One transcript tells you what one person knows. A handbook turns what several colleagues know into something the next person can follow.
Executive summaries
Every completed response gets an executive summary, written automatically as the respondent finishes and shown above their transcript: one short explanation of what matters, followed by at most three actions, decisions, or risks worth remembering.
A summary is a half-minute read, whatever the interview's length. It orients you; the transcript underneath is the record.
It is written in whatever language the interview's Results in setting names, so a French transcript can carry an English summary, and the page says so when the two differ. Quotes stay in the respondent's own words. Use Rewrite if the summary missed something you can see in the transcript.
Nuggets
As each colleague finishes, the AI also pulls out the reusable knowledge in what they said. Each piece becomes a nugget: one self-contained thing someone could act on.
Every nugget has a kind:
- How-to — ordered steps, the procedure as it is actually performed.
- Judgment call — when X rather than Y: thresholds, escalation rules, the "it depends" made explicit.
- Gotcha — what breaks, and the warning signs that it is about to.
- Why — the reason behind a rule. These stop the next person "improving" a process they do not understand.
- Who / where — the person to ask, the system to look in, the document that already exists.
And a status, which is the part to pay attention to:
- Confirmed — several colleagues independently described this the same way.
- Single source — one person said it and nobody has confirmed it. Still useful; just know that it is one account.
- Conflict — colleagues described the same thing differently. Both sides are shown, named. Somebody has to decide.
- Gap — it was asked, and nobody could answer. Often the most useful finding: a list of what nobody can currently explain.
Nuggets accumulate. Each new response is compared against what is already known rather than starting over, which is how a nugget goes from single source to confirmed.
Where to find them
Each interview has a Nuggets tab listing everything that interview captured, searchable and filterable by status.
They also appear on each transcript, under Knowledge captured here — the nuggets that conversation is the source of, so you can see what one person added without leaving their words.
Handbooks
The Handbook tab writes what has been captured up as a document someone can follow. It opens with a short title naming the job and a line saying when to reach for it, then sections, numbered steps, and the judgement calls and warnings around them.
It does not say who said what. The sources sit behind the small numbers on steps and notes: click one to see the nugget it came from, the exact words a colleague used, and a link into their transcript.
A handbook is written in whatever language the interview's Results in setting names, its own labels with it — on the page, on the shared link, and in the Markdown download.
Only confirmed and single-source knowledge is written into a handbook. Something two colleagues describe differently is not settled enough to instruct anyone with, and something nobody could answer has nothing to instruct. Those are listed at the end under Still unresolved, so you can see what is outstanding without it being mixed into the steps.
A handbook says when it has fallen behind, both on its own page and in the interview that produced it. Rewrite is offered only when something has changed — new nuggets captured, existing ones edited, or one that stopped being settled — and it tells you which, and how many. When nothing has changed, the button is off.
An interview has one handbook. Rewriting replaces it rather than adding a version, and it keeps the same share link, so a colleague you already sent it to sees the update. The date says Updated once it has been rewritten.
Every handbook your team has written is listed under Handbooks in the sidebar, whichever interview produced it. Each entry shows what it is and when to reach for it.
Fixing what a handbook says
A handbook is written prose, so editing a nugget does not rewrite a handbook that already exists. Fix the nugget on the interview's Nuggets tab, then choose Rewrite handbook to pick the change up.
Sharing a handbook
A handbook is private until you say otherwise. Choose Create share link and anyone holding that link can read it — no sign-in, no account. They see the handbook and nothing more: not the interview, not the transcripts, not your team. The link looks like app.nuggetz.ai/h/9f3a….
Pasted into Slack, it previews itself. The card shows the handbook's own title and the line saying when it applies, so the colleague receiving it can tell what they have been sent before they open it.
What is unresolved is shown to them too, exactly as you see it — a handbook that hid its own uncertainty would be read as settled. What they cannot do is open your transcripts: a quote shows who said it, but only your team can follow it back to the conversation.
Stop sharing revokes the link immediately. A revoked link and a wrong one look identical to whoever tries them.
Taking it elsewhere
Markdown downloads the handbook as a .md file, ready to paste into Notion, Confluence, a wiki or a repo. Headings, numbered steps and warnings travel with it, the sources are listed at the end so a [3] in the pasted copy still leads somewhere, and anything unresolved stays labelled as unresolved.
What happens to deleted interviews
Deleting an interview deletes everything that came out of it: its transcripts, the nuggets captured from them, and the handbook written from those.
The confirmation says what will go before it happens — how many transcripts and nuggets, and the handbook by name. If the handbook has a live share link, it says that too: anyone you sent it to loses access, and the page they reach cannot tell them why. You can download the handbook as Markdown from that dialog before confirming.
Deleting a handbook on its own is different: the nuggets stay, and you can write it again.
When AI is unavailable
Summaries, nuggets and handbooks all need an AI model configured for your Nuggetz instance. Without one, transcripts are still complete and readable — you just read them yourself.