ResearchWorkflow

Earnings Prep

Last updated April 27, 2026

What it does

Before a name reports, point Cowork at the prior four transcripts, the most recent 10-Q, and your existing notes folder. Cowork pulls what management said last quarter on the topics you care about, compares it to current consensus, and flags what changed in the language. You get a one-page brief instead of four hours of reading.

What you need first

·A folder containing the last 4 earnings call transcripts (PDF or text)
·The most recent 10-Q in the same folder
·(Optional) Your prior internal notes on the name
·FactSet or AlphaSense MCP connected if you want consensus pulled live

The prompt

Copy this, swap in the ticker and the topics you actually care about for this name:

I'm preparing for {TICKER}'s upcoming earnings call.

In this folder you have the last four earnings transcripts and the most recent 10-Q. Do the following:

1. Pull what management said about {TOPIC 1}, {TOPIC 2}, and {TOPIC 3} in each of the last four calls. Quote the exact language.

2. Track how the language has evolved across the four calls. Flag any changes in tone, hedging, or specificity. Be especially attentive to phrases that softened or sharpened.

3. Cross-reference against the most recent 10-Q. Note any disclosure that adds detail to or contradicts what was said on the calls.

4. Pull the current consensus estimates for next quarter (revenue, EPS, and any segment-level numbers I should care about) from FactSet.

5. Produce a one-page markdown brief titled "{TICKER}-earnings-prep-{date}.md" in this folder. Structure: (a) Three things to listen for, (b) Language drift since last call with quotes, (c) Where consensus could be wrong, (d) My suggested questions for the call.

Quote everything inline. Don't paraphrase numbers — copy them.

What good output looks like

✓A real markdown file in your folder, not a chat response
✓Direct quotes from each transcript — verifiable, not paraphrased
✓Specific language drift called out (e.g., “modest pressure” → “continued headwinds”)
✓Consensus numbers with timestamps and the FactSet field name they came from
✓A short list of questions you'd actually ask if you were on the call

Cowork running the earnings prep prompt against the transcript folder — Step 1 — Cowork reads the transcripts and the 10-Q in your working folder.

Cowork pulling consensus estimates from FactSet via MCP connector — Step 2 — Consensus is pulled live from FactSet through the MCP connector, with the source field named.

Cowork synthesizing the brief and writing it to the working folder as a markdown file — Step 3 — The brief is written as a real markdown file in your folder, not a chat response.

The finished one-page earnings prep brief, rendered as a markdown file — The finished brief — three things to listen for, language drift with quotes, suggested questions for the call.

Make it a Skill

Once the prompt is producing the brief you want, save it as a Skill named earnings-prep. Next quarter, you point Cowork at the new transcript folder, type /earnings-prep AAPL, and the same brief structure runs against the new data. Schedule it to fire two days before each calendar earnings date and the brief is in your folder before you ask.

Common pitfalls

Don't skip the “quote inline” instruction. Without it, Claude will paraphrase, and you lose the audit trail.
Be specific about topics. “Margins” is too broad. “Gross margin trajectory and the factors management cited” is right.
Cross-check the consensus pull. First few times, verify the numbers match what FactSet shows in the terminal. Once you trust the connector, drop the check.