Why The Traditional 10-K Is Ripe For An AI Overhaul

Forbes

4 days ago

William Tarr is the co-founder and CEO of Tergle, an audit-focused AI software startup in San Francisco.
The final night of a year-end close feels like what I imagine air traffic control would be like if it ran on paper. Bank confirmations, leasing schedules and goodwill tests crowd a war room-like table while someone refreshed the SEC's EDGAR uploader every 10 minutes. Even in the age of AI, most of that choreography still looks the same, only the spreadsheets have become denser and the filing clock less forgiving.
The modern 10-K, an annual report most public companies are required to submit to the U.S. Securities and Exchange Commission (SEC), often runs hundreds of pages, embeds thousands of XBRL tags and must track business lines that range from China to the cloud. Yet, some firms continue to assemble it with the digital equivalent of masking tape.
An Area Poised For Disruption
Regulators have begun to signal that this status quo cannot last. On March 27, 2025, the SEC devoted an entire public roundtable to AI in finance to discuss how smarter tools may support automation, investor communication, risk management and more. In my view, the timing is no accident: The number of public companies forced to restate or withdraw financial statements climbed to a nine-year peak, with 140 restatements in just the first 10 months of 2024. That's more than double the figure four years earlier, the Financial Times reported (registration required). High-profile missteps, like Macy's $151 million in false bookkeeping entries, have made headlines and can erode investor confidence.
Behind those headlines lies a structural problem. Financial reporting has outgrown the batch-process mindset of the 20th-century ledger. Transactions now stream in real time, yet accountants reconcile them weeks later; investors consume numbers in seconds, yet issuers take up to 60 days to publish a comprehensive story. That mismatch creates a widening gap where errors or, worse, manipulations, can hide. AI is not a silver bullet, but it offers a fundamentally different posture: continuous, data-first scrutiny instead of episodic, paper-trail inspection.
Consider how a filing might evolve if AI became native rather than ornamental. Journal entries could flow through a language-model pipeline as they are booked, which would allow statistical and semantic engines to flag outliers the moment they appear, like an accrual growing 10 times faster than revenue or a related-party vendor with a residential address.
During drafting, another model could cross-reference every XBRL tag with the surrounding prose, catching the mislabeled lease liability before it summons an SEC comment letter. Even the narrative sections would change: Transformers trained on years of peer filings could alert preparers that their risk-factor discussion omits a topic the market now expects, such as generative AI cybersecurity.
Caution Is Warranted
Skepticism remains healthy, however. Language models can hallucinate, and finance leaders justifiably fear black-box decisions that lack an audit trail.
The answer is not to bury AI but to govern it. Every inference must be logged; role-based controls should separate data ingestion, model tuning and approval; and audit committees ought to receive AI exception reports alongside traditional control matrices. With those guardrails, I believe the technology looks less like a threat to professional judgment and more like a tireless junior who never sleeps and never tires of ticking and tying.
Firms can also explore best-practice playbooks that are beginning to crystallize, such as the Public Company Accounting Oversight Board's July 2024 generative AI "Spotlight,' which highlights the need to keep a human reviewer accountable for each AI-assisted step, among other strategies. The SEC's 2023 cybersecurity disclosure rules likewise require boards to describe how they govern technology risk—obligations that extend to any AI system touching financial statements. Companies operating across the Atlantic must map those workflows to the EU AI Act, which imposes explicit data governance, transparency and human oversight duties.
In practical terms, that means tightening access controls so sensitive ledgers never leak into public large language models (LLMs), documenting prompt engineering as rigorously as journal-entry support and training staff to recognize—and correct—model hallucinations with the same skepticism they apply to a junior analyst's shaky reconciliation.
Looking Ahead
Early adopters are already discovering practical entry points. Revenue-recognition waterfalls and inventory roll-forwards lend themselves to anomaly detection because the data is structured and the business logic is explicit. Optical character recognition systems that once balked at faint fax copies can now parse documents with high accuracy, opening a path to automate footnote support files long trapped in PDF purgatory. Once confidence builds, firms can expand into narrative coherence checks and real-time benchmarking, turning what used to be a frantic year-end scramble into a measured, incremental close.
For finance chiefs weighing the leap, the calculus is shifting from 'if' to 'when.' KPMG's "2024 Global AI in Finance" report (download required) found that 95% of CFOs expect to be using generative AI in some phase of reporting within three years, up from less than 40% at the time of the survey. In my view, the first movers will not only shorten close cycles, they will also accumulate proprietary training data that becomes a moat that's harder for latecomers to replicate than any single piece of software.
In the 1930s, the newly created SEC demanded annual transparency and, over time, the 10-K was born. Today, the form still serves, but I believe its preparation often doesn't match the velocity or complexity of modern commerce. Whether prompted by rising restatements, investor impatience or regulatory prodding, the shift toward using AI for disclosures feels inevitable. The firms willing to re-code their reporting DNA now can position themselves to greet that future with confidence.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?