Why do I still spend so long editing AI scribe notes?

Because reading is slower than writing when the words are not yours. You could skim your own note. A draft you did not write has to be checked line by line, and any line could be wrong. The transcription was never the slow part. The checking is, and almost nobody offers to take that back.

Where does the time actually go?

Not where the demonstration says.

Count it honestly on your next patient. The recording costs nothing, it happens while you work. The draft appears in seconds. Then you read it. Then you delete what is not true. Then you rewrite what is true but wrong in emphasis. Then you add the examination you did not say out loud.

In a physicians’ thread this year, the people who quit described the same shape. One said that after six months of configuring, the notes were still bloated. One said that in the time it took to edit the plan he could delete it and write his own. One said five to ten minutes a chart went on removing things nobody in the visit said.

None of them said the tool could not hear them.

Why is a longer note worse?

Because every extra sentence is one more you have to check, and one more a colleague has to read.

A physician in the same thread said colleagues’ notes were now ten times longer than when they typed them, and that he was not going to read a history that size.

In most writing, length reads as effort. In a chart it is the opposite. A long note hides the two lines that matter, and it stretches the review for you and for whoever is covering you at two in the morning.

These tools produce length because length looks like work. You pay for it twice. Once when you check it, once when someone else has to.

Does more configuring fix it?

The thread suggests not, and this is the part worth arguing with.

The clinicians who put the most effort into setup did not report the best results. One spent six months and still called the notes bloated. The fastest note in the room, about two minutes, came from someone doing the opposite. He stopped trying to make the tool sound like him and asked only that it hold the facts.

Configuration is an attempt to make a tool produce the note you would have written. That is a very high bar, and you are the one marking the homework.

What would make a draft quick to check?

Ask that instead of asking for a better draft.

A draft is quick to check when you know exactly where to look. That means it is short. It means it does not invent. It means the parts with consequence, the numbers and the plan, are either yours or clearly marked as not yet confirmed.

One surgeon reported a tool getting a lesion percentage wrong. That is a small error with a long tail. Once you have seen one, you read every line of every note forever, and the tool has cost you more than it saved.

Trust is not a setting. It is earned by a tool that does less and is right about it.

What can you change this week?

Stop asking for a finished note.

Let the tool write the history. Write your own assessment and plan, out loud, right after the patient leaves. Time five notes and compare.

If the number does not move, the problem is the tool, and you now have a specific reason rather than a feeling.

The short version

The editing is the cost, and it comes from reading words you did not write. A short draft you can check beats a long one that reads well. Solanto is being built for the short draft: it will write the history, leave the judgement to you, 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

Send this to the colleague still charting at seven. Ask how many of those minutes go on typing and how many on checking.