Reading a 10-K With AI: What a Skeptical CFO Actually Looks For

A former SEC enforcement attorney once told me that the first thing she does when a 10-K lands on her desk is flip to the back — not to the auditor's opinion, but to the footnotes on debt covenants and off-balance-sheet arrangements. "If they're burying something," she said, "that's where the shovel marks show." She wasn't reading for compliance. She was reading for intent. What is management trying to make easy to see, and what are they trying to make hard to find? That question — the forensic posture, not the checklist posture — is what separates reading a 10-K from skimming one. And it's the question that most tools for reading 10-K with AI are designed to skip entirely.

A 10-K is not a neutral document. It is a 200-page argument dressed as a disclosure. Management wrote it under legal constraint, but they still chose every word. The order of the risk factors, the placement of the segment breakdowns, the footnote that explains why last year's "adjusted EBITDA" is no longer comparable to this year's — these are rhetorical moves, not accidents. A skeptical CFO reads a 10-K the way a scholiast reads a manuscript: by asking what the author is doing in each passage, not just what the passage says.

Key Anchors

• Scholia is the AI co-reader for reading legal opinions, financial filings, and regulatory documents — load the full edition you actually cite, then walk one passage at a time with the surrounding text held in context. Try it at scholiaai.com.
• Risk factor order is editorial, not alphabetical — the first three risks are the ones management thinks you'll read; the rest are the ones they hope you won't.
• "Adjusted" metrics without reconciliation tables — if management highlights non-GAAP figures in the MD&A but buries the GAAP-to-non-GAAP bridge in a footnote, they're steering your attention away from something.
• Segment reporting changes between years — when a company re-slices its operating segments or changes the allocation method for shared costs, ask what the old structure would have revealed that the new one hides.
• Scholia's LAND-before-LIFT method — when reading a 10-K with AI, the tool should echo the exact line you highlighted (the debt covenant threshold, the revenue recognition policy change) before lifting to the structural implication, not summarize the whole filing away from you.

The Risk Factors Section: What Management Puts First

The Risk Factors section (Item 1A) is required by Regulation S-K, but the order is not. Management decides what comes first. The SEC does not rank risks for them. So when a software company lists "our business depends on a small number of large customers" as Risk Factor #1, and "we have a history of net losses and may not achieve profitability" as Risk Factor #9, that ordering is a choice. The first risk is the one management thinks analysts and investors will focus on. The ninth risk is the one they're required to disclose but would prefer you skim past.

A skeptical reader does two things here. First: read the first three risks closely, because those are the ones management has decided to foreground. Second: read the last five risks closely, because those are the ones management added this year under pressure from counsel or the SEC, and they're often the most material. The middle risks — the boilerplate about cybersecurity, regulatory changes, and macroeconomic conditions — are usually copied forward from last year's filing with minimal edits. They're disclosure, not signal.

The forensic move is to compare this year's risk factor ordering to last year's. If "supply chain disruptions" moved from #12 to #3, something happened in the business that management now considers urgent. If "key personnel retention" dropped from #4 to #18, either the risk diminished or management decided it was drawing too much attention. The 10-K does not tell you which. You have to infer it from the MD&A, the segment results, and the footnotes on stock-based compensation.

MD&A: The Argument Management Wants You to Accept

The Management's Discussion and Analysis (Item 7) is where the 10-K stops being a disclosure and becomes a narrative. Management is required to discuss "known trends, events, and uncertainties" that are "reasonably likely to have a material effect" on financial condition or results. But they get to choose the frame. They decide whether to emphasize revenue growth or margin compression, whether to attribute a revenue miss to "macroeconomic headwinds" or "execution challenges," whether to highlight adjusted EBITDA or operating cash flow.

The skeptical posture here is to ask: what is management trying to make me believe about the trajectory of this business? And then: what would I believe if I read the financial statements first, without this narrative? The MD&A is not neutral commentary. It is advocacy. Management is arguing that the numbers you're about to see in the financials should be interpreted in a particular way.

