The month-end close is a crucible for every finance team. For those managing multi-entity consolidations, the most punishing phase is arguably the intercompany reconciliation and elimination process. You know the drill: a labyrinth of trial balances, manual tick-marks in Excel, endless email chains with subsidiary controllers, and the nagging fear that a 50-cent difference in a payable will balloon into a material audit finding. The pain is acute—hours of manual matching, frustration over mismatched currency translations, and the sheer cognitive load of tracking which entries have been confirmed and which are still open. This friction doesn’t just waste time; it erodes trust in the numbers and delays the consolidated financial statements that the board demands.
Artificial intelligence—specifically large language models like Claude—offers a radically different approach. Instead of spending days reconciling intercompany accounts row by row, you can now feed your raw trial balance files and intercompany transaction logs into an AI tool that reads, analyzes, and matches them in minutes. The AI doesn’t get tired, doesn’t miss a line item, and can handle the complexity of partial matches, timing differences, and currency adjustments. It outputs a structured elimination schedule with flagged exceptions for human review. This isn’t about replacing the controller; it’s about eliminating the drudgery so your team can focus on analysis, judgment, and strategic decisions. For CFOs and controllers, this means a faster close, lower audit fees, and a significantly reduced risk of material misstatement.
The key to unlocking this capability lies not in the AI itself, but in how you instruct it. A vague request like “match these intercompany transactions” will yield mediocre results. You need a structured prompt that defines the rules of engagement, provides reference material, and specifies success criteria. The following two prompts are designed to be copy-paste ready for your team. They embody the anatomy of an effective Claude instruction—clear task definition, contextual files, reference examples, and a precise success brief. Use them to transform your next consolidation cycle.
Why Intercompany Matching Demands a Structured Prompt
Standard AI chat interfaces are conversational. But consolidation accounting is governed by strict rules—GAAP, IFRS, tax regulations, and internal policies. A conversational prompt will produce generic output. You need to give Claude a “playbook.” The structured prompt below forces you to articulate your specific matching logic: tolerance thresholds for currency differences, rules for partial payments, treatment of intercompany loans versus trade payables, and the format of your elimination journal entries. By uploading your trial balance files and a reference elimination schedule, you train Claude on your exact context. The result is output that is audit-ready, not just “interesting.”
Before you dive into the prompts, prepare your files. You will need your latest intercompany trial balance (in CSV or Excel), a list of intercompany transactions with counterparty identifiers, and—crucially—a reference file showing a correctly formatted elimination schedule from a prior period. This reference file is the single most important element. It teaches Claude what “good” looks like in your organization.
First, read these files completely before responding:
[consolidation_standard_IFRS.md] — Our internal IFRS consolidation policy, including elimination rules, currency translation procedures, and materiality thresholds
[intercompany_trial_balance_Q2_2026.csv] — Raw trial balance for all legal entities, with columns: Entity, Account Code, Counterparty, Currency, Amount in Local Currency, Amount in Reporting Currency
[reference_elimination_schedule_prior_period.xlsx] — A correctly formatted elimination schedule from our last quarter, showing the exact layout, column headers, and journal entry formatting that the audit team approved
Here is a reference for what I want to achieve:
The attached reference elimination schedule shows how we present matched transactions, unmatched items, and currency adjustment entries. It uses a three-section format: (1) confirmed matches with zero difference, (2) matches with timing differences under $5,000, and (3) unmatched items requiring manual review. Each entry includes a unique match ID, the two entity names, the original amounts, the elimination amount, and a cross-reference to the source file row.
Here’s what makes this reference work:
– Every match is assigned a unique alphanumeric ID (e.g., IC-2026-001) that links back to the source rows
– Currency differences are shown as a separate line item labeled “FX Adjustment” with a note explaining the rate used
– Unmatched items are sorted by absolute value descending, so the largest risks appear first
– The schedule uses a consistent date format (YYYY-MM-DD) and no merged cells
– All elimination entries are pre-formatted as debit/credit pairs with the correct account codes from our chart of accounts
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A structured elimination schedule in markdown table format, approximately 50-200 rows depending on transaction volume. Include a summary section at the top with total matched amount, total unmatched amount, and number of exceptions.
Recipient’s reaction: The controller should be able to review and approve this schedule in under 30 minutes. The audit team should accept it as a supporting schedule without requesting reformatting.
Does NOT sound like: Generic advice, theoretical discussion, or conversational commentary. Do not include explanations of accounting concepts. Stick to the data and the matching logic.
