I used a general-purpose AI scribe for the better part of a year before switching to Nextvisit. The general scribe was fine. It saved me time. Three to four hours a day of charting became one to two. That was real, and I want to give the prior product credit. What I want to write about is what I did not realize I was missing until I switched, because that is the part I would have wanted someone to tell me before I made the change.
This is the first week. Not the polished case study version. The actual week, with the surprises and the small adjustments.
The day before, briefly
For context: I am a PA-C in outpatient psychiatry. My panel is roughly 60 percent medication management, 30 percent intake, and 10 percent therapy-medication combination visits. I share a workspace with three other clinicians and a billing coordinator. We use Osmind as our EHR. Volume is around 18 to 22 patients a day, with one intake per day on most days and two on some.
The general scribe I had been using produced a clean four-paragraph SOAP note. I edited every note. The edits were not large; mostly I rewrote the assessment and plan in the format I was trained to use, and I added the MSE language the model never quite got right. I told myself this was fine.
Day one
The Nextvisit setup took about 25 minutes. Most of that was the EHR connection. I picked Provider+. The first patient was a 45-year-old established on an SNRI for recurrent depression with a partial response, here for a check-in.
What I noticed first was that the draft note had a mental status exam in the MSE field, not paraphrased into the HPI. That is a small thing. It changed my afternoon. I did not have to rewrite the MSE. I had to read it, correct one or two phrases, and move on. Not “edit.” Read.
The diagnosis suggestion was F33.1 with the partial-response specifier reflected in the assessment text. The general scribe had been suggesting F32.9 for this patient for months. I had been overriding it and, honestly, sometimes forgetting to override it. The first F33.1 suggestion was a small wake-up about how much I had been compensating without realizing.
I finished four notes during the day, between visits, with a few minutes to spare. The remaining notes I closed before I left the office. No 9 PM laptop session. First time in a while.
Day two
Day two was an intake. Sixty minutes. New patient with a complicated history: childhood trauma, two prior medication trials with side effects, a family history that mattered, and a current substance use pattern that needed careful documentation.
The general scribe, for intakes like this, used to produce a tightly written summary that was a paragraph too short and missed structure. I had been spending 20 to 30 minutes restructuring intake notes after the visit.
The Nextvisit intake draft had every section. HPI with onset, course, modifying factors. Past psychiatric history with the prior trials and responses. Family history. Social history. Substance use as its own section, with the pattern documented in the language the patient used. The MSE was structured. The risk assessment was structured, with risk factors, protective factors, and the clinical judgment in the format I was trained to write.
I edited for clarity. I did not rebuild structure. The note was 90 percent done when I sat down to it.
Day three through five
Days three to five were where the personalization started to show up. The system was learning, in a non-PHI way, the phrasing I prefer. By Friday, the drafts read more like notes I would have written myself. I edit-corrected fewer phrases each day. Not zero. But a meaningful drop.
I also started to use the Treatment Pulse view on three of my long-term patients. This was not in my workflow before. I had never had a tool that summarized a patient’s symptom progression and treatment response across visits without me reading every note. The first three I looked at were useful. One showed a clear gradual worsening on a stable regimen that I had been chalking up to seasonal pattern; looking at the trajectory, it was not. I changed the regimen the following week.
What was unexpected
Three things.
First, I underestimated how much energy I spent on structural restructuring with the general scribe. The “saved time” I had been measuring was the savings from not typing. The cost I had not been measuring was the cognitive lift of restructuring every note. Removing that cost made my day quieter, not just shorter.
Second, the AI Timeline surfaced patterns across visits that I would have caught only by re-reading old notes, which I rarely had time to do. Several of those patterns were clinically actionable.
Third, the diagnosis codes were correct on first suggestion at a higher rate than I expected, especially for established patients. This started to matter at the billing-coordinator level too. Our denials on F-codes dropped in the first 30 days. I did not measure this carefully; the billing coordinator did, and she mentioned it before I would have noticed.
What I would tell someone considering the switch
If you are using a general scribe and it is “fine,” it is worth asking how much of your editing is structural restructuring versus content correction. If it is mostly structural, a behavioral-health-specific scribe will save you energy you did not know you were spending. If it is mostly content, the gain is smaller, and you can probably get there with template work in your current tool.
For me, the switch was the right call. The chart looks like a psychiatric chart on the first draft. The day is shorter and quieter. The 9 PM session is gone.