Blog: Scale Is Earned, Not Scheduled

At this point, it’s likely that you’ve run an AI pilot that worked, but it may be less likely that what you discovered in the pilot has become standard protocol across your sites. 

When Dr. Cara Bishop, vice president of clinical services at Banyan, is an example of a pilot made practice. They had early success in their pilot. Their 25-clinician documentation pilot across two sites in June 2025 showed that time on notes dropped 30 to 42%. Today, 73 clinicians, nurses, group facilitators, and medical staff across seven sites use the same tools as standard practice, time savings run 50 to 65%, and first-pass compliance has held at near perfect since October 2025. In a recent Kipu webinar, she described the principle that got Banyan there as “scale is earned, not scheduled.” 

In practice, that means expansion follows validated results rather than a date on a project plan, so that growth compounds what you’ve already measured instead of multiplying the problems you already have. Here is how Banyan earned it. 

Before AI, a biopsychosocial assessment took a seasoned Banyan clinician 78 minutes to complete, and once supervisory review and corrections were added, a single document consumed more than two hours. First-pass compliance sat around 93%, which sounds respectable until you price the remaining 7% in claim denials, audit exposure, and supervisor hours spent on correction instead of clinical development. 

Cara’s read was that Banyan had a clinical problem masquerading as an administrative one, because the system around clinicians was never built to support the standard they were held to. That diagnosis is why clinical leadership owned the project from day one, building around clinical standards and testing with clinicians, with IT as a partner rather than the lead. Banyan went straight at its most complex, highest-stakes document, and Cara’s advice to leaders who feel stalled is to do the same. She says, “find your biggest, most demanding piece of documentation and work it all the way through.” A pilot on something safe proves a small point, and small points rarely move an executive team. 

Banyan’s build ran in three phases, and none of them changed a single form or documentation standard. The AI was shaped to fit workflows clinicians already trusted, because, as Cara puts it, the moment you bend clinical standards to accommodate a tool, you’ve compromised the one thing AI is supposed to protect. 

Phase one was iterative prompt development with the Kipu Intelligence team, with nothing deployed until it held up to clinical scrutiny. Phase two introduced weighted rules, since not every element of a note carries equal regulatory weight, aligned to ASAM Fourth Edition, CARF, Joint Commission requirements, and Banyan’s own model of clinical excellence, resulting in a document that’s defensible under payer or auditor review. 

Phase three was the operational game changer. Compliance review moved from weeks after the fact to the point of documentation, so that a clinician now sees a compliance score against all four sources before signing, and a supervisor runs the same audit again before approving. The cycle of corrections and revisions that once consumed supervisory time was nearly eliminated, while every AI-supported document still receives clinical review, with clinicians making the final call. 

Before expanding, Banyan validated its results across 22 internal pilot studies, cross-checking clinician-reported time against EMR timestamps. As the prompts and weighted logic matured, the pilot’s 30 to 42% savings grew to a consistent 50 to 65%, biopsychosocial documentation fell from 78 minutes to 39, and first-pass compliance moved from 93% to near perfect within Banyan’s audit parameters. 

The clearest sign that the pilot had become practice was a change in what clinical directors were asking. Where they had once asked whether to use the tools, they began asking how to use them better, and supervisors started building the note audit into their standard second-signature workflow on their own because it made their work faster and more defensible. 

The result Cara talks about most belongs to new hires, who earn AI access only after completing an unassisted baseline evaluation. Once they have it, they see their own clinical reasoning reflected back in precise, compliant language the first time, and the learning curve to reach Banyan’s documentation standard has dropped 61%. Supervisors who once spent their time on corrections are now having the clinical conversations that supervision is meant for. 

Banyan hasn’t declared the work finished. Chart Assistant for utilization review is on its current frontier, and in its first three days the UR prep team secured 29 days of residential treatment across three cases it had expected to be denied. As Cara reminds leaders whose efforts have stalled, a rushed expansion gives you a bigger version of whatever problem you started with, while an earned one gives you a bigger version of the result. 

Watch the full conversation with Dr. Cara Bishop on demand, and download Banyan’s white paper on building a clinician-led AI documentation program:

Request a personalized consult below.

About the Authors

Travis Moon Travis Moon, Kipu's Content Marketing Strategist, is a seasoned leader in healthcare IT and content marketing, specializing in the behavioral healthcare sector. He develops impactful, data-driven campaigns that support healthcare professionals and enhance patient outcomes. With over a decade of experience, Travis has led strategic content initiatives for major healthcare organizations, including the launch of data visualization tools and thought leadership campaigns.