Synopsis: Keebler Health CEO Isaac Park dives into the impact artificial intelligence (AI) is having across the healthcare sector.
AI agents are popping up in every demo reel, but Isaac Park, CEO of Keebler Health, says hospitals are still working out the kinks. Clinicians see the potential—less paperwork, faster insights—but the workflows are far messier than a slick slide deck suggests.
The first hurdle is mixing probabilistic tools with deterministic tasks. Large‑language models make “best guesses,” yet billing codes and lab values have to be right every time. Treating one system like a Swiss‑army knife inevitably disappoints; sometimes a simple rules engine beats an eloquent prompt.
Park also pushes back on end‑of‑career doom‑scrolling. Past tech shifts killed some roles but created more over time; AI will likely reshape jobs rather than erase them outright.
C‑suites are split. Some executives won’t budget a dollar until they see hard proof; others are already counting the head‑count they believe agents will replace. Reality, Park says, probably sits somewhere between those extremes.
Nobody, however, is casual about patient data. Security and HIPAA compliance dominate every sales call, and any vendor that can’t safeguard clinical documents is quickly shown the door.
Where agents shine today is in taming unstructured text. Summarizing medical records, flagging chronic‑disease evidence, or turning live dictation into structured notes already saves doctors hours each week—and hasn’t triggered a malpractice crisis.
Looking ahead, Park expects an ecosystem of specialized agents orchestrated by a control layer rather than one monolithic model. Healthcare will adopt that pattern slowly, he warns, because regulated industries demand rock‑solid guardrails before anything touches production.
Finally, there’s the spreadsheet nobody wants to open: GPU costs, token fees and energy bills. Until hospital finance teams can map clear ROI, many pilot projects will stay in the sandbox.
In short, AI agents are coming to medicine—but only after the industry figures out where probabilities help and where certainty still reigns.
