Financial due diligence in M&A is a race against the clock with the highest possible stakes. Buy-side teams typically receive a target’s data room — trial balances, monthly management accounts, aged AR and AP ledgers, bank statements, payroll registers, revenue contracts — with four to six weeks to form a view on quality of earnings, working capital normalization, and hidden liabilities. The work is not intellectually difficult; it is volumetrically brutal. Analysts spend 70 percent of their time reconciling exports, chasing variances, and reformatting the same numbers into yet another schedule, and only 30 percent on the judgment the deal actually pays for. When a seller delivers 14,000 rows of general ledger detail across three entities and two currencies, the clock does not stop while you clean it.
Claude Cowork changes the economics of that timeline. Instead of asking a chat assistant to summarize a single file, Cowork operates across a folder of deal documents the way a junior analyst would: it reads every file completely, cross-references figures between the trial balance and the management accounts, flags inconsistencies, and drafts the schedules and memos your deal team reviews. You keep the judgment; the tool absorbs the mechanical load. For a CFO or controller who has run diligence before, the difference is not incremental — it is the difference between a two-week scoping phase and a two-day one.
The critical discipline is that Cowork performs at the level of the instructions it receives. Vague prompts produce vague schedules. The two prompt structures below are built to be copied, adapted with your deal specifics, and run against a structured data room folder — one for quality-of-earnings analysis, one for a red-flag exception report.
The data room discipline that makes Cowork work
Before you prompt, spend thirty minutes normalizing your folder. Convert everything to text or markdown where possible, name files predictably (tb_2025_entity_a.md, mgmt_accounts_monthly_2024.md), and place a context file at the root that states the deal thesis, the target’s accounting policies, and your materiality threshold. Cowork reads filenames and folder structure as signal. A messy data room produces a messy analysis; a disciplined one produces something a partner can review with minimal rework.
First, read these files completely before responding:
tb_2024_2025.md — full trial balance detail by entity and month for FY2024 and FY2025
mgmt_accounts_monthly.md — monthly P&L and balance sheet for the same period
addback_schedule_seller.xlsx.md — seller’s proposed addbacks with descriptions
context.md — deal thesis, accounting policies, materiality threshold of [$X]
Here is a reference for what I want to achieve:
[Upload a prior QoE bridge from a closed deal as markdown]
Here’s what makes this reference work:
It separates reported EBITDA from adjusted EBITDA in a single waterfall, labels every adjustment with a one-line rationale and a source reference, distinguishes recurring from non-recurring items, and shows a sensitivity range rather than a single point estimate.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: QoE bridge table plus 800-word narrative memo
Recipient’s reaction: the deal partner should be able to challenge any adjustment and find its source in under a minute
Does NOT sound like: a sell-side information memorandum or a marketing summary
Success means: every addback is either supported by source data or flagged as unsupported, with a quantified EBITDA impact
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.
Note what the prompt does: it forces Cowork to declare its plan before touching the numbers, which surfaces scope misunderstandings while they are still cheap to fix. It also demands source traceability on every adjustment — the single most common failure point in AI-assisted diligence, where a plausible-looking addback cannot be tied back to a ledger line. If Cowork asks whether to treat a one-off legal settlement as recurring, answer explicitly; that exchange is the value, not a delay.
From analysis to exceptions: the red-flag report
The second prompt targets the opposite motion. Rather than building a normalized number, it hunts for what is wrong: revenue recognized ahead of delivery, journal entries posted outside business hours or by unusual users, related-party transactions buried in AP, working capital swings that do not match the revenue trend. This is the schedule a CFO scans first because it determines whether the deal proceeds, reprices, or dies.
First, read these files completely before responding:
gl_detail_full.md — general ledger detail with posting dates, users, and descriptions
ar_aging_monthly.md — aged receivables by customer for the trailing 24 months
ap_ledger.md — accounts payable detail including vendor names and payment terms
bank_statements_2025.md — bank activity for the review period
context.md — deal thesis, accounting policies, materiality threshold of [$X]
Here is a reference for what I want to achieve:
[Upload a prior red-flag report from a completed diligence as markdown]
Here’s what makes this reference work:
It ranks findings by financial impact and likelihood, states the test performed, shows the evidence, and proposes a specific next action rather than a generic concern.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: ranked exception table plus a 600-word summary for the deal partner
Recipient’s reaction: the partner should know within two minutes which three findings matter most and what to do about each
Does NOT sound like: an audit opinion or a compliance checklist
Success means: every finding is quantified in dollars where possible and tied to a named source file and line reference
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 against the same data room and you get a complete first-pass view: a defensible earnings number and the exceptions that could move it. From there, the practical next step is to feed Cowork’s exception list into your management interview agenda and let it draft the follow-up question set, one question per finding, each citing the evidence. That closes the loop between analysis and negotiation — and it is where the hours you saved on reconciliation get reinvested into the judgment that actually prices the deal.
Published on 29 September 2026 on growwithgpt.com