Regard

With technology where it is today, things that once took years to build can be replicated in hours. The old sources of competitive advantage, the ones rooted in technical complexity, are giving way. Tools like Claude Code make it possible to go from idea to prototype in an afternoon. What AI can't do is create something out of nothing. What's been on my mind: follow the data. Teams that have the right underlying data, and can now act on it with the tools at their disposal, are the ones with the real advantage. The ones positioned to build solutions that actually rewire how we solve problems that are long overdue. A few inspiring me lately, all close to home: - AssistIQ , rebuilding how health systems capture supply and implant usage in the OR, closing the documentation gaps that cost hospitals millions in unbilled charges and leave supply chain running on incomplete data. -Pedestal Health, building the infrastructure so evidence compounds across drug development programs instead of rebuilding from scratch each time. The same patient who enters a Phase II study stays in the data through Phase IV and beyond. -Citizen Health, putting patients in control of their own longitudinal health record so they can continuously contribute their data to research and care, without starting over at every new encounter. - insitro , generating proprietary biological data at scale specifically to train ML models that predict drug outcomes and do something about it, rather than fitting AI to whatever data already exists. - Canopy and Regard, helping clinical and operational teams ask the right questions to better predict health events and get billing right the first time. This is the work that makes me believe we can build something that actually matters with AI, when we start from the ground up with the right data. Curious what that looks like in practice? Read this whitepaper from AssistIQ with Periop Leader Network (Formerly OR Manager). Many point solutions have been piled onto the exorbitantly costly problem of bad hospital usage data. This is the one solving it holistically. 📄https://lnkd.in/g8rtwHk8

64-year-old percussion teacher came in with classic sepsis. She looked toxic with high fever, tachycardia, soft BP, dirty urine. The ED and hospitalist team crushed the bundle and the rapid antibiotics really helped to put out the fire. All the sepsis training worked, and she had an amazing outcome. However, what followed was three days of prior auth, two peer-to-peers, and eventually a denial for insufficient documentation of medical necessity. The clinical picture was never in question, but the note just didn't capture everything we knew. The lactate was 3.2 but never referenced in the note. The creatinine went from 1.5 to 3.4, which the note described as "trending up," but acute kidney injury was never diagnosed. The source was a UTI, but the assessment said "sepsis," not "sepsis due to urinary tract infection." The diabetes, the chronic kidney disease, the prior hospitalizations were all buried in the chart, but the admitting team was too busy actually taking care of the patient to dig that deep. And that is where payers make their living. The denial engine is working exactly as designed. It lives downstream, after the note is locked, the chart is closed, and nobody has time to go back. Physician advisors show up to that peer-to-peer fight with an incomplete record of clinical decisions. This was all fixable if the documentation had been just a little more robust. The lactate in the A&P, not just the flowsheet. AKI written as a diagnosis, not an observation. "Sepsis due to urinary tract infection." would be five words that change the claim entirely. Even if we would have brought forward the comorbidities from all those clinic notes that nobody has time to read... AI does all of these easily while no human has the bandwidth to do that for every patient on every shift. She went home. Several of us in her clinical team went out to her open mic night, and her little band took a minute to thank each one of us. The sepsis bundle worked, but the old-school documentation didn't. The hospital sepsis scores won't reflect this save, and the hospital will not get the revenue it deserved. I didn't get into AI to cook the books. I just want my hospital and my colleagues (and all hospitals and all clinicians) to get credit for the amazing work they do each and every day.