Financial due diligence is the part of an M&A deal where time pressure and analytical depth collide. Buy-side teams typically get three to five weeks to normalize EBITDA, quantify working capital pegs, trace quality-of-earnings adjustments, and surface the risks that will either reprice the deal or kill it. The raw material arrives as a data room dump: trial balances, management accounts, bank statements, payroll registers, customer concentration schedules, and a dozen versions of the same revenue bridge. Someone on the team then spends the first week doing nothing but formatting, reconciling, and rebuilding schedules that the target’s finance function should have produced but rarely does.
The friction is not analytical. It is mechanical. Senior analysts burn their best hours on variance tie-outs and file hygiene instead of judgment calls, while the partner asks for a QoE summary before the analysts have even finished mapping the chart of accounts. Every hour spent restructuring a CSV is an hour not spent asking why gross margin stepped up in Q3 or why DSO dropped the month before the letter of intent was signed.
Claude Cowork changes the economics of that first week. Working against a structured context file that encodes your firm’s diligence standards, materiality thresholds, and reporting conventions, Claude can ingest the data room, reconcile management accounts to trial balances, flag unexplained variances, build normalized EBITDA bridges, and draft the first pass of the QoE report in your house format. You stay in the judgment seat; the formatting and reconciliation grind moves off your desk. The result is a diligence process that starts at analysis rather than at cleanup.
Why context files matter more than prompts in deal work
In M&A, generic output is dangerous. A normalized EBITDA schedule that ignores your firm’s treatment of non-recurring items, or a working capital analysis that uses the wrong peg convention, creates rework and, worse, credibility risk with the investment committee. Claude Cowork’s context file mechanism lets you encode your standards once: adjustment categories, materiality floors, the definition of run-rate synergies, the exact tab structure of your QoE deliverable. Every subsequent prompt inherits those constraints.
The two prompts below follow the anatomy-of-a-prompt structure. The first builds the normalized EBITDA and QoE foundation. The second pressure-tests working capital and produces the peg recommendation. Both are designed to be run against a data room extract and your firm’s context file.
First, read these files completely before responding:
[context.md] — my firm’s diligence standards, adjustment categories, materiality thresholds, and QoE report conventions
[trial_balance.csv] — target’s monthly trial balance for the trailing 24 months
[management_accounts.xlsx] — target’s monthly P&L and balance sheet as presented by management
[adjustment_log.md] — management’s own list of proposed add-backs with descriptions
Here is a reference for what I want to achieve:
[Upload prior QoE report as markdown] — a completed QoE from a closed deal in the same sector.
Here’s what makes this reference work:
Adjustments are grouped into five categories (non-recurring, owner-related, run-rate, accounting policy, pro forma). Each adjustment shows a dollar amount, a source reference, a rationale of two sentences maximum, and a confidence rating. The EBITDA bridge reconciles reported to normalized in a single table. Tone is factual, no hedging language, no adjectives.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Normalized EBITDA bridge table plus adjustment schedule, 1,200 to 1,800 words of supporting narrative
Recipient’s reaction: The deal partner should be able to challenge any adjustment and find the source reference immediately
Does NOT sound like: A sell-side CIM, a marketing document, or a generic consulting deck
Success means: Every adjustment above [materiality threshold] is sourced, categorized, and rated, with unexplained variances flagged for follow-up
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 that prompt after uploading the data room extract and your context file. Expect Claude to come back with clarifying questions about materiality thresholds, the treatment of owner compensation, and whether management’s adjustment log should be accepted at face value or rebuilt from source. Answer those questions before letting it execute. The clarifying-question step is where most of the value is created, because it forces you to state assumptions you would otherwise leave implicit.
From EBITDA to the working capital peg
Once the earnings side is normalized, the next battleground is the working capital peg. This is where deals get repriced at signing, and where sloppy analysis is most expensive. The prompt below produces a monthly working capital analysis, identifies the peg methodology, and drafts the negotiation position with supporting evidence from the trailing twelve months.
First, read these files completely before responding:
[context.md] — my firm’s peg methodology, excluded accounts, and target working capital conventions
[balance_sheet_monthly.xlsx] — 24 months of monthly balance sheet detail by account
[ar_aging.xlsx] — accounts receivable aging by customer for the trailing 12 months
[ap_aging.xlsx] — accounts payable aging by vendor for the trailing 12 months
[revenue_bridge.md] — monthly revenue with seasonality notes from management
Here is a reference for what I want to achieve:
[Upload prior peg analysis as markdown] — a working capital analysis from a closed deal.
Here’s what makes this reference work:
Twelve-month average, trailing six-month average, and LTM-end are all presented side by side. Excluded accounts (cash, debt, income tax, dividends) are listed with rationale. Seasonality is addressed explicitly with a monthly index. The recommendation states a single number with a stated confidence level and the two strongest counterarguments.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Working capital analysis with monthly trend table and peg recommendation, 1,000 to 1,500 words
Recipient’s reaction: The counterparty’s advisors should see the peg as data-driven, not negotiated
Does NOT sound like: An aggressive buyer position, a one-sided argument, or a textbook definition of working capital
Success means: A single recommended peg with three supporting methods, seasonality quantified, and the top three negotiation risks identified
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.
The practical tip: treat Claude Cowork as the first-year analyst you wish you had, not the partner. It will produce a complete, well-structured first draft, but every adjustment, peg, and variance still needs your sign-off. Build the review step into your workflow explicitly, and keep the source references intact so the investment committee can trace any number back to the trial balance in one click.
What to try next: run both prompts on a closed deal where you already know the answer. Compare Claude’s normalized EBITDA and peg to the final signed numbers. That calibration exercise will tell you exactly where your context file needs tightening before you point it at a live deal.
Published on 14 October 2026 on growwithgpt.com