Your first evidence
Upload a handful of call transcripts to a project and follow them through to the first theme Adduce synthesises.
By the end of this tutorial your project holds a transcripts collection, a few uploaded call recordings turned into quoted evidence, and the first themes Adduce clustered out of them. It takes about fifteen minutes, most of which is Adduce working while you watch. Transcripts are the fastest way in because they need no keys from another tool — only files you already have.
Before you begin
- Role: project Lead or Editor on the project you are working in. A workspace Admin holds Lead on every project.
- To hand: two or three call or interview transcripts as
.txt,.srtor.vttfiles. Each file must be under5 MB, and one upload takes at most25files. - Time: fifteen minutes, including the wait while Adduce reads the files.
Part 1 — Create a transcripts collection
Open the project and choose Sources. A project with nothing connected shows the connectors you can add and a single Add source button.
In Add a data source, choose Transcripts under Product signal. Every other connector on this screen wants a key or a token; this one wants files.
Give the collection a Collection name — Customer calls is a fine start — and a Source name saying where the calls came from. Choose Create; the wizard confirms the collection is ready and offers to open it.
Part 2 — Upload the transcripts
Open the collection from Sources and choose Upload transcripts. The accepted extensions are printed beside the button.
Pick your files and confirm. Each one appears in Files with the status Queued, then Processing while Adduce segments the text and reads it, then Extracted with a word count and the number of evidence items it produced. A file Adduce cannot read stops at Failed and stays in the list so you can delete it and try another export.
Part 3 — Watch the evidence arrive
Choose Pipeline in the sidebar. The chart counts evidence written over the last thirty days, stacked by the source it came from. Your new collection appears in it as soon as the first file finishes.
Choose Themes. Once the files are extracted, synthesis clusters the recurring evidence and names each cluster. Open the strongest theme: it carries its strength, its sentiment, how many pieces of evidence sit under it and how many sources they came from, and it quotes the evidence itself.
Why the numbers matter
A theme drawn from one file is a hunch. The source count is how you tell the difference: corroboration across separate sources, or across the same source over time, is what makes a claim worth acting on.
What you built
The project now holds a transcripts collection, one piece of evidence for every point a customer made in those calls, and a ranked list of themes over them. Every quote on a theme links back to the file it came from, so the argument is checkable rather than asserted. Uploading more files later re-runs synthesis, and the themes update in place.