Five stories this week, in priority order.
Story 1: AI being used for audit, but ROI not auditable
A Gartner survey found 93% of audit leaders and auditors report some AI use in their work, but only 15% say their department has deployed formal, structured AI use cases that run routinely in audits. Among 142 chief audit executives (CAEs) Gartner polled in May 2026, 54% have not started measuring the value AI generates, and only 7% tie it to cost metrics such as reduced external spend. Only 38% of a separate group of 161 CAEs have any level of AI strategy. Of 743 respondents asked where the tools get used, 60% apply AI to engagement preplanning and 60% to drafting audit reports. Only 30% use it for audit testing, the step that most directly changes conclusions. The top barrier, cited by 48%, was unclear internal expectations for tool use.
Why it matters: if your controller or internal audit function already uses AI informally for prep and drafting, and nobody has defined what good looks like, you have AI exposure with no ROI tracking.
What to do this week: ask your internal audit or controller lead which AI tools are in informal use today, and whether anyone independently checks testing-stage output before it lands in a report.
Source: Gartner, Sept. 10, 2026, via Help Net Security, Sept. 15, 2026
Story 2: Cash flow insight still highest priority for CFOs
A PYMNTS Intelligence report with Fynapse (2026 Certainty Project, August 2026 edition) found 45% of middle market firms went through at least three significant business changes in the prior 24 months, and 62% added products or services. Asked where they plan to invest next, CFOs put cash flow forecasting first at 52%, followed by real-time reporting (35%), reducing manual processes (33%), and unifying finance data (32%). AI-driven tools ranked lower as a standalone priority. Only 12% of firms said their finance systems were completely prepared for the next two years, though 62% considered themselves at least mostly prepared.
Why it matters: vendors lead with "AI-powered." The CFOs doing the buying rank faster, more accurate cash visibility well ahead of AI as a category.
What to do this week: score your next treasury or FP&A tool evaluation first on what it does to cash flow forecast accuracy and reporting latency. Treat "AI-powered" as a feature description.
Source: PYMNTS, Sept. 10, 2026 (PYMNTS Intelligence 2026 Certainty Project with Fynapse, Aug. 2026 edition)
Story 3: AI use may atrophy foundational skills in junior employees
James Tucker, who leads corporate finance and strategy globally at Boston Consulting Group and talks with hundreds of finance chiefs a year, told Fortune the old finance hiring model is being replaced. That model brought in large junior cohorts to do reconciliations and journal entries and promoted the best of them. The new one hires a smaller group for judgment and puts them on quality control of AI-built output. The problem is that AI now does the research and first-draft work junior staff used to build judgment on. A Harvard working paper cited in the piece finds generative AI adoption reduces junior hiring in AI-exposed roles, with far less effect on senior staff. In a separate BCG global C-suite study, half of leaders said they are already seeing de-skilling, and more than 60% expect it to become a material problem within three to five years. Tucker's fix: concentrate the roughly 10% of reconciliation work that resists automation on fewer people, so an experience curve still forms.
Why it matters: if AI has absorbed your reconciliations and first-draft memos with no plan for where junior staff still get real reps, your future controllers are not being trained.
What to do this week: ask your senior controller which manual tasks junior staff still do, and whether those reps are concentrated enough to build judgment.
Source: Fortune, Sept. 9, 2026 (interview with James Tucker, BCG); BCG global C-suite study, 2026
Story 4: Companies now responsible for ACH fraud monitoring
As of June 20, 2026, Nacha's expanded fraud-monitoring rule requires every corporate entity that originates ACH payments, not only banks, to run a documented, risk-based process for detecting fraudulent entries, reviewed at least annually. Phase 1, for the largest originators only, took effect March 20, 2026. The fraud numbers were already bad. AFP's 2026 Payments Fraud and Control Survey, released April 14, 2026 (the most recent full-cycle data available, not a this-week finding), found 76% of organizations experienced attempted or actual payments fraud in 2025, with ACH debits cited as a vector by 30%.
Why it matters: "our bank handles fraud monitoring" is no longer a complete answer. If your company originates any ACH payments, the documented process is now your obligation.
What to do this week: ask your controller or treasury lead whether anything in writing describes how your company monitors for fraudulent ACH entries, or whether that has been assumed to sit with the bank.
Source: Nacha, rule effective June 20, 2026 (Phase 1: March 20, 2026) • AFP 2026 Payments Fraud and Control Survey, released Apr. 14, 2026 (older benchmark, not a this-week finding)
Story 5: Hallucinations still showing up in legal citations
In Clinco, T.C. Memo. 2026-16, decided in February and reported by the Journal of Accountancy on Sept. 1, 2026, the U.S. Tax Court found three of four case citations submitted by a taxpayer's attorney to challenge an IRS notice were fictitious, invented, per Judge Mark V. Holmes, by apparent AI hallucination. The IRS flagged the fake citations in its response brief. Counsel never corrected the record and repeated one fictitious case in a later filing. The court stopped short of formal sanctions but warned that "submitting a brief with fictitious caselaw is a recipe for sanctions." It appears to be the first Tax Court case to directly address AI-hallucinated legal authority. This is an old lesson. One of the first widely reported AI stories, in 2023, was Mata v. Avianca, where a federal judge in New York fined two lawyers $5,000 for filing a brief with cases ChatGPT had made up. Three years later, the same thing is still happening.
Why it matters: CPAs face the same exposure under Treasury Circular 230 and the AICPA's Statements on Standards for Tax Services. Both require independent verification of AI-generated authorities and hold the practitioner responsible regardless of what the tool produced.
What to do this week: if your firm uses AI for tax research or memo drafting, confirm in writing that someone independently verifies every cited authority before it reaches a filing.
Source: Journal of Accountancy, Sept. 1, 2026 (Clinco, T.C. Memo. 2026-16); Mata v. Avianca, S.D.N.Y., sanctions order June 22, 2023
The CFO Operating Brief is published by PCS Insight, LLC. Content is general operational guidance for finance leaders and does not constitute investment, legal, tax, or accounting advice.