Becker's Health IT: Making AI actually perform in acute care settings

The following is an excerpt from a feature with Justin Mardjuki, CEO of Sayvant, in Becker's Health IT. You can find the full post on the Healthcare Dive website here.

‍

Ambient AI is exploding in popularity because charting overhead is growing and clinicians want time back. However, acute care adoption significantly lags ambulatory settings: KLAS estimates in their Ambient Speech 2025 report that 95% of adoption is in ambulatory care, with only 5% in acute care.

Why is this happening?

• More complex encounter workflows: Triage, initial assessments, test and order results, re-examinations, consultations and more complex problem lists demand multi-step encounters over several hours, with real-time changes in decisionmaking. PIT, split-flow/fast track, consult loops, re-evaluations during boarding, and critical patients that are crashing are all part of the job.
• Intricate documentation requirements: Acute care notes are not a simple reformatting of an ambiently captured conversation. SEP-1, stroke door-to-CT, critical care capture, MIPS measures are just a handful of the thousands of criteria defined by Quality, Risk, and Revenue Cycle Management teams:.
• Need for site-level and clinician-level customization: No two sites are the same – local stroke triggers and APP cosign rules differ hospital to hospital.

Acute clinical leadership and CMIOs need solutions that can model actual Emergency Medicine and Hospitalist Medicine workflows, absorb local guidelines and requirements, and integrate seamlessly into their EMR systems. Speech-to-text transcriptions alone are insufficient: you need clinical reasoning in workflow, and a note that stands up to Quality/RCM/CDI requirements.

‍