AI in finance: what 10-K and 10-Q filers say it does in the finance function, counted by process and quarter

In 10-K and 10-Q filings, AI used in finance departments shows up far more often as a word near finance vocabulary than as a described practice: 152 filings in 2026 to 2026-11-11 named both “artificial intelligence” and “financial close”, but in a sample of 168 filings that named AI beside a finance process, only 15 put the two in the same sentence.¹ Of those 15, one describes a company applying AI to its own accounting work, eight describe products or services sold to other companies’ finance teams, and six use the words for something else.¹

The counts below are of filings, not companies, and of words, not deployments. Each figure says how it was made.

The question and the process list

Six phrases stand for six pieces of finance work, chosen because each names the work rather than a line item: “financial close”, “account reconciliations”, “financial planning and analysis”, “invoice processing”, “tax compliance” and “cash forecasting”. The obvious phrase for payables, “accounts payable”, was left out because nearly every 10-K prints it on the balance sheet, so pairing it with AI would count almost every AI filing. Each phrase was searched together with “artificial intelligence” on forms 10-K and 10-Q.¹

Method: full-text counts, a sample, and the sentence test

The counts come from EDGAR full-text search, run one calendar quarter at a time from 2023-Q1, paged to the end, and reduced to distinct accession numbers, so a filing counts once however many times it uses the words.¹ A 10-K and the three 10-Qs a company files in a year are four filings. Both phrases can sit anywhere: an AI risk factor in Item 1A and “financial close” in Item 9A make a hit.

To see what the hits say, a seeded random sample of up to eight main documents per process and quarter, for the four quarters 2025-Q4 to 2026-Q3, was drawn from the search results and each document read on sec.gov.¹ Where a quarter had fewer than eight, all were read. A filing passes the sentence test when at least one sentence names an AI term (artificial intelligence, generative AI, machine learning, AI, agentic) and matches the process pattern: the site’s published topic rules for the close, payables, forecasting, tax and treasury, and “reconciliation” for the last.¹ Every passing sentence was then coded for stage (in use, piloting, planning, or none of these) by the verb rules in the site’s research script.

Filings naming AI by process and quarter

“Financial close” beside AI went from 15 filings in 2023-Q1 to 55 in 2026-Q1; “account reconciliations” from 35 to 72; “financial planning and analysis” from 39 to 147; “invoice processing” from 6 to 11.¹ Calendar years tell the same story without the first-quarter spike: financial close 52 filings in 2023, 85 in 2024, 111 in 2025 and 152 in 2026 to 2026-11-11.¹

Two phrases are left off the chart. “Tax compliance” is the largest, 585 filings in 2026-Q1 alone, and the sample shows why: its two passing passages were a list of laws a company must follow and a note on taxes on electricity, not the tax department’s work.¹ “Cash forecasting” is the smallest, no filings at all in 2023 and between 3 and 5 in each full quarter since 2025-Q1.¹

What a sample read on sec.gov says

The sentence test cuts the counts down hard. Of 32 sampled filings with AI and “financial close”, none had a sentence joining them; the same for 32 with “account reconciliations” and all 16 with “cash forecasting”.¹ FP&A passed in 7 of 32, invoice processing in 6 of 24, tax compliance in 2 of 32.¹

Where the close and reconciliation words sit

The two phrases that never met AI in a sentence have a home of their own in these filings. In 25 of the 32 sampled filings with “financial close”, the phrase’s first use sits within 1,500 characters of the words “material weakness” or a form of “remediate”; the same is true for 25 of the 32 with “account reconciliations”.¹ For FP&A, invoice processing, tax compliance and cash forecasting, the count is 2 of 104 filings put together.¹ The rule is mechanical (a window of text around the first use, read in the filing on sec.gov) and is printed in the site’s research script.

That placement explains most of the growth in the first chart. A small company that reports a material weakness in its financial close and reporting process, and also carries a risk factor about artificial intelligence, is a hit for both phrases, and both kinds of disclosure have become more common. The close shows up in these filings as a control that failed or is being repaired, written in the controls items of the 10-K and 10-Q, not as a process a model now runs. A finance leader reading the count as adoption would be reading the controls section as a product brochure.

What the sample does not show is as useful. No sampled sentence that names AI describes a model that posts journal entries, matches bank statements or drafts the close checklist, and none names a vendor whose AI it uses for the close.¹ The passages that come nearest are the payables products in the next group, sold by their makers to other companies.

Read one by one, the 15 passing filings sort into three groups.

A company using AI on its own finance work: one filing. The OLB Group’s 10-Q for the quarter ended 2026-06-30 is the only sampled passage in which a filer says it is building AI for its own accounting.²

Products and services sold to other finance teams: eight filings. XBP Global’s 10-K says its accounts payable and procure-to-pay solution “leverages artificial intelligence to simplify the complexities of supplier onboarding, management, and invoice processing,” and that it “performs AI-based three-way matching.”³ Global AI, Inc. filed the same description in a 10-K and two 10-Qs: a deployment at a customer “automates how supplier invoices are received, validated, and recorded across the customer’s finance systems,” with agents “processing invoices without manual intervention, and escalating exceptions for finance team review.”⁴ Paylocity’s 10-K describes “AI-powered Accounts Payable Automation” covering every step “from vendor onboarding and invoice processing to bill coding, approvals, payments” and reconciliation.⁵ Resources Connection, a consultancy, lists “financial planning and analysis enhancement” and “AI adoption” among its core solutions in a 10-K and two 10-Qs.⁷

The words meeting by chance: six filings. An executive’s biography that lists a past FP&A role and a later AI title; a list of laws that runs from foreign account tax compliance to artificial intelligence; a risk that AI-related demand makes it harder to forecast manufacturing capacity; labor forecasting in a payroll product; taxes on the electricity that AI equipment uses; and Kimball Electronics’ statement that it uses AI across its business, including “supply chain forecasting, back office productivity tools, and certain administrative functions.”⁶ Kimball’s is a statement of use, and the back office is in it, but the forecasting it names is supply chain forecasting, not FP&A.

Stage: in use, piloting, planning

The stage rule found 4 in-use statements among the 15 and no statement of a pilot or a plan; the other 11 state no stage.¹ Three of the four are Global AI’s descriptions of deployments at its customers, and the fourth is Kimball’s.⁴⁶ OLB’s “we are developing” falls outside the rule’s verbs and codes as no stage, which is why the sentence is printed above rather than counted. With 15 passages the split cannot carry a trend, and none is drawn here.

What filers say they keep manual

The sample was also searched for a human-review statement in any sentence naming AI: human review, oversight, judgment or intervention, “human in the loop”, review by personnel or management, manual review. 24 of the 168 sampled filings have one.¹ Global AI’s passage is the only one of the 15 that joins a finance task to a person in the same sentence, with exceptions escalated “for finance team review.”⁴

The Federal Reserve has put reconciliation on the record as an early use. Governor Christopher J. Waller said on 2026-09-29 that the payments industry “was also an early adopter of large language models (LLMs) for payment reconciliation and similar tasks.”⁸ The search found reconciliation beside AI in 167 filings in 2026 to 2026-11-11 and, in the 32 sampled, never in the same sentence.¹

What changed since the site’s first filings entry

The site’s filings log holds 67 published entries filed from 2025-11-04 to 2026-11-10, each a verbatim passage from a 10-K, 10-Q or 8-K that puts AI next to finance work, found by targeted searches and read in the filing. By the topic rules printed on the methodology page, fraud and deepfakes lead with 19 entries, then controls and audit with 12, AP and payments 11, the close 9, forecasting and FP&A 7, and tax and treasury one each; 15 entries fit no topic.

The log and the sample point the same way: passages that describe finance work done with AI are rare enough to be read one at a time, and in the sample most come from companies that sell the tools. The entries are grouped under the close, AP and payments and forecasting and FP&A, and the passages where filers describe AI in their own finance work are collected in AI in their own finance work.

How to use the counts, and when not to

The full-text counts are a fair measure of how often the vocabulary of AI and the vocabulary of finance work appear in the same SEC filing, quarter by quarter. They rise every year for every phrase except cash forecasting, which is too small to say.¹ They are not a measure of how many finance departments use AI, and they are not a count of companies: a calendar-year filer contributes up to four filings a year.

The sample is a small, seeded draw from four quarters, and its figures apply only to the filings drawn. It does not support a share for all filers, and none is given here. Every count in this article is recomputed from the saved search results and the sample file by scripts/cfonews-ai/articles/compute.mjs, and the sample can be drawn again from the same seed by the research script beside it.