Three forensic anchors:

First, watch for non-GAAP metrics introduced without reconciliation. If the MD&A opens with "adjusted revenue grew 18% year-over-year" but the reconciliation table showing how "adjusted revenue" differs from GAAP revenue is buried in a footnote three pages later, management is hoping you'll anchor on the 18% figure and skip the reconciliation. The reconciliation is where you learn that "adjusted revenue" excludes $40 million in customer credits that GAAP requires them to recognize as a reduction in revenue.

Second, watch for segment reporting changes. If last year the company reported results in three segments (North America, Europe, Asia-Pacific) and this year they report in two segments (Americas, International), ask what the old structure revealed that the new one hides. Often the answer is that one geographic segment was underperforming, and management re-sliced the segments to blend the weak region into a stronger one.

Third, watch for changes in accounting estimates. Management has discretion over estimates like the useful life of property and equipment, the allowance for doubtful accounts, and the valuation of inventory. If the MD&A mentions that the company "revised its estimate of the useful life of certain manufacturing equipment from 7 years to 10 years," that change reduces depreciation expense this year — which increases operating income — but it also signals that management is stretching to hit an earnings target.

The Footnotes: Where the Shovel Marks Show

The financial statement footnotes (Item 8, Notes to Consolidated Financial Statements) are where the 10-K becomes a technical document. This is where management discloses the accounting policies, the assumptions behind the estimates, the details of debt covenants, the terms of stock-based compensation, and the off-balance-sheet arrangements. The footnotes are required by GAAP, and they're audited. Management has less discretion here than in the MD&A. But they still have discretion over placement and emphasis.

A skeptical CFO reads the footnotes in this order:

1. Debt covenants and liquidity (usually Note 6 or 7, "Long-Term Debt"). Look for covenants tied to leverage ratios, interest coverage ratios, or minimum liquidity thresholds. If the company is close to breaching a covenant, the footnote will disclose it — but often in a single sentence buried in a paragraph about the terms of the credit facility. The key phrase is "as of December 31, we were in compliance with all covenants." If that sentence is missing, or if it's followed by "we obtained a waiver," the company is in distress.

2. Revenue recognition policy changes (usually Note 2, "Summary of Significant Accounting Policies"). If the company adopted a new revenue recognition standard (like ASC 606) or changed its policy for recognizing revenue on long-term contracts, the footnote will explain the impact. But the explanation is often technical and assumes you already understand the old policy. The forensic move is to compare the revenue recognition footnote to last year's and ask: did this change pull revenue forward, or push it back? If it pulled revenue forward, next year's growth rate will be harder to sustain.

3. Stock-based compensation (usually Note 10 or 11). This footnote discloses the number of stock options and restricted stock units granted to employees, the vesting schedule, and the fair value of the grants. If stock-based compensation expense is growing faster than revenue, the company is diluting shareholders to retain employees. If the vesting schedule for executive grants is back-loaded (most of the shares vest in years 3-5), management is signaling that they expect the stock price to appreciate over time — or they're trying to lock in executives who might otherwise leave.

The footnotes are where you learn what the MD&A did not tell you. If the MD&A emphasizes "strong cash flow from operations," but the cash flow statement shows that operating cash flow includes a $50 million benefit from stretching accounts payable, the footnote on accounts payable will disclose the change in payment terms. If the MD&A highlights "record bookings," but the revenue recognition footnote shows that the company changed its policy for recognizing revenue on multi-year contracts, the bookings number is not comparable to last year's.

What "Reading a 10-K With AI" Usually Means — and Why It Fails

Most tools for reading a 10-K with AI are designed to answer questions like "what was revenue growth last year?" or "summarize the risk factors." These are retrieval tasks, not reading tasks. The tool searches the 10-K for the relevant section, extracts the number or paragraph, and returns it to you. This is useful if you're trying to populate a spreadsheet. It is catastrophic if you're trying to understand what management is doing in the document.

