Quarterly earnings calls compress weeks of financial analysis, narrative drafting, and executive alignment into a few high-stakes days. Controllers and FP&A teams typically rebuild the same preparation scaffolding every quarter: reconciling actuals against guidance, drafting MD&A language, anticipating analyst questions, and rehearsing answers with the CFO. The work is repetitive but unforgiving — a single inconsistent figure or an unprepared question can move the stock price and damage credibility with the Street.
The friction is rarely analytical. It is coordination. Version control across draft scripts, Q&A documents, and disclosure checklists consumes hours that should go to judgment calls. Analysts on the team spend days hunting through prior transcripts and board materials to reconstruct context that already exists somewhere in the organization.
Claude Cowork changes the economics of this work. By reading your actual files — prior transcripts, guidance history, internal models, disclosure policies — Claude builds the earnings preparation package from your source material rather than generic templates. It drafts, cross-checks, and flags inconsistencies, leaving your team to focus on the decisions only humans should make: what to disclose, how to frame it, and what to say when the questions get hard.
Building the source-of-truth file set first
Before any prompt, assemble a working folder: last quarter’s transcript, the current quarter’s preliminary results, your guidance history, the earnings script template, and your disclosure policy. Claude Cowork reads these completely before drafting. The prompts below assume this structure and force Claude to ask clarifying questions before producing output — which is where most of the value sits.
First, read these files completely before responding:
prior_transcript.md — full transcript of last quarter’s earnings call, including prepared remarks and Q&A
preliminary_results.md — current quarter preliminary P&L, segment data, and key metrics
guidance_history.md — guidance issued each quarter for the past eight quarters with actual outcomes
script_template.md — our standard prepared remarks structure and required disclosures
Here is a reference for what I want to achieve:
[Upload last quarter’s final approved script as markdown]
Here’s what makes this reference work:
Opening frames the quarter in one sentence before any numbers. Each segment follows the same order: performance, drivers, outlook. Guidance language is hedged with specific qualifiers, never absolute. Transitions between sections are one line, no filler. Closing invites questions without summarizing again.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Prepared remarks script, approximately 1,800 words, plus a 10-question anticipated Q&A section
Recipient’s reaction: The CFO should read it once and say “this sounds like me” — no rewrites needed on tone
Does NOT sound like: A press release, a marketing deck, or a generic analyst summary
Success means: Zero factual inconsistencies with preliminary_results.md and every guidance statement traceable to guidance_history.md
My context file contains my standards, constraints, audience. Read it fully before starting.
DO NOT start executing yet. Ask clarifying questions first.
Give me your execution plan (5 steps max) before you begin.
Stress-testing the Q&A before the Street does
The prepared remarks are the easy part. The call is won or lost in Q&A, where sell-side analysts probe guidance, margins, and anything the script glossed over. A second prompt should turn Claude into a hostile analyst: generate the toughest plausible questions from the transcript and results, then draft answers grounded in disclosed facts with clear escalation flags for anything the CFO should not answer live.
First, read these files completely before responding:
draft_script.md — the prepared remarks script for this quarter
preliminary_results.md — current quarter preliminary P&L, segment data, and key metrics
prior_transcript.md — last quarter’s Q&A, including questions we handled poorly
disclosure_policy.md — our rules on forward-looking statements and what cannot be said on a live call
Here is a reference for what I want to achieve:
[Upload a past Q&A prep document that worked well as markdown]
Here’s what makes this reference work:
Questions are grouped by theme, hardest first. Each answer opens with a direct response in one sentence, then supports with a specific number, then stops. Deferral language is pre-approved and consistent. Every answer cites the source file and line where the supporting figure appears.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Q&A document, 15 questions, each with a 60-90 word drafted answer and a source citation
Recipient’s reaction: The CFO should feel the questions are genuinely adversarial, not softball, and trust every number
Does NOT sound like: A list of generic investor questions or answers that speculate beyond disclosed guidance
Success means: Every drafted answer is traceable to a source file, and at least 3 questions are flagged as defer-to-IR or defer-to-counsel
My context file contains my standards, constraints, audience. Read it fully before starting.
DO NOT start executing yet. Ask clarifying questions first.
Give me your execution plan (5 steps max) before you begin.
Run both prompts at least a week before the call, not the night before. The clarifying questions Claude asks in step one often surface gaps — a metric nobody owns, a guidance comparison that does not reconcile — that are far cheaper to fix early.
Next step: after the call, feed the actual transcript back into the same folder. Claude can diff your prepared Q&A against what analysts actually asked, and that diff becomes the highest-value input for next quarter’s preparation.
Published on 2 November 2026 on growwithgpt.com