Earnings Call Digest

name: earnings-call-digest
description: Digests earnings call transcripts into what changed, management tone and Q&A highlights, optionally read through the lens of the user's position or thesis. Use when the user shares an earnings transcript or asks for an earnings call summary or recap.

Turn an earnings call transcript into the digest a buy-side analyst actually needs: not a summary of what was said, but an analysis of what changed, what management emphasized or dodged and what it means for the position. An analyst covering 40 names cannot read every transcript. This digest has to be trustworthy enough to trade around.

Inputs

• The transcript (pasted or uploaded). Required.
• Optional but high-value: the prior quarter's transcript, guidance or the analyst's notes. This is what makes "what changed" rigorous rather than inferred.
• Optional: position context. Long, short or watchlist, plus a one-line thesis. If provided, the digest is read through that lens.

Workflow

1. Read the full transcript before summarizing anything. The most important sentence is often a half-answer in the final Q&A exchange.
2. Build the digest in the format below, in order.
3. Quote management exactly, with speaker names and their role. Never present a paraphrase as a quote.
4. Maintain three visibly distinct registers throughout: stated facts (numbers, guidance), management framing (characterizations, adjectives, explanations) and your inference. Label inferences as such.

Output format

Headline: two or three sentences: the quarter and the market-relevant takeaway. Written the way an analyst would say it to their PM in the hallway, not like a press release.

Guidance and outlook delta: a table if prior-period material was provided:

If no prior material was provided, list only what management themselves described as changed and say the comparison is limited to management's own framing.

Results vs. Expectations: reported figures for the headline metrics as stated on the call and any metric management newly introduced, stopped reporting or redefined this quarter. A metric that quietly disappears is a finding. Flag it explicitly.

Management tone: evidence-based, not vibes: new phrases versus standard language, hedging words attached to previously confident statements ("we continue to expect" becoming "we currently anticipate"), topics volunteered versus topics only addressed when asked and any change in who answers what (a CEO suddenly deferring detail questions to the CFO is worth noting). Write each tone observation as evidence plus a labeled inference:

"On the full-year guide, the CFO moved from 'we are confident in the range' (prior call) to 'we currently anticipate landing within the range' (this call). Inference: conviction in the guide has weakened. Watch for further softening at the mid-quarter conference."

Q&A highlights: the three to five most substantive exchanges. For each: the analyst and firm, the real question underneath the polite phrasing, the answer with a direct quote and a judgment of whether the question was actually answered. Note any question that was ducked twice. That is where next quarter's surprise usually lives.

Example entry: "Chen (Marbrook): asked twice for gross-margin trajectory excluding the new-facility ramp. CFO answered with total-margin guidance both times. 'We see consolidated gross margin in the 41–42% range' and did not provide the decomposition. Not answered; second consecutive quarter this breakout was declined."

Thesis read (only if position context was given). For each element of the stated thesis: supports, challenges or neutral, with the specific evidence from this call. End with the one or two questions this call raises that the analyst should run down before the next print.

Watch items: dates, metrics, catalysts and follow-ups mentioned on the call, each with its timing.

Guardrails

• Do not import outside knowledge about the company, consensus estimates or the stock's reaction as if they were on the call. If the user asks for that context, add it in a clearly separated, labeled section.
• Do not give buy/sell recommendations. The thesis read describes evidence; the trade decision is the analyst's.
• If the transcript looks truncated, machine-garbled or missing the Q&A, say so first. A digest built on a partial transcript needs a warning label.