The Center for Audit Quality, together with KRC Research, published its institutional investor survey on AI use in August 2026. The survey was conducted in May 2026 and represents views from 100 investors with $500 million or more in assets under management.
Four findings from the survey directly change how CFOs should approach disclosure drafting.
Finding 1: 68% of institutional investors report extensive or moderate use of AI in investment research and decision-making, including 26% using it extensively across multiple stages.
Finding 2: 41% use AI regularly on company filings, specifically 10-Ks and 10-Qs. Broken down: 16% use AI on filings all the time, 25% often. The CAQ's description of this finding: "AI tools are increasingly serving as the first reviewer of corporate disclosures and will continue to do so as investors' use of AI grows."
Finding 3: 83% of investors expect their use of AI in investment research to increase over the next two years, with 42% expecting significant increases.
Finding 4: the most common AI uses on company filings are extracting financial or operational metrics (60%), summarizing MD&A and risk factors (58%), identifying inconsistencies between narrative and financial information (48%), detecting changes over time (47%), synthesizing insights across multiple disclosures (46%), and identifying anomalies or red flags (45%).
This blog is distinct from the SEC AI reads your filings blog (August 16) in this cluster, which covered the SEC enforcement unit's AI scanning of public company filings. That blog covered AI used by a regulator. This blog covers AI used by investors. The disclosure quality implications are related but distinct: the SEC's AI reader triggers comment letters and enforcement referrals; the investor's AI reader triggers analyst coverage decisions, stock price movements, and capital allocation. Both audiences are now AI-first readers of your financial disclosures.
What Did the CAQ's August 2026 Institutional Investor Survey Find About AI Use in Financial Reporting Analysis?
The CAQ's survey, confirmed from the CAQ's own publication page, the Lexology coverage from August 18, 2026, Cooley's TheCorporateCounsel.net blog from August 2026, and the CAQ Audit Committee Insights August 2026 edition, is the most comprehensive quantitative survey of AI use in institutional investor financial analysis published to date.
The survey population: 100 institutional investors in the United States, each with $500 million or more in assets under management, employed as professional investors with more than five years of experience, at the director level or higher. This is the specific audience that reads your 10-K and influences your company's stock price, analyst coverage, and cost of capital.
The AI use breakdown:
Overall investment research: 68% report extensive or moderate AI use, including 26% who use it extensively across multiple stages of their research process. Only 12% are aware of AI tools but not actively using them.
Company filings (10-Ks and 10-Qs): 41% use AI regularly on filings (16% all the time, 25% often). An additional 33% use AI occasionally on filings, meaning 74% of surveyed investors use AI on company filings at some frequency.
Earnings calls: 60% regularly use AI to analyze earnings call transcripts, the highest adoption rate for any specific application. This exceeds the 41% who use AI on 10-Ks and 10-Qs.
Financial modelling: 46% use AI for financial modelling and projections. Risk assessment: 49% use AI for risk assessment scenario analysis.
The adoption trajectory: 83% of investors expect their AI use in investment research to increase over the next two years, with 42% expecting it to increase significantly. The 41% who currently use AI on filings regularly will become a larger majority within the current planning horizon.
The AI tool landscape: 37% use hybrid in-house and third-party AI tools, 29% primarily third-party tools, and 26% primarily in-house tools. This means the AI reading your 10-K may be a commercial large language model (GPT, Claude, Gemini), a proprietary investment research AI, or a hybrid system. Each has different strengths and failure modes, but all share the common capabilities of text extraction, pattern recognition, and inconsistency detection.
What Does "AI as the First Reviewer of Corporate Disclosures" Mean in Practice?
The CAQ's specific language about 41% of investors using AI on filings: "AI tools are increasingly serving as the first reviewer of corporate disclosures and will continue to do so as investors' use of AI grows."
The phrase "first reviewer" has a specific meaning in the disclosure drafting context. Previously, the disclosure team's standard was to produce a document readable and informative for a human analyst who would read it in sequence, beginning with the cover page, proceeding through the business overview and risk factors, and then reading the MD&A and financial statements with focused attention on the year-over-year changes. That human reading sequence, with its associated cognitive patterns (humans skip sections they consider boilerplate, humans read year-over-year comparisons sequentially), shaped how financial disclosures were written.
An AI reader does not read in sequence. It processes the entire document simultaneously. It extracts all quantitative claims at once. It compares every number in the narrative section to every corresponding number in the financial tables without reading intervening text. It identifies semantic similarity between the current filing's language and the prior year's language, and between the current filing's language and peer company language, in seconds.
