Most enterprise AI rollouts begin with a pilot, a dozen volunteers, six months of evaluation, and a report recommending further evaluation.
Ryan Weyman, Branch Chief of Information Technology at Santa Barbara County’s Behavioral Wellness Department skipped that part.
He went live with an ambient AI scribe across every clinician at once. The tool is Eleos, purpose-built for behavioral health and wired directly into the department’s electronic health records. “We did kind of a big bang approach,” he says. “We went live with the entire department.”
Six weeks after the initial roll-out, AI has drafted nearly 4,000 notes. The clinician talks with the client, the software listens and drafts the progress note, and the clinician reviews, edits, and signs. In a dashboard that tracks progress, the department discovered that staff saved almost 700 hours in documentation time.
Average out the salaries, and those hours are worth roughly $50,000. The Eleos subscription runs about $16,000. That represents a three-to-one return in the first month and a half and Weyman expects that number to increase as adoption spreads.
And the ROI is the least interesting number in this story.
One of the department’s physicians told BWell leadership that having a scribe like this is the best thing that has ever happened to him professionally. Not the best software. The best thing.
At an ice cream social in Santa Maria, a staff member mentioned that the previous Friday was the first time in her entire tenure with the department that she had gone home at the end of a week without a pile of documentation she could not catch up on. She also mentioned it was the first time she had exceeded her productivity target since the department instituted those targets a year earlier.
Those two facts belong together because so-called “documentation debt” can act as an invisible bottleneck on clinical capacity. The note is not the work, but the note gates the work, and unfinished notes follow you home, ruin the weekend, and greet you Monday alongside a supervisor asking why they are late. Remove that and productivity improves without anyone being asked to work harder. This particular staff member did not get faster. She got unblocked.
Every client is asked for consent before the scribe is used, and training included teaching the clinician how to raise the question to the client: we are going to use an AI scribe, it is a bit like having someone else in the room, is that all right with you? Not one client has said no.
Phase two is where things gets even more interesting.
BWell has a Quality Care Management team that reviews charts by hand for regulatory compliance. There is not enough time in the day to review all of them, so they sample. Sampling means some errors are found late or never. The next module, SmartComply, checks seven regulatory requirements against the content of the note and the type of treatment delivered, at the moment the note is written.
Instead of someone circling back weeks later to tell you what you missed, the requirement is satisfied while you are still in the document and still remember the session. The compliance team stops playing gotcha and starts doing the work that actually raises quality: education and training. “It takes the guesswork out of it,” Weyman says.
So, what made the big bang work?
Rather than saying “let’s see where AI might help”, the target use case was a clear pain point. There was one universally loathed task and one tool aimed squarely at it with all involved staff participating at once. Clear benchmarks were set up before go-live so the results would be clear.
Despite the program’s early success, Weyman is not declaring victory. “Still a long way to go,” he says. Adoption is uneven, phase two has not shipped, and the additional clients those 700 hours could serve have not been reached yet.
But the direction is clear. AI did not replace a clinician, or make a clinical judgment, or talk to a client. It took away the paperwork that was standing between good people and the work they are trained to do.
And somebody got her Friday night back.
