Meet David
David is a financial planning and analysis analyst at a manufacturing company with annual revenues of approximately two hundred million pounds, producing precision-engineered components for the automotive, aerospace, and industrial equipment sectors. The company operates three production facilities across two countries, sells into both domestic and export markets, and is privately held by a family ownership group that maintains an active board with high expectations for the quality and clarity of financial reporting. David has been in the FP&A function for five years, having joined from an audit background that gave him strong technical accounting foundations but required him to develop the commercial and narrative skills that financial planning and analysis demands.
David's primary organisational responsibility is the production of the monthly financial reporting pack for the executive leadership team, which consists of the Chief Executive, the Chief Financial Officer, the Chief Operating Officer, the Group Commercial Director, and the heads of the three production facilities. This reporting pack is the primary financial communication between the finance function and the leadership team and is the basis for the performance discussions that take place at the monthly executive review meeting. The quality of the reporting pack, including the accuracy of the underlying analysis, the clarity of the narrative explanations, and the appropriateness of the commentary for an executive audience that includes both financially sophisticated and less financially sophisticated readers, reflects directly on the credibility of the finance function.
In addition to the monthly reporting cycle, David manages the quarterly re-forecast process, which updates the full-year financial projection in the light of actual performance to date and any changes in business assumptions, and he supports the annual budget process that sets the following year's financial targets and resource allocations. He also responds to a regular flow of ad hoc analysis requests from the CFO and other members of the leadership team, which typically require rapid turnaround and a high degree of clarity in presentation given that they are often discussed in leadership forums with limited preparation time.
David's primary pain points before establishing an AI practice are three. First, the time cost of narrative writing. The monthly variance analysis that explains the differences between actual financial results and the budget is technically straightforward once the numbers have been compiled and verified: the variances are visible in the data and the analyst understands what drove them from knowledge of the business. Translating this understanding into clear, precise, and appropriately contextualised written explanations that a senior executive audience can read and act on is a different and substantially more time-consuming task. David currently spends two to three full working days each month on the narrative writing component of the monthly report, a disproportionate allocation of time relative to the analytical value of the narrative compared to the underlying analysis. Second, the loss of historical context between reporting periods. Month-to-month financial performance in a manufacturing business is affected by a range of one-time items, timing differences, and specific business events that should be documented at the time they occur to allow accurate explanation of prior-period variances when they are referenced in subsequent months. Without systematic documentation, the context that was clear in the month of occurrence has often faded by the time it becomes relevant for comparison purposes. Third, the time cost of managing presentation formatting. Board-quality financial presentations involve significant formatting work that is often more time-consuming than the analysis that populates them, and this formatting work consumes analytical capacity that could be more productively deployed on the substance of the financial analysis.
The Before State
Before establishing an AI practice, David's financial reporting documentation reflects the accumulated practice of a busy finance function that has evolved pragmatically rather than by design. The folder structure for financial reporting files has grown around the annual budget cycle: there are year-level folders, but within those folders the organisation of monthly actuals, variance analyses, forecasts, and presentation files varies between periods, reflecting the different naming habits that have developed over the months and the influence of files inherited from a predecessor whose conventions differed from David's own.
The most significant structural problem in David's current documentation is the absence of a systematic record of the context that explains prior-period performance. When the CFO asks in October why a particular cost line behaved unusually in June, David must reconstruct the explanation from memory, from email correspondence that is difficult to search reliably, and from the comments that may or may not have been included in the June reporting pack. This reconstruction is time-consuming and is not always successful: specific items that were clear at the time are sometimes impossible to reconstruct fully from the available documentary record six months later. The consequence is occasionally imprecise or incomplete explanations of historical performance that reduce the credibility of the financial narrative and require additional follow-up.
The narrative writing process is currently entirely manual. David writes variance explanations, executive commentary, and scenario narratives from scratch in each reporting cycle, working from his analytical spreadsheets and his knowledge of the business. The process is thorough but slow, and the quality of the writing varies between periods depending on the time pressure David is under. In months where the variance analysis is complex, the time allocated to narrative writing competes directly with the time required to complete the analytical work, which means that one or both elements receives less attention than the quality standard requires.
The financial model infrastructure is maintained in Excel, which is appropriate for the analytical complexity of the work and will not change as a result of the AI practice. The models are well-structured and accurately reflect the company's financial logic, but they are extensive and the formula logic in some of the more complex sections, particularly the variable cost allocation and the intercompany elimination calculations, requires specialist knowledge of the model's construction to understand and audit correctly.
Knowledge Base Setup
David's knowledge base is built around a financial reporting cycle structure that reflects the temporal organisation of FP&A work. The structure is:
FinancialReporting / Year / Period / Actuals, Budget, Variance, Narratives, Presentations, Archive
The top level is FinancialReporting because all of David's primary deliverable work relates to this function. The year and period levels reflect the fundamental reporting calendar of the organisation: annual at the top, with the period level representing each reporting month or quarter depending on the reporting cycle. Within each period folder, five document type subfolders address the distinct categories of material produced in each reporting cycle.
Actuals holds the data files produced from ERP exports: the trial balance extracts, the management accounts data, and any supporting data from operational systems that feeds into the financial analysis. Files in this folder are the source data for all analytical work and are treated as read-only reference material: the files exported from the ERP are not modified, and analytical calculations are performed in separate workbooks that reference the Actuals files. This separation between source data and analytical workbooks is a fundamental principle of financial data governance that ensures the integrity of the source data record.
Budget holds the approved budget files for the period, including the full-year budget broken down by month, any budget amendments approved during the year, and the working files from the budget-setting process that may be needed for reference when explaining budget variances.
Variance holds the variance analysis workbooks: the calculations of actual versus budget differences by revenue line, cost line, and business unit, together with the supporting analyses that decompose material variances into their component drivers. These workbooks are the primary analytical work product of the monthly reporting cycle and are the foundation for the narrative explanations that accompany the executive reporting pack.
Narratives holds the written narrative documents: the variance commentary, the executive summary of period performance, the CFO briefing notes, and any other written analytical documents produced in the reporting cycle. These are maintained as separate documents rather than embedded within the spreadsheet workbooks because they are subject to review and revision cycles that are distinct from the spreadsheet analysis, and their management as standalone documents makes them more accessible to AI-assisted drafting tools.
Presentations holds the executive and board presentation files, including both the working drafts produced during the preparation cycle and the final version submitted and presented. Version control within this folder distinguishes drafts by version number and date, using the naming convention Period_ReportType_Version_Date.
Archive holds superseded versions, preliminary drafts, and supporting materials from previous reporting cycles that should be retained for the period record but are no longer active working documents.
David maintains four context documents that serve his AI practice across all reporting cycles.
The Business Model Overview Document is the foundational context document for all AI-assisted financial analysis work. It describes the revenue model of the business: the product lines and their revenue characteristics, the key volume and price drivers of revenue in each product segment, the seasonal patterns that affect revenue and production volumes across the calendar year, and the key external factors that influence the business's financial performance. It also describes the cost structure: the distinction between variable costs that move with production volumes and fixed costs that do not, the major cost categories and their primary drivers, the overhead allocation methodology, and the capital expenditure cycle. This document allows AI tools to interpret financial data in the context of the actual business dynamics that drive it, rather than treating the numbers as a decontextualised set of figures.
The Monthly Financial Narrative Log is the context document that addresses the historical context problem identified in the before state. It is updated each month as a running record of the significant variance drivers, one-time items, and specific business events that affected the period's financial results. Each monthly entry is brief: a structured record of the material variances and their explanations, noting any items that are one-time in nature and should not be treated as indicative of the run-rate business performance. This log becomes progressively more valuable as it accumulates a documented history of the business's financial performance, and it is the primary resource for answering questions about prior-period performance accurately and efficiently.
The Executive Reporting Preferences Document records the communication preferences of the key members of the executive audience. It notes the CFO's preference for direct, quantified explanations rather than qualitative commentary; the CEO's preference for performance narrative framed in terms of strategic implications rather than technical accounting explanations; the COO's focus on operational cost performance and efficiency metrics; and the preferences of the facility heads regarding how their facilities' performance is presented. This document allows AI-assisted narrative drafts to be calibrated to the actual preferences of the audience rather than to a generic executive communication style.
The Historical Context Notes Document records the specific accounting, commercial, and operational events that affect the comparability of current period results with prior periods: acquisitions, disposals, restructuring events, changes in accounting policy, customer gains and losses, commodity price shocks, and any other material developments that should be understood when interpreting year-on-year or period-on-period comparisons. Unlike the monthly narrative log, which records the routine explanation of period variances, the historical context notes address the less frequent but more significant events that affect comparability over longer periods and that are most likely to be misunderstood if they are not specifically documented.
Model and Tool Selection
David's primary AI tool is Microsoft Copilot, selected on the basis of the task type matching framework from Module 4.2. His most demanding AI-assisted tasks involve drafting extended analytical narratives synthesizing financial data, business context, and executive communication preferences. Copilot provides deep, native integration across the Microsoft 365 ecosystem. This integration allows David to bridge quantitative data in Excel directly into coherent, professional written explanations in Word and PowerPoint securely within the company's enterprise boundary. Copilot's capability to process structured numerical data alongside corporate context documents makes it highly appropriate for this intensive reporting workflow.
Within Excel, Copilot serves as an embedded assistant for formula construction and data synthesis, as addressed in Section 4 of Module 4.3. David deploys Copilot when working with unfamiliar functions or when constructing complex formulas requiring multiple conditions. He deliberately maintains full manual control over the core financial model logic. The proprietary business logic embedded in the company's financial models demands human expertise for accurate construction and strict verification. Text-based AI interactions with spreadsheet content cannot reliably map the intricate formula structures and cell reference dependencies of a complex multi-sheet model.
David's organisation enforces strict security restrictions on Enterprise Resource Planning (ERP) system access by integrated AI tools. The ERP system serves as the authoritative financial record and remains completely isolated from direct AI processing. This isolation reflects the deliberate financial control framework described in Section 5 of Module 4.3. All data submitted to Copilot in David's practice originates exclusively from Excel exports of ERP data. David reviews these exports for absolute completeness and accuracy prior to any AI interaction.
David utilizes Copilot within PowerPoint to generate draft slide content for executive and board presentations. Copilot applies AI assistance specifically to the textual content of the slides. The generated text, calibrated to the specific context and audience using the executive reporting preferences document, provides a structured starting point for the presentation narrative. David subsequently edits the narrative and manually executes all visual design, formatting, and layout adjustments to meet the organisation's strict board presentation standards.
Workflow One: Monthly Variance Analysis Report
The monthly variance analysis report is David's primary deliverable. This workflow benefits significantly from AI assistance. The process moves from data compilation through analytical calculation to narrative drafting. Copilot assistance is concentrated specifically within the narrative drafting phase.
Step one: Export and validate actuals data. David exports the period actuals from the ERP system directly into Excel using the standard management accounts configuration. He validates the export for absolute completeness. He confirms the trial balance totals agree, all posting periods are reflected, and all manual adjustments approved by the financial controller are incorporated. He saves the validated export file to the Actuals folder for the period. Copilot does not possess access to the ERP system or the export process. This step remains entirely manual.
Step two: Calculate variances and identify material differences. Working within the variance analysis Excel workbook, David links the actuals data to the corresponding budget figures. He calculates the variances by revenue line, gross margin component, overhead category, and business unit. He applies the organisation's materiality threshold to isolate the specific variances requiring narrative explanation. He then applies a secondary assessment evaluating the directionality of each variance. This determines its favourable or unfavourable status compared to the budget and its trajectory compared to the prior year and period. David performs this core analytical work manually. The output consists of a structured variance table serving as the direct input for the Copilot-assisted narrative drafting.
Step three: Draft variance explanations with AI assistance. David prepares the prompt for variance narrative drafting within Word or Excel using Copilot. He highlights the structured variance table and references the relevant sections of the business model overview document, the recent entries from the monthly financial narrative log, and the executive reporting preferences document. He constructs a prompt supplying this context material to request a narrative explanation of the specifically identified material variances. A representative prompt structure includes: "Draft a variance narrative for the selected period results table. Business context: [reference business model overview]. Prior period context: [reference monthly narrative log]. Focus the narrative on the revenue shortfall in Product Line B and the overhead cost overrun in Facility 2. The audience is the executive leadership team. The CFO requires direct, quantified explanations. Make the narrative factual and specific. Identify the primary driver of each material variance and quantify its contribution to the overall variance." This high level of prompt specificity significantly improves the quality of the Copilot draft.
Step four: Verify every figure in the AI narrative against source data. David reviews the Copilot-generated variance narrative figure by figure against the source variance analysis workbook. He confirms every monetary variance cited in the narrative directly against the workbook calculation. He checks every percentage change for arithmetic accuracy. He confirms every characterisation of a variance as favourable or unfavourable against the strict directionality of the actual figure. This verification step represents a non-negotiable, separately budgeted activity. The narrative enters the reporting pack only after every single figure achieves confirmed accuracy.