The implications of this reading sequence difference for disclosure drafting are specific and immediate. An AI reader will not give the benefit of the doubt to an ambiguous sentence that a sympathetic human analyst would have read favourably. It will flag the ambiguity. An AI reader will not apply context to a round number in the narrative that differs from a precise number in the table. It will flag the inconsistency. An AI reader will not skip the risk factor section because the analyst remembers the company's risk profile from the prior year. It will extract and compare every risk factor to prior year language and to peer language.
The 40% of investors who say AI has limited or no role in their decision-making today are the remaining human-first readers. Disclosures that would previously have been adequate for the full audience are now adequate only for the 40%. The 60% who use AI extensively or moderately are already applying a different standard.
What Is the AI Anomaly Detection Problem: How Investor AI Flags the Same Disclosures That Draw SEC Comment Letters
The survey finding that 45% of investors use AI specifically to identify anomalies or red flags in financial disclosures creates a direct connection between investor AI analysis and SEC comment letter risk.
The SEC comment letter record, covered in the companion blogs in this cluster, consistently focuses on three categories: generic language without company-specific detail (the tariff and geopolitical risk comment), inconsistencies between the narrative and the financial tables (the MD&A quantification comment), and statistical outliers relative to peer company disclosures (the non-GAAP reconciliation comment). All three of these comment letter triggers map directly to the AI capabilities investors are using most frequently.
A disclosure that draws an SEC comment letter for generic language will also be flagged by investor AI as lacking specific, extractable information. The same NLP models that detect generic language in peer benchmarking (described in the SEC AI reading blog) are available to institutional investors through commercial AI platforms.
A disclosure with internal inconsistencies between the narrative percentage claims and the financial table data will be flagged by the SEC's AI reader (confirmed from Law.com) and will be flagged by investor AI trained to identify inconsistencies between narrative and financial information (48% of investors specifically use AI for this purpose).
A disclosure with metrics that are outliers relative to peers will generate a comment letter from the SEC staff and will be flagged by investor AI trained to synthesise insights across multiple disclosures (46% of investors do this) and to identify anomalies relative to the investor's existing database of peer company metrics.
The practical consequence for disclosure committees: the disclosure that previously needed to satisfy only one reviewer type (human analysts at institutional investors, human staff at the SEC) now needs to satisfy two simultaneous AI readers with overlapping detection capabilities. The disclosure committee's internal quality check must anticipate both audiences.
MD&A Drafting for an AI Reader: What Changes When the First Reader Is an Algorithm, Not an Analyst
The two most common AI applications on company filings from the CAQ survey are extracting financial or operational metrics (60%) and summarising MD&A and risk factors (58%). Both of these applications have specific implications for how MD&A should be drafted.
For metric extraction: AI metric extraction works by identifying quantitative claims in the text and associating them with the underlying data context. A claim that says "revenue grew approximately 12%" is extracted as a ~12% revenue growth claim. If the financial table shows $4.5 billion to $4.8 billion (6.7% growth), the AI extraction identifies a numeric inconsistency. If the claim said "revenue grew 6.7%", the AI extraction would be satisfied.
The metric extraction implication for MD&A drafting: every percentage, dollar amount, and basis point claim in the MD&A must be the exact figure derivable from the financial table, not a rounded approximation that differs from the precise table figure. This is the same ecological error risk described in the companion SEC AI reading blog, applied to investor AI readers rather than enforcement unit AI readers. The precision requirement is identical.
For MD&A summarisation: AI summarisation works by extracting the most information-dense sentences from the text and synthesising them into a shorter summary. A disclosure that buries the most important operational development in the third paragraph of the fourth section of the MD&A will have that development either missed by the AI summary (if the density metric ranks it low) or appearing without the contextual qualifications that surrounded it in the original text.
The summarisation implication: the most material operational developments, business risks, and financial drivers should be explicitly stated early and directly in the MD&A, not embedded within longer paragraphs that contextualise them. An AI summarisation of an MD&A that front-loads material information will represent the company's story accurately. An AI summarisation of an MD&A that buries material information in dense contextual paragraphs will produce an incomplete or distorted summary that investors see before reading the underlying document.
Risk Factor Language for an AI Reader: Why Generic Language Is Now Detected and Penalized Instantly
The CAQ survey finding that 58% of investors use AI to summarise risk factors, combined with the survey finding that investor AI identifies anomalies and red flags in disclosures, creates a specific risk factor drafting implication.
An AI reading a risk factor section applies semantic similarity analysis: it compares each risk factor to a database of risk factors from prior year filings by the same company and from peer company filings. A risk factor that is semantically similar to the company's own prior year risk factor without meaningful updating is flagged as unchanged. A risk factor that is semantically similar to generic industry boilerplate rather than to the company's specific situation is flagged as non-specific.
Neither of these flags was previously visible to human analysts reading a single filing. A human analyst reading a current-year risk factor against the prior-year risk factor would detect obvious identical language but would not detect subtle rewording that preserves the same meaning. An AI semantic similarity analysis detects both.
The risk factor quality standard that AI readers impose is therefore higher than the standard previously imposed by human readers: each risk factor must be demonstrably specific to the company's current situation and must reflect material changes from the prior year's risk factor language. A risk factor that is materially identical to the prior year's language without corresponding business reason will be flagged as boilerplate by investor AI.
The SEC comment letter practice on risk factor specificity (covered in the companion Q2 season wrap blog) confirms that the SEC staff has been asking for company-specific risk factors for several years. The investor AI finding confirms that institutional investors are now applying the same specificity standard with AI tools. The audience for risk factor specificity has doubled: both the SEC staff and 58% of institutional investors are now reading risk factors with AI tools that detect generic language.
Earnings Call Transcript vs. 10-Q: What Happens When AI Identifies Inconsistencies Between Them
The highest AI adoption rate in the CAQ survey is 60% of investors regularly using AI to analyse earnings call transcripts. The 41% who use AI on 10-Ks and 10-Qs represents a lower but rapidly growing adoption rate. The combination of these two findings creates a specific risk for CFOs: investors are using AI to compare earnings call statements to 10-Q disclosures, and the AI is specifically trained to identify inconsistencies.
The inconsistency risk between earnings calls and 10-Qs: a CFO who makes optimistic AI capability claims on an earnings call that are not reflected in the 10-K's AI disclosure section is creating a CETU-adjacent exposure (as described in the companion SEC three AI bodies blog) and simultaneously creating an investor AI inconsistency flag. The investor AI that reads the earnings call transcript and the 10-Q will identify the divergence and flag it for the analyst's attention.
Concrete examples of earnings call vs 10-Q inconsistencies that investor AI detects:
A CFO describes the company's AI-driven operational efficiency on the earnings call using specific productivity metrics. The 10-Q's MD&A does not reference those metrics or discusses AI only in the risk factor section as a general risk. The inconsistency between the optimistic earnings call framing and the more cautious 10-Q framing is a flag.
A CFO reports on the earnings call that revenue for the current quarter will be "at the high end" of guidance. The 10-Q, filed several weeks later, discloses actual revenue that is below the midpoint of guidance. The inconsistency between the guidance-call language and the subsequent 10-Q is a flag.
A CFO discusses a specific business segment's performance on the earnings call using metrics not disclosed in the 10-Q's segment reporting. The investor AI that is summarising MD&A and risk factors will note the absence of the earnings call metrics from the 10-Q, generating a question about why the most-discussed business metric is not disclosed in the official filing.
The cross-document consistency check is now performed systematically by 60% of institutional investors using AI. The disclosure committee's review process must include a specific comparison of every material statement from the earnings call to the corresponding 10-Q disclosure, confirming that the 10-Q is not less informative than the earnings call on any material topic.
What Investor AI Specifically Looks For: Speed, Anomaly Detection, and Qualitative Consistency and What Triggers a Human Deep-Dive
The CAQ survey found that investors rate speed of analysis (80%) and reduction in manual workload (79%) as AI's top benefits, followed by identifying anomalies, patterns, insights, and risks (the specific percentages for the sub-benefits were not broken out separately in the survey summary available).
The most significant implication of the speed benefit for CFOs: investor AI compresses the timeline from filing to analysis. A 10-Q filed at 5:00 PM on August 11 was previously read by human analysts over the following day or two, with research notes distributed to portfolio managers by the following week. An investor with AI tools reads the same 10-Q within minutes of filing, extracting the key metrics, summarising the MD&A, identifying year-over-year changes, and comparing the filing to peer disclosures before the company's IR team has sent its first investor email.
The after-hours stock price reaction to earnings releases, which has always been influenced by algorithmic trading, now has an additional algorithmic layer: AI-assisted investor analysis that identifies the key takeaways before human analysts complete their manual review. The companies whose disclosures communicate clearly to AI readers will be correctly valued in the immediate post-filing period; companies whose disclosures are ambiguous, inconsistent, or generic may be mis-valued until human analysts correct the AI summary.
