Five stories this week, in priority order.
Story 1: AI decision-making, not AI use, separates working capital leaders from laggards
PYMNTS Intelligence released its fourth annual Working Capital Efficiency Index on September 28, based on CFOs and treasurers at North American firms with $50 million to $1 billion in revenue. The index fell 6% to 51.6, its first decline in four editions. Early customer receipts dropped to 12% of receivables from 35%.
AI use no longer separates anyone. 100% of top performers and 97% of bottom performers use AI in treasury. What separates them is what they would let it decide. 93% of top performers would let AI forecast a cash shortfall, against 69% of bottom performers. Nearly six in ten top performers would let it decide when to draw a credit line, and one in five would let it execute a transaction above $100,000, which no bottom performer would. Top performers run a 39-day cash conversion cycle against 63 days.
Why it matters: having an AI tool in treasury is now table stakes. The leaders are the ones that trust it with decisions, and their cash cycle is 24 days shorter.
What to do this week: ask your treasury lead which cash decisions your AI tool makes today and which it only reports on. For each one it only reports on, decide whether that is a deliberate control or a habit.
Source: PYMNTS Intelligence, Working Capital Efficiency Index, Fourth Edition: North America, released Sept. 28, 2026 (CFOs and treasurers at firms with $50M–$1B revenue)
Story 2: AI-generated invoice fraud is getting more sophisticated
A September 10 disclosure, not a this-week finding. