For CFOs and financial controllers, the annual impairment test under IAS 36 is one of the most stressful, time-consuming, and high-stakes compliance exercises on the calendar. The process is deceptively complex: you need to build discounted cash flow models for each cash-generating unit, determine appropriate discount rates, forecast terminal values, run sensitivity analyses, and document every assumption in a way that satisfies both your external auditors and local regulators. When you have multiple CGUs across different geographies and business lines, the manual workload becomes crushing. A single spreadsheet error—a wrong growth rate, a mislinked cell, a stale WACC—can trigger a write-down that wipes millions off your balance sheet and sends your share price tumbling.
The pain is compounded by the sheer volume of documentation required. Auditors expect a clear audit trail showing how you derived each value-in-use calculation, including the rationale for your growth rates, the basis for your discount rate, and the results of your sensitivity testing. Most teams spend weeks assembling these workpapers, often pulling all-nighters in the final push before the reporting deadline. And if the auditors push back on a single assumption, the entire cycle starts again.
This is where AI changes the game. By using a structured prompt framework with models like Claude or GPT-4, you can automate the heavy lifting of IAS 36 impairment testing. The AI doesn’t just generate numbers—it builds a complete, auditable impairment model based on your specific financial data, applies the correct IAS 36 methodology, runs sensitivity scenarios, and produces a professional documentation package that stands up to auditor scrutiny. The result: what used to take two weeks of manual work can be completed in two hours, with fewer errors and deeper analytical insights.
How the AI Tool Works: From Data to Documentation
The key to making AI work for impairment testing is not asking it to “do the calculation” in a single, vague prompt. Instead, you feed it your structured financial data—historical revenues, EBITDA margins, capital expenditure plans, and market risk premiums—along with clear instructions on the IAS 36 methodology you want applied. The AI then builds the model step by step, from forecasting free cash flows to calculating the value-in-use and comparing it to the carrying amount. It also generates the sensitivity tables and narrative explanations that auditors expect to see in your impairment assessment report.
Below are two ready-to-use prompts that demonstrate the approach. The first prompt handles the core calculation and model construction. The second prompt focuses on sensitivity analysis and documentation. Both follow the structured “Anatomy of a Prompt” format that ensures the AI understands your context, constraints, and success criteria before it begins executing.
First, read these files completely before responding:
[CGU_data.md] — Contains historical financial data (revenue, EBITDA, D&A, capex, working capital) for each CGU for the past 5 years, plus current carrying amounts and useful lives of assets.
[market_assumptions.md] — Contains risk-free rate, equity risk premium, beta values, cost of debt, and country risk premiums for each CGU’s jurisdiction.
[IAS36_methodology.md] — Contains our firm’s interpretation of IAS 36, including our policy on terminal growth rates (max 2.5%), discount rate calculation (WACC approach), and treatment of goodwill.
Here is a reference for what I want to achieve:
[Upload reference file: “2025_Impairment_Model_Template.xlsx” converted to markdown showing the structure of a completed impairment model with five-year DCF, terminal value calculation, value-in-use summary, and sensitivity tables.]
Here’s what makes this reference work:
– Clear separation between input assumptions, calculations, and output summaries
– Explicit formulas for each cell (no black-box calculations)
– Sensitivity table showing impact of +/- 1% changes in WACC and terminal growth rate
– Audit trail column showing source of each assumption (file name, paragraph reference)
– Professional formatting with section headers and page breaks
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Complete impairment model in structured markdown format, approximately 15-20 pages when rendered, with all calculations shown step-by-step.
Recipient’s reaction: Our audit partner should be able to follow every assumption back to its source and verify all calculations without needing to rebuild the model.
Does NOT sound like: A generic template with placeholder values. Every number must be calculated from the specific data in my input files.
Success means: The model passes audit review on first submission, with zero material adjustments required.
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.
Why Structure Matters for Financial AI Prompts
The prompt above works because it forces the AI to understand the complete context before generating anything. By requiring the AI to ask clarifying questions and provide an execution plan, you prevent it from making assumptions about your methodology or data that could lead to incorrect calculations. This is critical for impairment testing, where a misunderstanding about the discount rate calculation or the treatment of corporate overhead can produce materially wrong results.