The trigger for human deep-dive: the CAQ survey is explicit that AI informs most investors' decision process with human oversight (60%). For the remaining 40%, AI has limited or no role. This means the AI analysis is typically a first filter that determines which filings receive more intensive human attention. A filing that AI summarises as unremarkable receives less human analyst time. A filing that AI flags as containing anomalies, inconsistencies, or peer outliers receives more intensive human scrutiny.
This investor attention allocation pattern has a specific implication: a disclosure that is flagged by investor AI as containing an anomaly or inconsistency will receive both investor AI scrutiny and intensified human analyst scrutiny. The combination produces the most unfavourable investor attention outcome for the company.
What Is the Auditor's Role When AI Reads Disclosures First and Why the CAQ Says This Strengthens the Case for Audit
The CAQ's Audit Committee Insights August 2026 edition specifically addressed the investor AI survey in the context of audit quality and the auditor's role. The CAQ's position: investors want to know if companies are utilising AI and that it is governed properly, and the auditor's role in providing assurance over AI-related ICFR controls becomes more important as AI becomes embedded in financial reporting processes.
The CAQ's framing: audited financial statements retain high trust from institutional investors. The annual institutional investor survey (March 2026 edition) confirmed that 90% of investors rely on and 91% trust the reliability of audited financial statements. The August 2026 AI survey confirms that investors are using AI to analyse those audited statements. The combination produces a specific message for audit committees: the AI readers that investors are deploying are reading audited financial data, and the audit provides the assurance that the data those AI tools extract is reliable.
The auditor's role in the AI-as-first-reviewer environment: the COSO GenAI guidance and FEI AI-ICFR framework (covered in the companion blogs) require that AI tools used in financial reporting processes have adequate ICFR controls. The auditor assesses those controls under the amended AS 2201 standard (effective December 15, 2026). When investor AI reads a 10-K's financial metrics, it is reading metrics produced by a financial reporting process that the auditor has either assessed or has not. The presence of an unqualified audit opinion is the assurance that the financial data the investor AI is processing is reliable.
The audit committee implication: when briefing the audit committee on the CAQ survey findings, include the connection between the investor AI findings and the year-end audit scope. The AI tools that 41% to 74% of institutional investors are using to analyse your financial statements are reading data that your auditor is responsible for assuring. The completeness and effectiveness of the year-end audit over AI-assisted financial reporting processes is therefore directly connected to the quality of the data investor AI is reading.
What Three Changes Should Your Disclosure Committee Make Before the Q3 10-Q?
The CAQ survey findings produce three specific disclosure process changes for the Q3 2026 10-Q, each directly responsive to one of the top investor AI applications.
Change 1: implement the AI consistency pre-flight check as a named close deliverable. The six AI applications on company filings that rank highest in the survey (metric extraction, MD&A summarisation, narrative-financial inconsistency detection, change detection, cross-disclosure synthesis, and anomaly identification) are all addressed by the pre-filing AI consistency checklist described in the companion SEC AI reading blog. That checklist was written from the enforcement perspective. It applies equally from the investor perspective, because the same disclosure failures that draw SEC comment letters also draw investor AI anomaly flags. Make the checklist a named deliverable in the Q3 close calendar, assigned to a specific team member, completed from the final EDGAR submission package, and signed off before the SOX 302 certification is executed.
Change 2: run an MD&A front-loading review. Using the CAQ's finding that investor AI summarises MD&A by extracting the most information-dense sentences, review the Q3 MD&A draft and confirm that the most material operational developments, business risks, and financial drivers are explicitly stated in the first paragraph of each section, not embedded in qualifying language mid-paragraph. Each section's first sentence should contain the material conclusion; subsequent sentences should provide context and detail. This structure produces AI summaries that represent the company's story accurately.
Change 3: conduct an earnings call vs 10-Q consistency check. Assign a specific team member to read the most recent earnings call transcript alongside the Q3 10-Q draft and confirm that every material statement from the earnings call is either directly supported by or acknowledged in the 10-Q. Any earnings call statement that contradicts, weakens, or is absent from the 10-Q disclosure is a cross-document inconsistency flag for investor AI. Address each inconsistency before filing: either add the supporting disclosure to the 10-Q or update the earnings call language to reflect the more accurate 10-Q characterisation.
Frequently Asked Questions
Do institutional investors use AI to read 10-Ks and 10-Qs?
