AI vs. the 10-K: Part 2 of How to Conduct a Forensic Examination of Company Accounts
Automate Quarterly and Annual Report Reviews without Missing the Nuance
This is the second installment in a two-part series on automating annual and quarterly earnings report reviews with rigorous, forensic discipline. Part I (click here) covered the backbone of the statements: accruals quality, cash conversion, revenue recognition, capitalization choices, working-capital stress tests, leverage and covenants, and off-balance-sheet exposures. In Part II we move from core accounting to signals and governance: compensation and dilution, segment economics, auditor messages, narrative shifts, related-party transactions, reserves, and the bridge from guidance to economics. Throughout, we build techniques based on the classic security analysis texts: Benjamin Graham’s quality of earnings and margin of safety, Richard Penman’s clean-surplus intuition, Koller and colleagues’ return on invested capital, White–Sondhi–Fried’s statement linkages, and Schilit and Perler’s pattern recognition. Each section gives you condensed copy-paste prompts that keep artificial intelligence focused on facts, calculations, and testable hypotheses.
The full single-pass, all steps, Master Prompt to conduct a comprehensive forensic examination of a company will be released in The Inferential Investor’s professional prompt library. Subscribe to ensure you don’t miss its release.
8) Stock-Based Compensation and Dilution: Counting the Real Cost of Talent
Why this matters
Stock-based compensation is either the cheapest growth capital a company can access or a slow leak in owner value. It never shows up as a cash outflow today, but it shows up as a claim on tomorrow’s earnings via share count expansion. Benjamin Graham warned against “paper profits” that are not anchored in owner earnings; here the “paper” is literal - new stock certificates. If reported free cash flow looks strong but fully diluted shares creep higher every year, the economic pie is not growing as fast as the headline suggests.
Goal
Measure the total economic cost of equity compensation, the trajectory of dilution, and the company’s policy for offsetting issuance with repurchases. Tie the numbers to incentives: what behaviors are being rewarded?
Forensic checklist
Track stock-based compensation expense relative to revenue and gross profit for at least three years.
Reconcile basic and diluted share counts; account for options, restricted stock units, and performance awards.
Compare buyback spend with new shares issued to see whether management offsets or amplifies dilution.
Note cash tax benefits from option exercises that can flatter cash flow from operations.
Read the compensation discussion to understand performance metrics (revenue growth, free cash flow margin, total shareholder return) and vesting cliffs.
Copy-paste prompt (ready to use, attach accounts)
You are a forensic equity analyst focusing on stock-based compensation (SBC) and dilution.
Inputs: 10-K (or annual report) sections on SBC expense, share count reconciliation, equity plans, and repurchases for the last 3 fiscal years.
Tasks:
1) Extract:
   - SBC expense by year.
   - Basic weighted-average shares, diluted weighted-average shares, and ending shares outstanding by year.
   - Number of options, RSUs, PSUs outstanding and their weighted-average exercise price (if applicable).
   - Share repurchases (shares and dollars) by year and average repurchase price.
2) Compute:
   - SBC as a percentage of revenue and as a percentage of gross profit.
   - Dilution rate = (Diluted shares – Basic shares) / Basic shares each year.
   - Net share change = Ending shares outstanding (current year – prior year).
   - “Offset ratio” = Shares repurchased / Shares issued (including SBC settlement), indicate >100% (offsetting) or <100% (dilutive).
   - Estimated buyback dollars required to fully offset next year’s expected SBC issuance (assume current share price = [USER_INPUT_PRICE]).
3) Diagnose and explain:
   - Is SBC trending higher, stable, or lower relative to revenue?
   - Is management fully offsetting dilution? Provide evidence.
   - Identify any cash tax benefits from option exercises that affect cash flow from operations.
Output:
- A labeled table with all extracted figures and computed metrics for 3 years.
- A 150–200 word conclusion on sustainability of compensation practices and expected owner dilution next year. Avoid recommendations; cite exact note headings and page numbers.Illustrative example
High-growth software firms often run stock-based compensation at eight to fifteen percent of revenue in early scale. Some also repurchase heavily to “neutralize” dilution. The optics can mislead: a large buyback funded with operating cash flow creates the appearance of generous free cash flow while silently recycling it into employee compensation. Your job is to show the net change in fully diluted shares and the offset ratio so decision-makers see the true economic trade-off.
9) Segment and Geographic Economics: Where Value Is Actually Made
Why this matters
Consolidated totals hide capital allocation quality. Graham and Dodd taught analysts to separate a business into parts and reassemble the whole only after assessing the parts’ earning power. Segment disclosure is the closest we get to that x-ray inside the earnings report. A company can be a laggard in one segment and a star in another; the mix shift often determines valuation far more than the consolidated growth rate.