Success means: The output matches the format and quality of the reference schedule exactly, with all intercompany balances eliminated except for those flagged as exceptions. Zero arithmetic errors.
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.
Refining the Output: From Matching to Journal Entry Generation
The first prompt gets you a clean elimination schedule. But the real prize is automating the journal entries themselves. Many controllers still manually copy elimination amounts from a spreadsheet into the ERP system—a step that introduces transcription errors. The prompt below extends the AI’s capability: after matching, it generates the actual debit and credit journal entries in a format that can be uploaded directly into your consolidation software. This prompt also introduces a “confidence score” for each match, which helps your team prioritize which exceptions to investigate first. It is designed for the second pass of your consolidation process, after the initial matching is complete.
Note the addition of a “feedback loop” in this prompt. You can include a small sample of corrected matches from a prior cycle. This teaches Claude to learn from your team’s previous judgment calls—for example, how you handled a disputed intercompany loan that was settled in a different currency. Over time, the AI’s accuracy improves because you are effectively curating a training dataset with each use. This is the difference between a static automation tool and an adaptive AI assistant that gets smarter the more you use it.
First, read these files completely before responding:
[matched_elimination_schedule_Q2_2026.csv] — The output from the previous matching step, containing match IDs, entity pairs, amounts, and exception flags
[erp_journal_entry_template.xlsx] — Our SAP journal entry upload template, with required columns: Document Date, Posting Date, Company Code, Account, Debit Amount, Credit Amount, Text, and Reference
[correction_log_prior_cycles.csv] — A log of 20 corrections our team made to AI-generated entries in the last two quarters, showing the original entry, the corrected entry, and the reason for the change
Here is a reference for what I want to achieve:
The attached correction log shows that most errors occurred when the AI used the wrong account code for intercompany loans versus trade payables. The pattern shows that loans should always use account 14100 (Intercompany Receivable) and 24100 (Intercompany Payable), while trade items use 12100 and 22100. The reference template also shows that the “Text” field must include the match ID and a brief description (e.g., “IC-2026-001: Entity A to Entity B, trade payable elimination”).
Here’s what makes this reference work:
– Each journal entry has a one-to-one correspondence with a match ID from the elimination schedule
– Debits and credits are balanced for each entity pair; the consolidated elimination entry is a single line per entity
– The posting date is always the last day of the reporting period
– The reference field contains the subsidiary invoice number for traceability
– Entries for unmatched items are omitted (they require manual review before posting)
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A CSV file with approximately 50-200 rows, formatted exactly to match the ERP upload template. Include a header row with the exact column names from the template.
Recipient’s reaction: The finance team should be able to upload this file directly into SAP and have zero validation errors. The audit trail should be clear from the reference field.
Does NOT sound like: Markdown tables, bullet points, or any formatting that is not CSV-compatible. Do not include explanations, notes, or warnings in the output itself—only data rows.
Success means: The uploaded file passes SAP’s journal entry validation on the first attempt. The total of all debits equals the total of all credits. Every entry has a valid account code from our chart of accounts.
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 First AI-Powered Consolidation
Start small. Do not attempt to automate your entire intercompany elimination process in one go. Pick a single entity pair—for example, your US parent and your UK subsidiary—and run the first prompt with their trial balances. Review the output against your manual process. You will likely find that the AI handles 80-90% of the matches correctly. The remaining 10-20% will reveal gaps in your prompt—perhaps you need to specify a tolerance for currency rounding, or you need to upload a more detailed chart of accounts. Adjust your prompt, re-run, and iterate. After two or three cycles, you will have a prompt that works reliably for that entity pair. Then expand to the next pair.
A second tip: invest time in creating a high-quality reference file. The reference elimination schedule you upload is the single biggest driver of output quality. If your reference file is messy—inconsistent formatting, missing data, manual overrides not documented—the AI will replicate those flaws. Clean your reference file first. Use a period where you are confident the manual process was correct. This is your “gold standard.” Once the AI matches the gold standard consistently, you can trust it to handle new periods.
Finally, do not skip the “Ask clarifying questions” step built into these prompts. When Claude asks you for clarification—for example, “Should I treat intercompany dividends differently from trade payables?”—answer it directly in the chat. Each question and answer refines the AI’s understanding of your specific rules. This interactive refinement is far more powerful than trying to cram every rule into a single prompt. It also builds institutional knowledge: the conversation history becomes a playbook for future consolidations.
Try the first prompt this week with your most recent trial balance data. You will be surprised at how quickly the AI grasps your matching logic. Within a few cycles, you will wonder why you ever did it manually.
Published on 27 June 2026 on growwithgpt.com