Step five: Add business context that AI cannot derive from the numbers. The verified draft is reviewed for the business context that the AI tool cannot supply from the financial data and context documents alone: the specific operational developments that occurred during the period and explain the variances, the commercial intelligence about customer behaviour that David has obtained through conversations with the sales team, the production issues that the COO is already aware of and that provide the causal explanation for the cost overruns visible in the data. These contextual additions are the component of the narrative that most directly reflects David's value as an analyst: the ability to connect the financial data to the operational reality of the business in a way that makes the numbers meaningful rather than merely accurate.
Step six: Anticipate executive questions and prepare talking points. David uses Copilot to anticipate the questions the CFO and the executive team will likely raise during the variance analysis review. A representative prompt is: "Based on the variance narrative and the results described, what are the three to five questions a commercially minded CFO expecting direct, data-driven answers would most likely ask at the executive review meeting?" The Copilot output provides a structured list of probable questions. David reviews and supplements this list with questions anticipated from his personal knowledge of the CFO's current concerns and the strategic business context. David prepares a concise response note for each probable question. This provides the factual basis for an immediate answer and ensures precise responses to executive inquiries during the meeting.
Step seven: Final review and assembly of the reporting pack. The completed narrative, the variance tables, the supporting exhibits, and the talking points are assembled into the reporting pack format. David reviews the assembled pack as a complete document before submitting it to the CFO for final review, checking for internal consistency across sections and for the overall narrative coherence of the pack as a communication of the period's performance.
Workflow Two: Building and Communicating Forecast Scenarios
Scenario analysis is a regular component of the FP&A function's work, and the communication of scenario results to an executive audience that needs to understand the financial implications of alternative business assumptions is one of the tasks where the gap between analytical capability and communication clarity is most often felt. The workflow addresses how AI assistance supports the scenario communication task without substituting for the financial modelling work that generates the scenarios.
Step one: Clarify the scenario request and its parameters. When an executive requests a scenario analysis, David confirms the specific assumptions to be modelled, the financial dimensions to be analysed, the time horizon relevant to the analysis, and the format in which the results should be presented. Clarity about the scenario parameters before beginning the modelling work prevents the waste of completing a scenario analysis that does not address the actual question the executive had in mind.
Step two: Update the forecast model with the new assumptions. David updates the forecast model in Excel to reflect the scenario assumptions, working through the model's assumption cells and verifying that the downstream calculations correctly reflect the new assumptions throughout. The model logic, including the relationships between revenue assumptions and variable cost calculations, the timing of headcount reductions relative to the period in which the savings are realised, and the interaction between volume assumptions and the overhead absorption calculation, is verified by David through direct inspection of the formula structure. This verification is performed manually: the formula logic of the financial model is not submitted to AI review for the reasons addressed in Section 4 of Module 4.3.
Step three: Use AI for formula assistance on specific technical questions. When the scenario analysis requires unfamiliar formulas or targeted technical checks on specific calculation logic, David utilizes Copilot within Excel for bounded assistance. These requests remain highly specific and strictly scoped. A prompt asking whether a specific formula correctly accounts for the timing of headcount reductions given the model's period structure provides sufficient specificity for Copilot to generate a useful response. Broad requests asking the AI to review the overall model logic fail to produce reliable results due to the structural limitations addressed in Module 4.3.
Step four: Draft the scenario comparison narrative with AI assistance. Using the finalised scenario results in Excel alongside the business model overview context document, David prompts Copilot within Word or PowerPoint to draft a structured comparison narrative. A representative prompt structure includes: "Draft a scenario comparison narrative for an executive audience. Base case: [reference selected data table]. Scenario A: revenue decline of ten percent assuming zero cost interventions. Key metrics: [reference data]. Scenario B: revenue decline of ten percent with the outlined cost reduction actions. Key metrics: [reference data]. Business context: [reference business model overview document]. Explain the financial implications of each scenario in clear, non-technical language. Identify the key differences between the scenarios and their primary drivers. Present the analysis in a format suitable for a ten-minute executive discussion." This highly structured input enables Copilot to generate a scenario narrative securely grounded in the specific analytical figures and appropriately framed for the executive audience.
Step five: Add the strategic implications that AI cannot derive from the financial data. The AI scenario narrative will describe the financial outcomes of the scenarios accurately but will not derive the strategic implications that require business judgment beyond the financial data. The assessment of whether a ten percent revenue decline is recoverable within the planning horizon, whether the cost actions in Scenario B are operationally feasible given the current production environment, and what the scenarios imply for the organisation's investment and financing plans are judgments that David adds to the AI draft from his knowledge of the business and his understanding of the strategic context in which the analysis is being presented.
