USE CASE

Turn earnings call notes to report, then check every number

An earnings call produces more notes than anyone reads twice. The useful output is a short report where every figure and every quote can be traced to the transcript. That is a structuring job before it is a writing job.

Capture the raw material

Keep the transcript, your live notes, and the presentation deck in one place. Date each one and note which version of the transcript you used; transcripts get corrected.

Label the call's shape early: prepared remarks, guidance, Q and A, and closing. That structure is what the report will mirror, so getting it right saves the rewrite.

Tag before you write

Tag four things while the call is fresh. Guidance and changes to it. Hard figures with their units. Quotes that show how management thinks. Open questions that nobody asked.

Tags are more useful than a summary because they are checkable. A summary tells you what someone thought the call meant; a tag tells you where to look.

Draft from tags, not from impressions

Write the report from the tags. Each paragraph should answer one question an investor would ask: what changed, what management expects, what the numbers show, and what is still unknown.

Put the strongest counter-evidence in the report, not in your private notes. A report that only supports the thesis is not useful to anyone who has to make a decision.

Check every number

Run a simple check pass: every figure in the report appears in the transcript, every percentage has a base, and every quote matches the wording. Numbers copied by hand are where most errors enter.

Ask AI to list the figures it used and where each one came from. Reviewing a list is faster than re-reading the report.

Where this breaks

Reports break when the transcript is mistranscribed, when guidance is quoted out of context, and when a metric is compared across periods with different definitions. Each one needs a check, not a better prompt.

If the call had a live translation, keep the original transcript next to the translated one. Quote from the original when the wording matters.

How to run it on Sigla

Keep the transcript, the tags, and the draft in one workspace. Ask AI to tag the call, then draft section by section from the tags. The built-in browser is for pages that need a sign-in, such as a corrected transcript.

For the wider research process, read the AI investment research workflow. If you are writing for a fund audience, the AI research workflow for VC analysts shows how the memo and the call notes connect.

FAQ

Can AI transcribe the call?

It can work from a transcript you provide. Check the provider's transcript against the recording before you quote anything; names and numbers are where transcripts fail.

How do I stop the report from overstating the call?

Keep the open questions and the counter-evidence in the report. A draft that only supports one reading is a thesis, not a report.

What should every figure include?

The value, the unit, the period, and the transcript location. A number without a period is not a comparison.

Next step

Take one recent call, tag it before writing anything, and draft only the paragraph that covers guidance.

SIGLA DESKTOP

Download Sigla and run this workflow on your desktop.

Windows · macOS (Apple silicon and Intel). Download first, then start from a request or a template.