Notice how the prompt explicitly defines the success criteria: the model must pass audit review on first submission. This shifts the AI’s focus from simply generating numbers to producing a document that is defensible, transparent, and professionally formatted. The reference file provides a concrete example of what “good” looks like, while the “Does NOT sound like” section prevents the AI from falling back on generic templates that would waste your time.
First, read these files completely before responding:
[impairment_model_results.md] — Contains the value-in-use calculations from the model built in the previous prompt, including base-case VIU, carrying amounts, and headroom for each CGU.
[audit_committee_requirements.md] — Contains our board’s specific format preferences: executive summary first, then detailed methodology, then sensitivity analysis, then appendix with all source data.
[prior_year_disclosures.md] — Contains last year’s impairment disclosure notes from the annual report, showing the level of detail and narrative style expected.
Here is a reference for what I want to achieve:
[Upload reference file: “Board_Pack_Impairment_Q2_2025.pdf” converted to markdown showing a completed board presentation with tornado charts, waterfall charts, and narrative explanations for each sensitivity scenario.]
Here’s what makes this reference work:
– Executive summary limited to one page with key numbers and conclusion
– Sensitivity analysis shows impact of WACC changes from -1.5% to +1.5% in 0.25% increments
– Each CGU gets its own section with a “reasonableness check” comparing implied multiples to market peers
– Narrative explains why each assumption is appropriate (e.g., “Terminal growth rate of 2.0% reflects long-term GDP growth in the region, consistent with central bank forecasts”)
– Appendix includes a data dictionary defining every input variable
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Board-ready documentation package in structured markdown, approximately 25-30 pages with charts described in text form, plus an appendix of all assumptions.
Recipient’s reaction: The audit committee should feel confident approving the impairment results without needing to call a special session or request additional work.
Does NOT sound like: A mechanical report that simply states numbers. Every conclusion must be supported by a narrative explanation that pre-empts likely questions (e.g., “Why did you use 2.0% growth when last year was 2.5%?”).
Success means: The audit committee approves the impairment results unanimously at the scheduled meeting, with no follow-up questions.
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 Implementing AI in Your Impairment Workflow
Start with a single, simple CGU to validate the AI’s output before scaling to your full portfolio. Pick a CGU where you already have a manually calculated impairment model—compare the AI’s value-in-use number to yours. If they match within an acceptable tolerance (typically 1-2% due to rounding differences), you have confidence the methodology is correct. If they diverge, investigate the assumptions: the most common errors are inconsistent treatment of lease liabilities, incorrect calculation of the tax shield in WACC, or mismatched growth rate periods.
Second, always include the “Ask clarifying questions first” step in your prompts. This is not optional—it is the single most important guardrail for financial calculations. When the AI asks questions, it reveals what it does not understand about your context. For example, it might ask: “Should the terminal value be calculated using the Gordon Growth Model or the exit multiple approach?” or “How should I treat the corporate center costs that are not allocated to individual CGUs?” These questions force you to clarify your methodology upfront, which prevents errors downstream.
Finally, treat the AI’s output as a first draft that requires human review—not as a final deliverable. The AI is excellent at structure, calculation, and formatting, but it cannot know the qualitative business factors that might affect your impairment conclusion. For example, if you are planning a major restructuring that will change cash flows in year three, the AI will not know this unless you explicitly include it in your input files. Use the AI to handle the heavy mechanical lifting, and reserve your expertise for the judgment calls that only a human CFO can make.
What to Try Next
Once you have mastered the impairment model prompts, extend the same approach to related financial reporting tasks. The same structured prompt framework works for purchase price allocation under IFRS 3, goodwill tracking and allocation, deferred tax asset recoverability analysis, and even going concern assessments under IAS 1. Each of these areas involves the same combination of financial data, accounting standards, and auditor expectations—making them ideal candidates for AI automation.
Start building your library of reference files: your historical financial data, your accounting policy manuals, and examples of prior-year disclosures that received clean audit opinions. The more high-quality reference material you provide, the better the AI’s output will be. Over time, you will develop a set of prompts that can generate a complete impairment analysis in under an hour, freeing your team to focus on strategic analysis rather than spreadsheet mechanics.
Published on 16 July 2026 on growwithgpt.com