Quarterly earnings call preparation is one of the most compressed, high-stakes workflows in corporate finance. Within a window of roughly ten days, a lean IR and FP&A team must reconcile preliminary results, finalize MD&A language, build the analyst Q&A defense, pressure-test guidance, and rehearse the CEO and CFO on a script that cannot afford a single factual slip. The inputs are scattered across ERP exports, board decks, prior call transcripts, analyst models, and a dozen email threads. The output is a 60-minute live event where a misstated metric can move the stock and invite scrutiny from the SEC.
The traditional process is a serial relay: analysts draft, controllers verify, IR rewrites, legal reviews, and the CFO re-reads everything the night before. Each handoff loses context. Each version spawns another file named with a date suffix. The real work — anticipating the three hardest questions and answering them with precision — gets squeezed into whatever time remains after the mechanical assembly is done.
Claude Cowork changes the shape of that workflow. Because it works directly with your files and folders, it can act as a persistent preparation partner: reading your actuals, your prior transcripts, and your guidance history, then drafting, cross-checking, and stress-testing every artifact in the earnings package. It does not replace the controller’s judgment or the CFO’s voice — it removes the assembly tax so that judgment and voice get more time. This post covers two structured prompts you can adapt: one for building the earnings script and Q&A defense, and one for running a hostile-analyst stress test before the real call.
Why structure matters more than model choice
Earnings preparation fails in predictable ways: numbers that do not tie across documents, tone that drifts between the press release and the call script, and Q&A answers that sound defensive because nobody rehearsed the hard version. A vague prompt like “help me prepare for the earnings call” produces generic advice. A structured prompt that names your files, your success criteria, and your constraints produces a working draft that a controller can actually review.
The two templates below follow the same anatomy: task and success criteria first, then explicit file reads, then a reference pattern, then a success brief, then a hard stop for clarifying questions. That last instruction is the most important one. Claude Cowork will ask about segment definitions, non-GAAP reconciliations, and guidance framing before it writes a word — which is exactly the conversation your team should be having on day one, not day nine.
First, read these files completely before responding:
[actuals_q3_2026.xlsx] — preliminary financial results by segment, with GAAP and non-GAAP reconciliations
[prior_call_transcript_q2_2026.md] — full transcript of last quarter’s earnings call, including analyst Q&A
[guidance_history.md] — every guidance figure we have issued for the past eight quarters and whether we beat, met, or missed
[analyst_models_summary.md] — consensus estimates and the three most influential analyst models with their key assumptions
[press_release_draft.md] — current draft of the earnings press release
Here is a reference for what I want to achieve:
[Upload last quarter’s final Q&A prep document as markdown]
Here’s what makes this reference work:
Every anticipated question is grouped by theme (revenue, margin, guidance, capital allocation), each answer leads with the direct number before any explanation, answers are capped at 90 seconds of spoken length, and each entry includes a “bridge” line that redirects to our strategic narrative if the analyst pushes further. Tone is confident and specific, never defensive.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Full call script (8-10 minutes of prepared remarks) plus 20-25 anticipated Q&A entries with answers
Recipient’s reaction: The CFO should read it and think “I can deliver this without editing” and “nothing on this call can surprise me”
Does NOT sound like: A press release read aloud, a defensive legal memo, or a generic IR template
Success means: Every figure in the script traces to actuals_q3_2026.xlsx, and at least 18 of the 25 Q&A entries match questions that actually get asked
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.
Notice what the first prompt forces: a traceable link between every spoken number and a source file, a hard cap on answer length, and a measurable success criterion. When Claude Cowork returns the draft, your controller’s job shifts from writing to verifying — a much faster and more reliable task.
Stress-test before the street does
The second prompt is the one most teams skip, and it is the one that pays off most. Once the script exists, run it through an adversarial pass. The goal is not reassurance; it is to surface the two or three questions that will actually hurt if the CFO has not thought them through. This is where Claude Cowork’s file access matters again — it can compare your prepared answers against the guidance history and flag any place where the script implies a commitment you have not actually made.
First, read these files completely before responding:
[earnings_script_draft.md] — the full prepared remarks and Q&A defense package
[guidance_history.md] — every guidance figure issued for the past eight quarters and actual outcomes
[short_interest_and_sentiment.md] — current short interest, recent analyst rating changes, and notable bear-case commentary
[legal_review_notes.md] — compliance constraints on forward-looking statements and non-GAAP usage
Here is a reference for what I want to achieve:
[Upload a red-team review memo from a prior quarter as markdown]
Here’s what makes this reference work:
It ranks issues by severity (call-breaking, damaging, minor), quotes the exact line from the script that creates the problem, explains the specific follow-up an analyst would use to exploit it, and proposes a rewritten line that closes the gap without sounding evasive.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A ranked red-team memo, maximum 12 issues, each with quoted line, exploit path, and rewritten alternative
Recipient’s reaction: The CFO should finish reading and say “I am glad we caught that now” at least twice
Does NOT sound like: A generic risk checklist or a legal disclaimer exercise
Success means: At least three issues are ones the team had not previously identified, and zero rewritten lines conflict with legal_review_notes.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.
A practical tip: run the stress test at least 48 hours before the call, not the night before. The rewritten lines need a legal pass, and the CFO needs one full rehearsal with the revised answers. Teams that build this into the timeline consistently report that the live call feels like the third or fourth time they have had the conversation — because it is.
What to try next: after this quarter’s call, save the transcript and compare it against your Q&A prep document. Feed both back into Claude Cowork and ask it to identify which anticipated questions missed and which unanticipated ones appeared. That gap analysis becomes the input for next quarter’s prompt, and the preparation loop compounds. Over four quarters, your Q&A defense stops being a scramble and starts being an institutional asset.
Published on 5 October 2026 on growwithgpt.com