How do I test an AI scribe before I pay for one?

Run the same recorded visit through every tool you are considering, then count the lines you rewrote. Not how the note reads at first glance. The number of lines. A psychiatric nurse practitioner gave that advice to colleagues comparing scribes, and it is the best trial method in any room we have read. Add one more number: the monthly price, expressed in patient visits.

Why one recorded visit?

Because it is the only fair test. Every tool demos well on its own sample. Another clinician in the same thread suggested finding a recorded psychiatric evaluation online and running it through each tool in turn. Same words in, different notes out. Now you are comparing tools, not visits.

If you do not want to use a real patient's audio, and you should not, record yourself role-playing a visit with a colleague, or use a published teaching video. What matters is that every tool hears the same thing. And say it out loud: if you gesture through the exam, no tool will catch it.

Why count lines instead of reading the note?

Because a note that reads well can still cost you twenty minutes. The first read tells you whether the tool writes nicely. The count tells you what you will pay every visit, in your own time. That is the whole argument in why the editing is the real cost.

Keep the count honest. A line you deleted counts. A line you rewrote counts. A line you added because the tool left it out counts double, because you had to notice it was missing.

What about the price?

One nurse practitioner priced her scribe in a unit every owner understands: about one patient visit a month. Another said a tool she liked had simply become too expensive and she was moving. Put every price in visits and the comparison writes itself. A tool that costs one visit and saves ten minutes a day is a different decision from one that costs three.

Then read the contract before the trial ends. the five contract questions are the ones people skip.

What should I ignore?

Feature lists, especially the ones that promise to suggest your billing level or update a diagnosis for you. One tool in that thread does both. A code you did not choose is still your code, and a diagnosis the software proposed is still your diagnosis. Judge the tool on the note, and keep the judgment yours.

How long should the trial take?

An afternoon. One recording, three tools, a tally. Add a week of real visits for the winner before you sign anything longer than a month. If you want a shorter version, the ten-minute check is built for exactly this.

The short version

Run one recorded visit through each tool, count the lines you rewrote, and price each one in patient visits. Solanto will put its price on the page in that unit, and it will write the history and leave the judgment to you, in Cloud mode or in Private Mode, where the note never leaves your device.

The price is on the page, not behind a call. Early access is open: solanto.ai/pricing

Send this to the colleague who is about to pick a scribe from a Reddit thread. Ask them how many lines they rewrote.