutes introduce me to your it team tell me who the office manager is and then tell me who your billing team is and everything else is kind of taken care of" [1]. Katz treats understanding how clinics actually think, clinicians and administrators alike, as the prerequisite for anyone building and fundraising in this market [1].

On leading with revenue, not savings

Katz deliberately sells new revenue before cost reduction. The triage tool and virtual care management are pitched on outcomes and revenue generation, and expense reduction has been left for later because "it is much harder uh I think at least for a a business to offer a service to clinics that is cheaper than what they're what they have now" [1]. Marginal savings do not overcome switching inertia, and he speaks from the buyer's side: managing a clinic, he passed on moving after-hours calls from a $33,000 vendor to a $28,000 one because the meetings required to re-specify patient routing were not worth $5,000 [1]. A service the clinic has never bought before, which produces incremental revenue, is an easier sale, so the strategy is to "come through the front door" with services that currently have no direct competitor and expand from there [1].

On taking the risk off the clinic's balance sheet

The virtual model exists because clinics cannot afford to hire their way to a full service line. In his own practice, they acquired a rheumatology group and then lacked capital for orthopedics and physical therapy, so they took a large office, offered discounted rent to an orthopedic group and a physical therapy group, handled the triage and tried to capture savings and better payer rates [1]. The productised version transfers the hiring risk: Flagler triages the patients and employs the physicians, who subcontract under the clinic, so the clinic books the revenue and pays a fee for service that is less than it generates, growing the service line "without having to take any underlying risk" [1]. He also discusses growth through partners rather than direct expansion alone [4].

On how doctors' attitudes to AI flipped

The adoption curve has been all-or-nothing. Launching in 2022, before ChatGPT entered general awareness, the reaction was flat disbelief: "I just remember every doctor saying like this is ridiculous this will never work" [1]. By late 2023 and into 2024 the pendulum had swung so far that selling became easy because "they believe AI can do more than what it can actually even do today" [1]. Katz's summary of the two eras is that there is "no middle ground" [1]. Notably, the resistance he encounters is not from physicians but from staff whose roles are being compressed, medical assistants and phone teams "who essentially see that their job that required let's say a team of 10 can now be done with team of four or five" [1].

On why small clinics beat health systems to adoption

He expects SMB clinics rather than hospitals to implement new technology first, and grants that he is biased by who he sells to, but the argument is structural. Small practices have one to three decision makers and no legal team obliged to tick every box, while hospitals present both a data problem, since "it's a nightmare to get into epic," and an integration problem across dozens of specialties unless the solution sits in an isolated division [1]. Clinics are technically tractable because they are homogeneous: a handful of specialties, typically fewer than eighty to a hundred doctors, and "they code the same way it's it's very symmetrical" [1]. The commercial consequence is decisive: clinic sales cycles run one to three months against one to two years for hospitals, which is why some sub-billion-dollar startups have only two to four hospital customers as their entire client base [1].

On independence versus consolidation

Asked whether AI is a David versus Goliath story for private practice, Katz declines the framing because Flagler serves both sides, selling to private equity backed platforms that are acquiring clinics and to mom-and-pop practices trying to stay independent [1]. He settles on being "the tide that's going to raise All Ships," and defines success in either direction: practices that want to stay independent will be able to, and those that want to sell will do so at a better multiple [1].

On behavioral health as the missing MSK measurement

Katz argues mental health is inseparable from musculoskeletal and pain care, particularly for interventional pain and ortho spine patients on high-dose opioids, where physicians need to know whether there is substance use disorder, depression or escalating dosage, and "if you are a doctor with 2,000 patients that you have to manage every year it's impossible to keep track of all of that right and right now clinics aren't doing it" [1]. His answer is the same turnkey pattern: a ten-question CMS-approved form pushed to every scheduled patient, giving a gauge of mood and stability, which he sees as "one of the things we could be doing today to help battle the opioid crisis" [1]. He is candid that no physician wants patients on opioids but that alternatives sometimes do not exist for severe pain, which makes monitoring the obligation [1]. He is equally candid about the gap that remains: "I don't think that we've been monitoring that well enough and I don't think we've created very good analytics around how to measure depression" [1], with better triage of high-risk depressed patients to the right provider set as a 2025 focus [1]. That, he says, is the nature of startups, where solving one problem exposes a million more and the work becomes deciding what takes priority [1].

Takeaways

• The roll-up thesis was abandoned when a triage tool built to serve it proved sellable on its own, converting an acquisition strategy into a software one [1].
• Adding filters to a risk-triage model turned it into a procedure recommendation engine covering both medical and financial eligibility, validated in a pilot at Rush Medical [1].
• Structuring messy clinical notes is the real engineering problem: "the hardest part isn't building an algorithm that can recommend procedures and acting like a doctor the hardest part is making sense out of a patient note" [1].
• Every product must clear three tests: improve outcomes, move revenue or expenses, and add zero steps to the physician's workflow [1].
• Sell new revenue first, because marginal cost savings never justify the switching friction inside a clinic [1].
• Flagler employs the subcontracted physicians and therapists so clinics can extend service lines without hiring risk, paying a fee for service below what they generate [1].
• Physician sentiment moved from "this will never work" to overestimating what AI can do, with real resistance coming from medical assistants and phone staff whose teams shrink [1].
• Small clinics adopt faster than health systems: one to three decision makers, homogeneous coding, and a one-to-three-month sales cycle versus one to two years [1].
• Behavioral health screening via a ten-question CMS-approved form is his practical lever against the opioid crisis in MSK care [1].