Workflow Three: Responding to Ad Hoc Analysis Requests
Ad hoc analysis requests represent a significant and unpredictable component of David's working week. The requests typically arrive with short turnaround requirements from the CFO or the CEO, they address specific questions about financial performance that require targeted analysis rather than the comprehensive reporting coverage of the monthly pack, and the quality of the response is assessed as much on clarity and concision as on analytical depth.
Step one: Clarify the question and the required output. Before pulling data, David confirms the specific question being asked, the level of analytical depth required, the format in which the response should be delivered, and the timeframe within which it is needed. Many ad hoc requests that appear straightforward on initial reading contain embedded ambiguities about scope or methodology that, if not resolved before the analysis begins, result in a completed analysis that does not address what the requestor actually needed.
Step two: Pull and validate the relevant data. David identifies the data required to address the specific question, exports it from the relevant source systems, and validates it for completeness and accuracy before using it as the basis for analysis. Where the request requires data from multiple sources, the consistency of the data across sources is checked before analysis proceeds.
Step three: Decompose the variance with AI assistance. For requests involving the decomposition of a financial variance into its component drivers, David highlights the relevant data and submits a structured decomposition prompt to Copilot within Excel. A representative prompt for a gross margin variance decomposition includes: "Decompose this gross margin variance into its component effects. Data provided: [reference selected data table]. Separately identify the volume effect, the price effect, the product mix effect, and the input cost effect. Present the decomposition in a table. Explain in plain language the meaning of each effect and quantify its contribution to the overall variance." This structured decomposition prompt generates a structured analytical output to support David's subsequent manual review.
Step four: Verify all AI arithmetic. AI tools can make arithmetic errors, particularly in decomposition calculations that involve multiple steps and intermediate calculations. Every figure in the AI decomposition output is verified against the source data and checked for internal consistency: the component effects should sum to the total variance, and each component effect should be derivable from the source data by a calculation that David can trace and confirm. Where errors are identified, they are corrected in the analysis before the response is drafted.
Step five: Draft the executive response with AI assistance. Using the rigorously verified analytical output, David prompts Copilot within Word or Outlook to produce a concise executive response calibrated specifically to the requesting audience. A representative prompt includes: "Draft a concise response to the following question from the CEO: [insert question]. The analysis shows: [reference verified decomposition]. Maintain a direct and data-driven tone. Lead with the direct answer to the question followed immediately by the supporting analysis. Ensure readability under two minutes. Exclude technical accounting language. Use specific numerical figures throughout the text." David reviews the resulting draft to confirm factual accuracy, appropriate executive language, and narrative completeness before transmission.
Quality Control Checklist
David applies a quality control checklist to all AI-assisted work before it is incorporated into the reporting pack, presented to the CFO, or submitted in response to an executive request.
Does every figure in the AI narrative match the figures in the source analytical workbook? Financial communications to executive and board audiences are relied upon for decision-making, and a figure that does not match the analytical source undermines the credibility of the entire communication. Every monetary figure, every percentage variance, and every directional characterisation in an AI narrative is verified against the source data in the analytical workbook before the narrative is used.
Has the AI correctly identified favourable and unfavourable variances throughout? The directionality of a variance, whether it represents better or worse performance than the benchmark, affects the entire interpretation of the financial results. A narrative that characterises an unfavourable variance as favourable, even if the figures cited are arithmetically correct, will mislead the executive audience in a way that could affect their assessment of period performance. Variance directionality is checked systematically throughout every AI narrative before submission.
Are the scenario assumptions mathematically consistent throughout the scenario model and the narrative? Scenario analyses involve multiple assumptions that interact with each other through the financial model's logic. An AI narrative based on scenario outputs should be consistent with those outputs throughout: the revenue figure cited in the narrative should match the model output, the cost implications cited should reflect the model's cost structure, and the bottom-line impact should follow arithmetically from the revenue and cost figures presented. Where the narrative introduces any figures not directly sourced from the model output, those figures are verified before the narrative is used.
Does the narrative make sense to a reader who is not immersed in the data? Financial narratives are produced by analysts who have been working closely with the underlying data and who are therefore at risk of inadvertently assuming knowledge that the executive reader does not have. Before finalising any executive narrative, David reads it with the perspective of a reader who is encountering the period's financial results for the first time, checking whether the narrative is self-contained and comprehensible without reference to the underlying spreadsheets.