Yes. The CAQ's August 2026 institutional investor survey (conducted May 2026 with 100 investors with $500 million+ AUM, published in partnership with KRC Research) found that 41% of investors use AI regularly on company filings including 10-Ks and 10-Qs. An additional 33% use AI occasionally on filings, meaning 74% of surveyed investors use AI on company filings at some frequency. The CAQ described this finding as: "AI tools are increasingly serving as the first reviewer of corporate disclosures."
What does the CAQ's August 2026 investor AI survey show?
68% of institutional investors report extensive or moderate AI use in investment research and decision-making. 41% use AI regularly on 10-Ks and 10-Qs (16% all the time, 25% often). 83% expect AI use to increase over the next two years. The top AI benefits are speed of analysis (80%) and reduction in manual workload (79%), followed by anomaly and pattern identification.
What types of financial disclosures do investors use AI to analyse?
The six leading AI applications on company filings from the survey: extracting financial or operational metrics (60%), summarising MD&A and risk factors (58%), identifying inconsistencies between narrative and financial information (48%), detecting changes over time (47%), synthesising insights across multiple disclosures (46%), and identifying anomalies or red flags (45%).
How does AI investor analysis change what CFOs must include in MD&A?
MD&A must be front-loaded with material conclusions in the first sentence of each section, because AI summarisation extracts the most information-dense sentences. Every percentage and dollar claim in the narrative must precisely match the corresponding financial table figure, because AI metric extraction identifies numeric discrepancies. Earnings call statements must be consistent with 10-Q disclosures, because 60% of investors use AI to analyse earnings call transcripts and the AI identifies inconsistencies between them.
What anomalies does investor AI flag in financial disclosures?
Based on the survey's top applications: inconsistencies between narrative and financial information (48% of investors specifically use AI for this), anomalies or red flags in specific metrics (45%), changes over time that are inconsistent with disclosed business developments (47%), and peer outliers where the company's disclosed metrics diverge from industry peers (the cross-disclosure synthesis application).
Does AI investor review affect a company's stock price or analyst coverage?
The CAQ survey describes investor AI as the first reviewer that determines which filings receive intensive human analyst attention. A filing flagged by AI as containing anomalies or inconsistencies receives more intensive human scrutiny. Given the speed advantage (AI reads a filing within minutes of submission versus hours or days for human analysts), AI-assisted analysis compresses the timeline from filing to investor reaction. Companies whose disclosures communicate clearly to AI readers are correctly valued in the immediate post-filing period; companies with disclosure quality issues may experience adverse initial reactions before human analysts can correct the AI summary.
Key Takeaways
- The CAQ's August 2026 institutional investor survey (100 investors with $500M+ AUM, conducted May 2026) confirmed that 41% of investors use AI regularly on 10-Ks and 10-Qs (16% all the time, 25% often), and 74% use AI on filings at some frequency. The CAQ's description: "AI tools are increasingly serving as the first reviewer of corporate disclosures."
- 68% of institutional investors report extensive or moderate AI use in investment research; 83% expect that use to increase significantly over the next two years. The 41% who currently use AI on filings regularly will become a larger majority within the investor base's planning horizon.
- The six leading AI applications on company filings: extracting financial metrics (60%), summarising MD&A and risk factors (58%), identifying narrative-financial inconsistencies (48%), detecting changes over time (47%), synthesising across multiple disclosures (46%), and identifying anomalies or red flags (45%). Applications 3 and 6 (inconsistency detection and anomaly identification) directly overlap with SEC comment letter triggers and SEC enforcement AI capabilities.
- Investor AI and SEC AI read the same disclosures with overlapping pattern recognition capabilities. A disclosure that draws an SEC comment letter for generic language or internal inconsistency will also be flagged by investor AI before the SEC comment letter arrives.
- Three disclosure process changes before Q3 close: (1) implement the AI consistency pre-flight check as a named close deliverable, (2) front-load MD&A with material conclusions in the first sentence of each section for accurate AI summarisation, and (3) conduct an earnings call vs 10-Q consistency check to eliminate cross-document inconsistencies that investor AI flags.
- The CAQ survey reinforces the case for audit quality: investors use AI to read audited financial data, and the audit provides assurance that the data investor AI processes is reliable. The ICFR controls over AI-assisted financial reporting processes (covered in the COSO, FEI, and PCAOB blogs in this cluster) are directly connected to the quality of the data 41% to 74% of institutional investors are now reading with AI tools.
