For any CFO or financial controller, the due diligence phase of a merger or acquisition is where deals go to die—or where they quietly become overpriced. The friction is brutal: you receive a data room with 15,000 documents, many of them scanned PDFs, messy Excel exports, and inconsistent accounting treatments across target entities. Your team spends three weeks pulling together quality of earnings (QoE) adjustments, net working capital (NWC) baselines, and debt-like items. By the time you have a clear picture, the seller’s exclusivity window is closing, and you are negotiating from a position of time scarcity rather than analytical strength.
The core problem is not a lack of intelligence—it is a lack of structured, iterative collaboration with the data. Traditional tools force you to either read everything (slow) or sample and risk missing a material contingency. Claude Cowork changes this dynamic. It acts as a senior analyst who never sleeps, reads every page, and—critically—asks you clarifying questions before it executes. You are not prompting a chatbot for a summary; you are co-building an audit trail. Claude can reconcile intercompany balances, flag non-recurring expenses, and stress-test your NWC assumptions against historical seasonality. The result is a due diligence process that is faster, more defensible, and far less prone to the “we missed that indemnity clause” regret.
This post gives you two production-ready prompt templates. They are built for the way a financial professional thinks: with a success criterion, a reference framework, and an explicit instruction to pause and ask questions before touching the data. Use them as your starting point, then adapt them to your specific deal structure.
Why Claude Cowork Beats a Standard Chatbot for QoE
A standard chatbot gives you a generic answer. Claude Cowork, when prompted with the structure below, behaves like a peer reviewer. It first reads the files you designate, then it extracts the patterns from a reference document you provide, and only then does it ask clarifying questions. This is crucial in M&A because every deal has a unique accounting framework—cash vs. accrual, IFRS vs. GAAP, or a target that capitalizes software development costs. The prompt forces Claude to acknowledge those variables before it produces a single adjustment. You avoid the trap of receiving a beautifully formatted report built on the wrong assumptions.
First, read these files completely before responding:
[target_trial_balance.xlsx] — the last 36 months of monthly P&L data, including natural account descriptions and cost center tags.
[management_adjustments.xlsx] — the seller’s own reclassification entries, including any one-off addbacks they claim are non-recurring.
[prior_qoe_template.pdf] — a clean, two-page QoE bridge from a previous deal that the committee approved without pushback.
Here is a reference for what I want to achieve:
The prior QoE template shows a waterfall chart starting from reported EBITDA, then listing each adjustment with a reference number, a one-line rationale, and a column for “recurring” or “non-recurring.” The final row shows normalized EBITDA and a footnote with the key assumptions (e.g., FX rate used, inventory valuation method).
Here’s what makes this reference work:
– Every adjustment has a unique ID (e.g., ADJ-001) that links to a supporting schedule.
– Non-recurring items are separated from normalizing adjustments (e.g., owner’s salary above market rate).
– The tone is neutral; no judgment words like “aggressive” or “conservative” — just facts.
– The bridge footnotes disclose any changes in accounting estimates from prior periods.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A structured QoE bridge in Excel format (or markdown table if Excel is not possible), with a maximum of 20 adjustments, each with a reference, amount, direction, and rationale. Include a summary tab with normalized EBITDA and a bridge from reported to normalized.
Recipient’s reaction: The investment committee should be able to spot the single largest non-recurring item in under 30 seconds and understand why it is excluded from run-rate EBITDA.
Does NOT sound like: A generic audit report with vague wording like “various adjustments” or “miscellaneous expenses.”
Success means: I can send this bridge to the external auditors without them asking a single clarifying question about the source of any adjustment.
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.
Moving from QoE to Net Working Capital and Debt-Like Items
Once you have normalized EBITDA, the next battleground is the NWC target and the identification of debt-like items. This is where sellers often try to inflate the baseline NWC to reduce the purchase price adjustment. The friction is real: you need to define “cash” (exclude restricted cash), “debt” (include finance leases, but what about vendor financing?), and “working capital” (which receivables are truly collectible?). A typical analyst spends two days just categorizing the trial balance. Claude Cowork can compress that to hours—but only if you give it a precise reference for how you define debt-like items, and only if it asks you the right clarifying questions first.
The prompt below is designed for that exact scenario. Note how it forces Claude to ask about your deal-specific definitions before it categorizes a single account. This is the difference between a useful tool and a dangerous one—because a wrong categorization of a debt-like item can swing the purchase price by millions.
First, read these files completely before responding:
[target_balance_sheet.xlsx] — the trial balance as of the closing date, including all asset, liability, and equity accounts with sub-ledger detail.
[spa_nwc_definition.docx] — the exact definition of NWC from the Sale and Purchase Agreement, including exclusions and any specific accounting policies referenced.
[opening_balance_sheet.xlsx] — the audited balance sheet from 12 months prior, to identify any reclassifications or changes in presentation.
Here is a reference for what I want to achieve:
A three-column schedule: (1) account name, (2) balance per target books, (3) adjustment and rationale. The final summary shows: Total Cash (excl. restricted), Total Debt (incl. all interest-bearing liabilities and finance leases), and NWC (defined as current assets minus current liabilities, excluding cash and debt). Each line item has a reference to the SPA clause that governs its treatment.
Here’s what makes this reference work:
– The schedule is built from the trial balance, not from the target’s own presentation, to avoid inherited misclassifications.
– Every excluded item (e.g., deferred revenue) has a footnote citing the SPA section.
– The tone is legalistic but clear—no room for interpretation.
– The output separates “clearly debt-like” from “disputed items” so we know where to negotiate.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A single Excel workbook with three tabs: (1) NWC calculation, (2) Debt-like items, (3) Disputed items log. Each tab must have a column for “Evidence” linking to the source ledger page.
Recipient’s reaction: The seller’s CFO should see that our position is grounded in the SPA language, not in accounting judgment calls that they can challenge.
Does NOT sound like: A generic balance sheet analysis with no contractual backing.
Success means: We have a negotiation-ready schedule that reduces the expected dispute window from 30 days to under 7 days.
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.
Practical Tips for Your Next Deal
Here is the practical tip: do not run these prompts on the full data room immediately. Instead, start with a subset—say, one quarter of the trial balance or one legal entity—and validate Claude’s output against your own manual calculation. Once you confirm the logic is aligned with your deal’s specific accounting policies, scale it to the full dataset. This iterative approach builds trust in the tool without risking a costly error.
What to try next: after you have the QoE bridge and the NWC schedule, ask Claude to generate a draft of the “Seller’s Disclosure Schedule” exceptions based on the discrepancies it finds between the target’s representations and the actual ledger data. That is a high-value, low-effort win that typically takes a senior analyst a full day. With the prompts above, you will have that draft in under an hour.
Published on 28 August 2026 on growwithgpt.com