For most finance teams, the monthly business review is a ritual of reconstruction. The data exists in the ERP, the variance commentary lives in someone’s inbox, the prior month’s deck sits in a shared drive, and the executive narrative is rebuilt from scratch every thirty days. Controllers and financial analysts spend the better part of a week pulling numbers, chasing department heads for explanations, reformatting tables, and rewriting the same story with new figures. By the time the deck is finished, the analysis is already stale and the team is too exhausted to defend it properly.
The friction is not analytical — it is mechanical. The same slides get rebuilt, the same variance thresholds get re-explained, the same formatting rules get re-applied by hand. Meanwhile, the genuinely valuable work — interpreting what the numbers mean for the business and preparing for the questions the CFO will actually ask — gets squeezed into whatever time remains. This is precisely the kind of repetitive, structure-heavy work that Claude Cowork is built to absorb.
Claude Cowork is Anthropic’s agentic workspace for knowledge work: it reads your files, follows your standards, and produces finished deliverables rather than suggestions. For monthly business reviews, that means you can hand it your actuals, your budget, your prior deck, and your commentary log, and receive a structured first draft — variance tables built, narrative framed, exceptions flagged — that your team refines instead of assembles. The time saved is real, but the bigger gain is consistency: every month’s review follows the same logic, the same thresholds, and the same executive tone, regardless of who is on rotation.
What Cowork needs from you before it can be useful
The quality of a Cowork output is bounded by the quality of its inputs. Before prompting, assemble a small set of reference files: the prior month’s deck, the current month’s trial balance or management pack, a variance threshold policy (what counts as material and at what level it must be explained), and a short log of commentary collected from department heads. Save them as markdown where possible so Claude can read them completely rather than parsing slides visually.
The prompt below follows the anatomy-of-a-prompt structure: it defines the task and success criteria, lists the files to read in full, provides a reference for the target output, and — critically — instructs Claude to ask clarifying questions and present a plan before executing. That last instruction is what separates a usable draft from a plausible-sounding mess.
First, read these files completely before responding:
prior_mbr_deck.md — last month’s full deck, including slide order, section headers, and commentary style
management_pack_[MONTH].md — current month actuals vs. budget vs. forecast by P&L line and by department
variance_policy.md — materiality thresholds, required explanation levels, and escalation rules
commentary_log.md — raw commentary collected from department heads this month, unedited
Here is a reference for what I want to achieve:
Use prior_mbr_deck.md as the structural and tonal reference. Match its section sequence, its table layouts, and its narrative voice.
Here’s what makes this reference work:
It opens with a one-line headline takeaway before any table. Variances are stated in both absolute and percentage terms, always against budget and prior month. Every material variance has a two-sentence explanation: what happened, and what management is doing about it. Non-material items are aggregated, never listed individually. Tone is direct and unhedged — no “slightly below” language.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: full slide-by-slide draft in markdown, approximately [N] slides, matching the prior deck’s structure
Recipient’s reaction: the CFO should be able to read the draft and immediately identify the three issues that need discussion, without hunting
Does NOT sound like: a data dump, a generic template, or commentary that restates the numbers without interpreting them
Success means: fewer than [N] slides require structural rework, and every material variance has an owner-attributed explanation
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.
From first draft to executive-ready: the second pass
The first prompt produces structure. The second prompt produces judgment. Once the draft exists, the highest-value use of Cowork is adversarial: have it interrogate the deck the way a CFO would. This is where most monthly reviews fall apart — the numbers are right, but the narrative collapses under three pointed questions. A second pass that stress-tests the story before the meeting is worth more than an hour of formatting polish.
First, read these files completely before responding:
mbr_draft_[MONTH].md — the draft deck produced in the previous step
variance_policy.md — materiality thresholds and escalation rules
cfo_question_log.md — questions asked in the last [N] monthly reviews, verbatim
management_pack_[MONTH].md — the underlying actuals for cross-checking every claim in the draft
Here is a reference for what I want to achieve:
A red-team review document, not a revised deck. Model it on the structure of a post-mortem: findings ranked by severity, each with the slide reference, the risk, and a suggested fix.
Here’s what makes this reference work:
Findings are ranked, not listed. Each finding names the specific slide and the specific number or claim at risk. Suggested fixes are concrete rewrites, not “consider clarifying.” It distinguishes between errors (the deck is wrong) and vulnerabilities (the deck is right but will invite a hard question).
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: ranked findings list, maximum [N] items, each under [N] words
Recipient’s reaction: the presenter should feel prepared, not criticized — every finding comes with a usable answer
Does NOT sound like: generic presentation advice, or a rewrite of the deck itself
Success means: every question in cfo_question_log.md that applies to this month is either answered in the deck or flagged with a prepared response
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 pass at least two business days before the review, not the night before. The findings often surface a commentary gap that requires a real conversation with a department head — and that conversation is the actual work the monthly review exists to force. Cowork compresses the assembly and interrogation; it does not replace the judgment.
What to try next: build a reusable context file that captures your organization’s review standards — slide order, materiality thresholds, tone rules, recurring executive concerns — and load it at the start of every monthly cycle. Once that file exists, both prompts above shrink to a few lines of month-specific input, and the entire review process becomes a two-hour exercise instead of a two-week one.
Published on 14 November 2026 on growwithgpt.com