Would this analysis withstand a detailed challenge from the CFO? This is the overarching quality standard that subsumes the specific checks above. The CFO's review of financial reporting is a rigorous examination of both analytical accuracy and analytical reasoning. An analysis that could not withstand this review should not be submitted, regardless of how well it reads at a surface level.
The After State
After twelve weeks of consistent AI practice, the most significant change in David's work is the reduction in narrative writing time. The combination of current context documents and structured drafting prompts has reduced the time required to produce the first draft of the monthly variance narrative from the two to three days previously required to approximately half a day of drafting and verification. The total time invested in narrative work per month, including the drafting, verification, business context addition, and editing steps, is approximately one full day, representing a recovery of one to two days per month from the narrative writing component of the reporting cycle.
This recovered time has been reinvested in two areas. First, in the depth of the analytical work underlying the variance analysis: David now has capacity to investigate the drivers of material variances more thoroughly before the reporting pack is assembled, which has improved the quality of the explanations in the pack and reduced the number of follow-up questions requiring additional analysis after the executive review meeting. Second, in the response quality for ad hoc analysis requests: the improved availability of time for analytical work means that ad hoc requests are addressed with more analytical depth than was possible when the narrative writing cycle consumed most of the working days around the period-close.
The financial models remain David's manual domain. AI assistance has not changed the approach to model construction and audit: these activities require the professional accounting judgment and model-specific knowledge that only the analyst can provide, and the data governance framework of the organisation rightly restricts AI tool access to the ERP data that underlies the models.
Common Mistakes for Financial Analysts
Incorporating AI-produced figures without verifying them against the analytical source. Financial communications are precise documents in which figures carry specific meaning and are subject to scrutiny by audiences who are capable of identifying discrepancies. An AI tool producing a variance narrative from a data table can make arithmetic errors, can misread the directionality of a variance, or can cite a figure from one line of the table when the context requires a figure from a different line. None of these errors are detectable from the AI output alone: they require comparison with the source data. A financial communication that reaches the CFO or the board containing figures that do not match the analytical source is a professional failure that undermines the credibility of the finance function, and no AI efficiency gain justifies the reduction in verification rigour that allows such errors to occur.
Failing to maintain the monthly financial narrative log. The monthly narrative log is the context document that makes AI-assisted financial narrative genuinely useful over time rather than merely efficient in the current period. Its value accumulates as it builds a documented history of the business's financial performance that can be referenced accurately when explaining prior-period comparisons. When this log is not maintained consistently, the AI tool working on a variance narrative for the current period cannot accurately contextualise the current results against prior periods, and the narrative it produces will treat each period in isolation rather than as part of an ongoing financial story. The fifteen-minute weekly maintenance habit from Module 4.1, applied to the narrative log at the close of each reporting period, is the investment that preserves the cumulative value of this context document.
Assuming AI understands company-specific accounting policies, chart of accounts structure, or cost allocation methodology. Every organisation of any complexity develops accounting policies, cost allocation approaches, and reporting conventions that are specific to its business model and its history. The way revenue is recognised across product lines, the methodology for allocating overhead to production facilities, the treatment of intercompany transactions, and the specific definitions of the metrics used in management reporting are all properties of the individual organisation that AI tools trained on general financial text do not know. A prompt asking an AI tool to explain a gross margin variance without providing the business model overview and cost structure context will produce a narrative based on generic assumptions about how gross margins work in manufacturing businesses generally, not on the specific cost structure of this organisation. The business model overview context document exists precisely to address this gap, but it must be kept current and must be provided in the prompt for every AI-assisted analysis task.
Presenting AI-produced scenario narratives without adding the strategic interpretation. Scenario analysis is valuable to executive audiences not primarily because it calculates the financial outcomes of different assumptions, which the executives could derive themselves from the model outputs, but because it provides an analytical interpretation of what those outcomes mean for the strategic choices facing the business. An AI tool can describe what a ten percent revenue decline does to operating profit under different cost action assumptions. It cannot assess whether the cost actions are operationally feasible, whether the revenue decline scenario is consistent with the current market intelligence, or whether the analysis changes the strategic priorities the leadership team should be focused on. These interpretive judgments are the analyst's contribution to the executive discussion, and a scenario narrative that omits them in favour of an AI-produced financial description alone adds less value than the analysis the executive team needs.