For any controller or CFO in the construction sector, the month-end close is rarely just a numbers exercise—it is a forensic audit of judgment calls. IFRS 15 (Revenue from Contracts with Customers) demands that you recognize revenue over time only when you can reasonably measure progress, enforce enforceable rights to payment, and identify performance obligations that are distinct. In practice, this means sifting through hundreds of change orders, variation claims, and cost-to-cost calculations, each of which carries its own risk of misstatement. The friction is real: project managers track physical progress, while your finance team tracks incurred costs, and the two datasets rarely reconcile without hours of manual adjustment.
Generative AI, specifically large language models (LLMs) like Claude, does not replace your judgment—it amplifies your capacity to apply it consistently. By feeding the AI a structured prompt that includes your contract terms, cost reports, and the specific IFRS 15 guidance applicable to your jurisdiction, you can generate a draft revenue recognition memorandum, flag potential performance obligation splits, and even propose the correct progress percentage based on cost-to-cost versus output methods. The AI does the heavy lifting of pattern recognition across your contract portfolio, freeing you to focus on the high-stakes exceptions that require human expertise.
The key to making this work is not asking the AI a vague question like “help me with IFRS 15.” Instead, you need to provide it with the same structured context you would give a new senior accountant on your team: the contract’s payment terms, the expected margin, the risk of customer cancellation, and the specific clauses that affect enforceability. The prompts below are designed to do exactly that—they are not magic spells but structured briefs that force the AI to think like a construction revenue specialist before it writes a single word.
Why Most AI Outputs Fail in Construction Accounting
The most common failure mode is that the AI produces generic revenue recognition text that could apply to a software subscription or a consulting engagement. Construction contracts are uniquely messy: they involve retentions, mobilisation fees, unapproved change orders, and cost overruns that may or may not be recoverable. A prompt that does not explicitly reference these elements will yield output that is technically accurate but practically useless. The following pre-box prompt templates are built to eliminate that ambiguity by forcing the AI to acknowledge the specific contract mechanics before it proposes any accounting treatment.
First, read these files completely before responding:
[contract_summary.md] — Contains the contract value, payment schedule, retention clause, and specific performance obligations as written in the signed agreement.
[cost_report_q2.md] — The latest job cost report showing actual costs incurred to date, committed costs, and the project manager’s estimated percentage of physical completion.
[ifrs15_checklist.md] — A condensed version of IFRS 15 paragraphs 35-45 (over time recognition) and 56-63 (measuring progress) tailored for construction.
Here is a reference for what I want to achieve:
[Upload a previously approved revenue recognition memo from a similar project as markdown, or describe the format: 3-page memo with sections for Contract Overview, Performance Obligations, Recognition Method Justification, Progress Measurement, and Risk Factors.]
Here’s what makes this reference work:
The memo explicitly cites the specific IFRS 15 paragraph numbers for each judgment call. It includes a sensitivity table showing revenue recognized under cost-to-cost versus output method. It flags any unapproved change orders as a separate risk item rather than burying them in the main calculation. The tone is neutral and factual, with no optimistic language about recoverability.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: 3-page professional accounting memo, approximately 1,200 words.
Recipient’s reaction: The audit partner should be able to follow our logic without asking follow-up questions about the contract terms.
Does NOT sound like: A textbook explanation of IFRS 15. It must reference the specific numbers and clauses from my contract and cost report.
Success means: The memo is accepted as supporting documentation for the Q2 2026 revenue figure with no more than one clarifying question.
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.
This first prompt is designed for the initial recognition memo. Notice how it forces the AI to distinguish between the contract’s legal terms and the cost data—a distinction that is critical under IFRS 15 because you cannot recognize revenue on unapproved variations unless you have a high probability of recovery. The prompt also requires the AI to ask clarifying questions before executing, which is essential because a generic response will not capture the specific risk profile of your project.
Handling Change Orders and Variations Under IFRS 15
Change orders are where most construction revenue misstatements occur. Under IFRS 15, a variation is recognized only when it is highly probable that including it will not cause a significant revenue reversal. That “highly probable” threshold is a judgment call, and it depends on your historical experience with the client, the legal enforceability of the claim, and the political dynamics on the project site. The second prompt below focuses on this specific pain point—generating a risk assessment and proposed accounting treatment for a batch of pending change orders.
First, read these files completely before responding:
[change_order_log.md] — A table listing each change order number, description, value, date submitted, client’s verbal response, and whether the work has already been performed.
[client_history.md] — A summary of our past three projects with this client, showing the percentage of change orders that were ultimately paid in full, partially paid, or rejected.
[contract_general_conditions.md] — The section of our contract governing change order approval procedures, including the 30-day written notice requirement and dispute resolution clause.
Here is a reference for what I want to achieve:
[Upload a previous change order assessment memo, or describe the format: a 2-page matrix with columns for Change Order ID, Work Performed (Yes/No), Client Communication, Probability Assessment (High/Medium/Low), and Recommended Revenue Recognition Amount.]
Here’s what makes this reference work:
The matrix uses a simple probability scale rather than vague language like “likely” or “probable.” Each change order has a one-line justification citing either the client’s written approval, past payment behavior, or the specific contract clause. The memo concludes with a total revenue adjustment figure that reconciles to the general ledger.
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A 2-page assessment matrix with a 300-word narrative summary.
Recipient’s reaction: The CFO can quickly see which change orders are recognized and which are deferred, and why.
Does NOT sound like: A legal opinion. This is an accounting assessment, not a contract law analysis.
Success means: We can post the Q2 revenue adjustment with confidence that the audit team will agree with our probability assessments.
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.
When you run this second prompt, expect the AI to ask about the timing of client communications—specifically whether any verbal approvals were given before the 30-day notice period expired. That is a good sign; it means the model is internalizing the contract’s procedural requirements. If the AI does not ask about that, you should challenge its output and refine your context files. The goal is not to get a perfect answer on the first try but to have a structured conversation where the AI forces you to clarify your own assumptions.
One practical tip for using these prompts effectively: do not run them in isolation. Start with the first prompt to establish the baseline revenue recognition for the base contract. Then run the second prompt for change orders. Finally, ask the AI to reconcile the two outputs by generating a combined journal entry with supporting calculations. This sequential approach mirrors how your team actually works and produces documentation that is audit-ready because each assumption is traceable to a specific input file.
What to try next: after you have used these prompts for one quarter, save the AI’s output and your edits as a new reference file. Feed that edited version back into the prompt template the following quarter. This creates a feedback loop where the AI learns your specific presentation style, your firm’s risk appetite, and the particular quirks of your largest clients. Over time, the AI will require fewer clarifying questions, and your memo drafting time will drop from four hours to forty-five minutes.
Published on 31 July 2026 on growwithgpt.com