Goal
Quantify the economics of each segment including growth, margin, capital intensity and determine whether the consolidated trend is driven by mix or by within-segment improvement. Track restatements or re-definitions that muddy comparability.
Forensic checklist
Extract revenue, operating income, and assets by segment for at least three years.
Compute growth, margin, and implied return on assets or return on invested capital where disclosures allow.
Build a “mix bridge” that shows how consolidated margin changes split into mix shift versus within-segment changes.
Track any restatements of prior periods for comparability and any re-labeling of segments that may camouflage softness.
Map geography: exposure to currencies, sanctions, local pricing power.
Copy-paste prompt (ready to use, attach accounts)
You are a forensic analyst performing segment and geographic economics analysis.
Inputs: 10-K segment tables (revenue, operating income, assets), geographic breakdowns, and any segment restatement notes for the last 3 fiscal years.
Tasks:
1) Extract by segment and year: revenue, operating income (or segment profit), and segment assets (if disclosed).
2) Compute by segment:
   - Revenue compound annual growth rate (CAGR).
   - Operating margin = Operating income / Revenue.
   - Implied return on segment assets = Operating income / Segment assets (if assets disclosed).
3) Build a margin bridge:
   - Decompose the change in consolidated operating margin into (a) mix effect and (b) within-segment margin change.
   - Show contribution of each segment to the consolidated change (basis points and percent).
4) Geography:
   - Extract revenue by region and note currency exposures.
   - Flag concentration risks (any region >30% revenue or profit).
5) Restatements and re-definitions:
   - Identify any segment definition changes year-over-year and summarize management’s rationale and impact.
Output:
- A compact table by segment (3-year history) with growth, margin, implied returns.
- A one-paragraph explanation of what drove consolidated margin changes (mix vs. improvement).
- A bullet list of geographic concentration and currency risk items with page cites.Illustrative example
Consumer electronics businesses have watched services expand as a share of revenue and profit. Even with modest top-line growth, the consolidated margin rose because the services segment carries structurally higher margins and low capital intensity. An artificial intelligence routine that produces a clear margin bridge lets humans focus on cleaner insights rather than gather and process numbers.
10) Auditor Signals: Opinions, Critical Audit Matters, and Restatements
Why this matters
The auditor’s report is short, but the density of signal is high. Clean opinions with recurring critical audit matters on revenue estimation or impairment tell you where judgment is concentrated. Restatements do not always mean malfeasance; they often reveal operational complexity that deserves extra margin of safety.
Goal
Harvest the auditor’s explicit and implicit risk markers and set a monitoring agenda that matches judgmental hot spots.
Forensic checklist
Record the opinion type (unqualified, qualified, adverse) and any going-concern language.
Catalog critical audit matters, summarize the accounting estimates involved, and note whether they repeat year-to-year.
Identify any restatements or revisions, the reasons, and the quantitative impact.
Cross-reference critical audit matters to your analytical work plan (for example, if revenue estimation is a critical audit matter, prioritize receivables and deferred revenue diagnostics).
Copy-paste prompt (ready to use, attach accounts)
You are analyzing auditor signals, critical audit matters (CAMs), and restatements.
Inputs: Auditor report, CAM section, and any restatement/revision notes for the last 3 fiscal years.
Tasks:
1) Extract the opinion type and whether there is going-concern language.
2) For each CAM:
   - Summarize the estimate/judgment area (e.g., revenue estimation, allowance, impairment).
   - Quote the auditor's description (≤40 words) with page reference.
   - Indicate whether the CAM is new or recurring.
3) Restatements/revisions:
   - List restated items, the reason, and the quantified effect by period.
4) Map to monitoring plan:
   - Link each CAM to the forensic modules in this series (e.g., revenue CAM → revenue recognition test; impairment CAM → capitalization and impairment test).
Output:
- A table of CAMs with status (new/recurring) and citations.
- A bullet list of restatements with magnitude.
- A 120–180 word note on how these auditor signals should change analyst attention and thresholds for concern.
Illustrative example
A recurring critical audit matter on revenue tied to complex multi-element arrangements does not prohibit investment. However, it does tell you exactly where to focus your sensitivity analysis and what disclosures to track for small but compounding changes.
11) Management Discussion and Analysis Language: Narrative-Numbers Consistency
Why this matters
Words are soft data that precede hard data. Management often foreshadows demand, pricing, and competitive dynamics in the Management Discussion and Analysis before it fully shows up in ratios. Penman’s advice is to reconcile the numbers to the narrative.
Goal
Run a redline between the current Management Discussion and Analysis and the prior year’s. Detect added and removed phrases in sections about demand, pricing, supply chain, competition, capital allocation, and risk. Verify claims against the statements.