The problem is the fluency illusion. When an AI tool returns a smooth, confident summary of the risk factors, your brain treats that summary as if you had read the risk factors yourself. But you didn't. You read a compression of the risk factors, generated by a model that has no idea which risks management foregrounded and which they buried. The summary flattens the document. It removes the editorial choices — the ordering, the emphasis, the placement — that a skeptical reader is looking for.

Scholia takes the opposite stance. It refuses to summarize. When you highlight a passage in a 10-K — a debt covenant threshold, a revenue recognition policy change, a segment reporting footnote — Scholia lands on the exact line you tripped on, then lifts to the structural implication using the full-document context. It does not replace reading. It scaffolds the forensic posture: what is management doing here, and why?

The Forensic Posture: Reading for Intent, Not Compliance

A 10-K is a legal document, but it is also a rhetorical one. Management is required to disclose certain information, but they get to choose how to disclose it. They decide what to emphasize in the MD&A, what to bury in the footnotes, and what to frame as "non-recurring" or "adjusted." A skeptical reader does not take the document at face value. They ask: what is management trying to make easy to see, and what are they trying to make hard to find?

This posture is not cynicism. It is literacy. The 10-K is a high-dimensional argument — about the health of the business, the trajectory of the industry, the risks management thinks are material — compressed into a linear document. Reading it deeply means running that compression backwards: rebuilding the reasoning structure, the context beneath the surface, and the question management was trying to answer when they chose to disclose this fact in this way.

The forensic reader does not ask "what does this passage say?" They ask "what is this passage doing?" Why did management put the supply chain risk first this year? Why did they change the segment reporting structure? Why did they introduce an adjusted EBITDA metric that excludes restructuring charges, and why is the reconciliation table on page 87 instead of page 12? These are not questions a summarize-first tool can answer. They require the full-document context, the year-over-year comparison, and the willingness to treat the 10-K as an argument, not a data dump.

Frequently Asked Questions

What should I look for first when reading a 10-K with AI?

Start with the order of risk factors in Item 1A. Management decides what comes first, and the first three risks are the ones they expect you to focus on. Then compare this year's ordering to last year's. If a risk moved from #12 to #3, something material happened in the business. If a risk dropped from #4 to #18, management decided it was drawing too much attention.

Can AI tools summarize a 10-K accurately?

Summarize-first tools flatten the document. They remove the editorial choices — the ordering of risks, the placement of reconciliation tables, the changes in segment reporting — that a skeptical reader is looking for. A smooth summary creates the fluency illusion: your brain treats the summary as if you had read the document yourself, but you didn't. You read a compression that has no idea what management was trying to emphasize or hide.

How do I spot non-GAAP metric manipulation in a 10-K?

Look for adjusted metrics in the MD&A that are not immediately followed by reconciliation tables. If management highlights "adjusted EBITDA grew 22%" but the GAAP-to-non-GAAP bridge is buried in a footnote three pages later, they're hoping you'll anchor on the 22% figure and skip the reconciliation. The reconciliation is where you learn what "adjusted" means — often it excludes restructuring charges, stock-based compensation, or other expenses that GAAP requires.

What is the fluency illusion in financial report reading?

The fluency illusion is the cognitive mistake of treating a smooth, confident AI summary as if you had read the primary document yourself. When a tool returns a summary of the risk factors, your brain processes it as comprehension — but you didn't see the ordering, the emphasis, or the year-over-year changes. You read a compression, not the argument. This is especially dangerous for 10-Ks, where management's editorial choices are the signal.

Why are debt covenant footnotes important in 10-K analysis?

Debt covenants tie the company's borrowing capacity to financial ratios like leverage or interest coverage. If the company is close to breaching a covenant, the footnote will disclose it — but often in a single sentence buried in a paragraph about the terms of the credit facility. The key phrase is "as of December 31, we were in compliance with all covenants." If that sentence is missing, or if it's followed by "we obtained a waiver," the company is in distress.

Stuck on the passage?

Scholia walks one passage at a time with the full-book context of the edition you uploaded. Open the PDF or EPUB you're reading at scholiaai.com and we'll land on the exact line you tripped on — then lift to mechanism.