Dhruv Vasishtha pushes exception management
- Dhruv Vasishtha, a 25Madison health adviser and former firsthand product chief, argued healthcare AI should target operational exceptions instead of full automation. - He pointed to care gaps like missing labs, unreconciled medications, and off-pathway patients, where software can surface deviations for staff review. - The pitch lands as hospitals struggle to scale AI beyond pilots and measurable returns remain rare. (qventus.com)
Healthcare AI works best when it watches for what went wrong, not when it tries to run every step itself, Dhruv Vasishtha said in a recent post and interviews. (x.com) (elion.health) Vasishtha is an operating partner at 25Madison and a former senior vice president of product at firsthand, a value-based care company focused on people with serious mental illness. At firsthand, he said his teams built around decision support, risk stratification, and data workflows rather than replacing care teams outright. (25madison.com) (businesswire.com) The core idea is exception management: software follows a standard care path, then flags the cases that fall off it. Vasishtha’s examples included patients with missing lab work, unreconciled medication lists, or other gaps that need a human to step in. (x.com) That is different from the broader promise that artificial intelligence will automate an entire clinical or administrative workflow from end to end. In healthcare, every handoff touches billing rules, safety checks, and patient histories that change from case to case. (sciencedirect.com) (forbes.com) Hospitals have been looking for narrower uses that fit into existing operations. In a November 21, 2024 interview with Elion, Vasishtha said health systems under margin pressure were becoming more willing to buy tools for transitions of care, throughput, and labor-intensive workflows. (elion.health) The timing lines up with a wider industry problem: lots of pilots, not much scale. Qventus said in an April 2026 report based on more than 60 health system technology leaders that 42% were deploying AI across multiple use cases, but only 4% had reached scaled implementation with measurable outcomes. (qventus.com) (beckershospitalreview.com) Exception management fits that reality because it narrows the job. Instead of asking a model to schedule, document, triage, and close the loop on every patient, it asks software to identify the few charts, referrals, or appointments that need attention first. (forbes.com) (314e.com) That logic maps closely to behavioral health, where intake and scheduling often break on edge cases. Vendors serving that market describe higher no-show rates, incomplete records, and insurance or acuity issues that require escalation instead of a fully automated script. (prettygoodai.com) (caliberfocus.com) Vasishtha’s argument is less about replacing staff than deciding where staff time is most useful. In healthcare operations, the clean cases can move on rails; the messy ones are where the labor, risk, and cost still sit. (x.com) (elion.health)