Forensic checklist
Redline changes; extract new and removed phrases by topic.
Map assertions (“strong cash generation,” “pricing power”) to cash flow from operations, free cash flow, margins, and working-capital ratios.
Track the tenor of risk factors: are new risks elevated, or old ones softened?
Copy-paste prompt (ready to use, attach accounts)
You are analyzing Management Discussion and Analysis (MD&A) language for narrative-number consistency.
Inputs: Current-year MD&A and prior-year MD&A; risk factors section for both years.
Tasks:
1) Redline comparison:
   - List added phrases and removed phrases in sections on demand, pricing, supply chain, competition, capital allocation, and liquidity.
2) Evidence check:
   - For each management claim (e.g., "improved cash generation"), extract and report the relevant metric (CFO, FCF, margin, DSO/DIO/DPO) and whether it confirms or contradicts the claim.
3) Risk factor evolution:
   - Identify newly introduced risks and risks that were removed or de-emphasized.
4) Output a "narrative change log":
   - 8–12 bullets, each with: (topic, change summary, metric corroboration, page cites).
Constraints:
- Keep each bullet ≤35 words.
- No adjectives without a specific metric attached.
Illustrative example
During supply chain normalization, many industrials quietly reduced references to logistics constraints before the cash conversion cycle fully reverted. Tracking removals in phrasing and validating against days payable and inventory turns, helped distinguish real easing from an optimistic tone.
12) Related-Party Transactions and Governance: Terms Tell the Truth
Why this matters
Conflicts of interest live in the detail that is often overlooked. Related-party services, leases, loans, and guarantees can look immaterial at first glance yet determine who gets paid first when the cycle turns or whether shareholders or management are being prioritized by the company. Graham’s “character of management” has to be inferred from incentives and related-party footprints.
Goal
Enumerate related-party dealings, quantify them, and judge whether terms appear arm’s length. Tie board composition, incentive metrics, and shareholder rights to the financial risks you have identified elsewhere.
Forensic checklist
Catalog each related-party transaction: counterparty, nature, amount, pricing basis.
Track trend and concentration; identify any related-party share of cost of goods or operating expenses.
Note board independence, dual-class share structures, change-of-control protections, and poison pills.
Map executive compensation metrics to reported vulnerabilities (for instance, revenue-only incentives where revenue recognition is judgmental).
Copy-paste prompt (ready to use, attach accounts)
You are reviewing related-party transactions and governance signals.
Inputs: Related-party note; corporate governance and compensation disclosures; shareholder rights summary.
Tasks:
1) List all related-party transactions (counterparty, nature, amounts for 3 years, pricing basis if disclosed).
2) Identify trends and concentration (any item >5% of COGS or Opex).
3) Governance snapshot:
   - Board independence ratio, dual-class structure (if any), key shareholder rights (poison pill, special meeting thresholds).
   - Executive compensation metrics (revenue growth, FCF, ROIC, TSR).
4) Risk assessment:
   - Flag any transactions that could impair cash generation or skew reported margins.
   - Link incentive metrics to potential accounting judgment areas.
Output:
- A table of related-party transactions with amounts and terms.
- A 120–180 word governance note highlighting any conflicts and monitoring items, with page cites.
Illustrative example
In some roll-ups, the operating company leased headquarters or facilities from an outside entity controlled by executives. The rent looked modest annually but compounded to a material share of operating costs across the cycle. Identifying these arrangements sheds light on management incentives and prevents the density of accounts from hiding economic reality.
13) Provisions, Contingencies, and Legal Matters: Where Smoothing and Shocks Hide
Why this matters
Reserves are the analyst’s paradox. Under-reserved positions convert to sudden hits; over-reserved balances become “cookie jars” that smooth future earnings. Schilit and Perler dedicate an entire category of shenanigans to timing and reversals. Your task is to check whether the underlying earnings reality and the story lines up.
Goal
Evaluate reserve adequacy and transparency. Link reserve levels to the activity bases they are supposed to cover (allowance to receivables, warranty to product failure rates, litigation to exposure ranges). Track reversals.
Forensic checklist
Extract roll-forwards of allowances and provisions with openings, charges, utilizations, and reversals.
Compute ratios such as allowance to receivables and warranty provision to relevant sales.
Read legal contingencies for range disclosures; note any “unable to estimate” language and whether it repeats.
Tie changes in reserve ratios to changes in underlying activity or quality metrics.
Copy-paste prompt (ready to use, attach accounts)
You are evaluating provisions, contingencies, and reserve adequacy.
Inputs: Notes on allowance for credit losses, warranty reserves, returns/refunds, and legal contingencies (3-year history).
Tasks:
1) Extract roll-forwards: beginning balance, additions, utilizations, reversals, ending balance for each reserve category.
