Should an AI scribe write my assessment and plan?

No. The clinicians who cut their note time the most gave the scribe the history and kept the assessment and plan. They stopped asking it to sound like them and asked it only to hold the facts. One said his note now takes about two minutes. The ones who spent months tuning the tool still rewrote the note at the end of the day.

That is the whole finding, and it is the opposite of how these tools are sold.

Why does editing take longer than writing?

Because reading is slower than writing when the words are not yours.

A note you wrote can be skimmed. You know what you meant, so your eye only has to confirm it. A draft you did not write has to be read line by line, because any line could be wrong and you are the one signing it.

One surgeon in a physicians’ thread described reviewing his drafts as being an English teacher. He could proof his own notes quickly, because he wrote them. The tool handed him a draft and took the time back in checking.

So the real cost of a scribe is not the subscription. It is how much of the note you still have to read.

What should the scribe actually write?

The history. The part where the work is recall, not judgement.

The history is what you were told, in the order you were told it. You cannot skip recording it and you gain nothing by typing it yourself. If you forget a medication timing from four hours ago, the recording does not.

The assessment and plan is different. It is the part that is you: what you think is going on, what you are going to do, and why. It is also the part a colleague actually reads.

Clinicians in that thread arrived at the same split independently. One psychiatrist writes his own assessment and plan and lets the tool keep the history. Another said the tool captures the facts, so he stopped fussing over the history and spends his effort where it counts.

What happens when you stop caring how the note sounds?

Your editing time falls.

The most useful comment in the thread came from a psychiatrist who made a decision. Did he care that the note sounded good, or that it held the right facts? Once he let go of the first, his editing collapsed. He put the note at about two minutes.

That is worth sitting with. What was costing him hours was not accuracy. It was style. He was correcting a draft that was already true, because it did not sound like him.

Nobody reads your history for prose. They read it for what happened.

What do you lose by keeping the plan?

Less than you think, and you keep the part that matters.

Dictating your own plan takes a minute or two, because you already made it. You made it in the room. The typing was never the hard part.

What you get back is a note that reads like a clinician wrote it, because one did. You also stop proofreading the section with the most consequence. A wrong number in a plan is a different kind of wrong from a clumsy sentence in a history.

How do you try this tomorrow?

Three things, and no new tool.

First, at the end of the next visit, let whatever you use write only the history. Delete its plan before you read it, so you are not tempted to edit it.

Second, dictate your plan in your own words, straight after the patient leaves, while it is still in your head.

Third, note the time. Compare it with your usual. Do it for five patients before you judge it.

If your tool cannot be told to stop writing the plan, that is a fact about the tool, not about you.

The short version

No. An AI scribe should write the history and leave the assessment and plan to you. You can force most tools into that shape by hand. Solanto is being built to work that way by default: it will write the history, keep the judgement yours, and run in Cloud mode or in Private Mode, where the note never leaves your device.

Solanto has not shipped yet. The price is on the page rather than behind a call, and early access is open: solanto.ai/pricing

Pass this to the colleague who told you scribes do not work. Ask them which half they gave away.