How AI works in Contentmunch

AI that waits to be asked.

It answers from what your team wrote, you decide what stays, and it never edits your notes behind your back.

How it behaves
Why it's built this way

AI is only useful in a knowledge base if you can trust it. So it's an assistant you ask, not a layer that rewrites things.

Every AI feature in Contentmunch follows the same few rules. They're below, each with the piece of the product that shows it.

01

Only when you ask

Summaries, quizzes, flashcards, drafts and answers are generated when someone asks for them, then saved with the note so everyone sees the same one. Using a tool counts toward a usage limit. The one thing that runs on its own is the search index, described under Where your text goes.

02

From your notes, not the internet

Ask about the exact passage you're reading. The answer is grounded in that passage and the note around it, so it says what your team wrote, in your terms.

03

You decide what stays

An answer is shared only if you choose Keep, and then it joins the note's Questions & Answers for everyone. Revise proposes edits to a note but changes nothing until you accept each one, and it never touches blocks that are shown to particular readers.

04

It tells you when it's stale

Each result remembers the text it came from. If the note changes afterwards, the result says so, stays readable, and offers to regenerate. Nobody acts on an old summary without knowing it's old.

05

It respects who can see what

AI works from the general version of a note. Text marked for admins or editors isn't part of what summaries and answers are made from, and it is left out of the search index too. Suggested edits follow the same rule: restricted sections are replaced by a marker before anything is sent.

06

Admins hold the controls

Each AI tool can be switched off for a workspace, and when it is, it isn't shown at all. Limits keep use predictable: an hourly limit per person and a daily limit an admin sets.

The tools

Five places AI shows up, each where you'd use it.

On a passage

Ask about this

Select a sentence you don't follow and ask. Keep the answer for everyone, or discard it.

On a note

Summary and mind map

Made on request and stored with the note, marked out of date once the note changes.

On a note

Quizzes and flashcards

Multiple-choice questions and flip cards that belong to one note, not a separate bank.

In the editor

Revise

Describe a change and review the proposed edits one by one. It can be undone and never touches audience blocks.

On Home

Draft a note

Answer a short question and it writes a first draft. It asks up to three follow-ups, only if your answers would change the draft, and saves nothing until it's written.

And that's the list

No hidden extras

If an AI tool isn't on this page, it isn't in the product.

Where your text goes

Exactly when AI sees your notes.

Contentmunch uses outside AI providers for these four things and nothing else.

  • When you use a toolThe passage or note, and your question, are sent to the AI provider your workspace uses, to produce the result. Nothing is sent unless someone asks.
  • When a note is savedIts text, without any part restricted to particular readers, is turned into a numerical index (an embedding) so search can match meaning as well as words. Only the first part of a very long note is used. This runs on its own, because search needs it, and only for organizations with AI search switched on.
  • When you searchYour query is turned into the same kind of index, to find notes with a similar meaning.
  • What is left outText marked for admins, editors or other particular readers is not part of any of the above. This includes suggested edits: restricted sections of the open text are replaced by a marker before anything is sent.
Plain limits

What it won't do, and where it stops.

Being clear about limits is part of being trustworthy.

  • It won't change a note on its own.Edits are proposals until you accept them.
  • It won't share an answer you didn't keep.Discarded answers are gone; kept ones are marked as such.
  • It won't summarise a very short or very long note.A note under about 400 characters is too short; a very large one should be split.
  • It has a rate.Generating is limited per person each hour, with a daily limit set by an admin.

Questions about AI

Can we turn it off?

Yes. An admin can switch off any AI tool for the whole workspace. A tool that's off isn't shown anywhere. The search index described above is separate: it is what search uses to match meaning.

Who sees an answer I keep?

Everyone who can read the note. It's saved under the note's Questions & Answers. An answer you don't keep isn't saved.

What if the note changes after a summary is made?

The summary is marked out of date, stays readable, and offers to regenerate.

AI you can ask, and trust to stay in its lane.

We're onboarding a small number of teams directly. Tell us a little about yours and we'll follow up.