2) Compute ratios:
   - Allowance / Accounts Receivable.
   - Warranty reserve / Relevant product sales.
   - Returns reserve / Revenue for affected lines.
3) Identify:
   - Any significant reversals and management’s stated rationale.
   - Legal matters with disclosed ranges or repeated “cannot estimate” language.
4) Assess whether reserve changes track underlying drivers (e.g., higher chargebacks, defect rates).
Output:
- A table of roll-forwards and ratios for 3 years.
- A 150–200 word assessment of reserve adequacy and transparency, with page cites and specific follow-up questions.
Illustrative example
Consumer lenders that pivoted into riskier customer cohorts sometimes let allowance ratios lag while receivables surged. The best analysts can spot the divergence and its associated risks, early by pairing the provision changes with originations mix.
14) Guidance, Consensus, and the Economics Bridge: What Must Be True
Why this matters
Valuation is about tomorrow’s cash flows. Management guidance and sell-side consensus are hypotheses about those cash flows. The forensic task is to translate guidance into unit drivers (eg price, volume, mix, cost) and test whether the required balance sheet conditions are in place to achieve that guidance or whether there are impediments.
Goal
Build a driver-based bridge from last year’s results to guided ranges. List the two or three “must-be-true” conditions and the fastest falsification tests.
Forensic checklist
Extract guidance ranges for revenue, margins, capital expenditure, and cash conversion; note any key performance indicators disclosed.
Create a bridge that decomposes changes into price, volume, mix, cost, and operating expense.
Translate the bridge into working-capital and capital-expenditure implications.
Identify early-warning ratios (for example, days sales outstanding must fall by a set number of days to hit cash flow targets).
Copy-paste prompt (ready to use, attach accounts)
You are building a guidance-to-economics bridge.
Inputs: Guidance ranges and key performance indicators from MD&A or earnings release; last fiscal year’s financials.
Tasks:
1) Extract guidance ranges for revenue, gross margin, operating margin, capital expenditure, cash flow, and any disclosed KPIs.
2) Build a bridge from last year to the guidance midpoint:
   - Attribute changes to price, volume, mix, cost of goods, operating expense (state assumptions if not disclosed).
3) Working capital and capex implications:
   - Estimate the change in DSO, DIO, DPO, and capex needed to support the bridge.
4) “Must-be-true” list:
   - Provide 3 conditions that must occur for the midpoint to be achieved and the quickest falsification checks (metrics and thresholds).
Output:
- A clear bridge table and a bullet list of must-be-true conditions with monitoring metrics.
- Keep assumptions explicit; avoid recommendations.
Illustrative example
A device maker guiding to higher margin without changing pricing usually implies a mix shift toward services or premium products and/or a reduction in input costs. If working capital ratios do not improve and services mix does not rise, the thesis is off-track. The point is not necessarily to predict, but to make the prediction testable and identify signals which suggest the company may be off track.
Synthesizing the Signals
At this stage the artificial intelligence system has produced a fourteen tightly scoped memos with tables, page-anchored citations and helpful interpretations. The analyst’s value is in synthesis and materiality:
Reconcile contradictions. Strong free cash flow with rising days sales outstanding might be explained by deferred revenue. A margin expansion with rising stock-based compensation might be a wash at the owner level.
Prioritize by value impact. A small lease nuance rarely moves intrinsic value. A change in revenue recognition for the largest segment often does.
Establish thresholds. Define in advance what would count as confirmation or refutation of the company’s guidance (for example, “if receivables grow faster than revenue for two consecutive quarters and days sales outstanding rises by more than ten days, reduce confidence”).
The classics all converge here. Benjamin Graham’s margin of safety is a discipline about what can go wrong and how you would know early. Richard Penman reminds you to connect the story to the numbers and to keep the accrual engine clean. Koller and colleagues pull you back to return on invested capital and growth that does not destroy value. White–Sondhi–Fried make you respect the plumbing between statements. Schilit and Perler ensure you do not get charmed by polished narratives.
Conclusion
Use AI to process and interpret while your own judgement focuses on causality, materiality and monitoring.
Forensic security analysis is not about paranoia or cleverness. It is about doing the basics such as extraction, reconciliation and cross-checking every single time, with no skipped steps. Artificial intelligence lets you scale those basics across a watchlist and focus your human attention on causality and materiality. With the module prompts above and the prompt library’s master report prompt, you can institutionalize that discipline in a fast, consistent, focused manner.
If you found this insightful, consider becoming a subscriber for more techniques, access to the copy/past equity research prompt library and the latest updates on AI capabilities in Finance:
Read more about The Inferential Investor here.
As always,
Inference never stops. Neither should you.
Andy West
The Inferential Investor.



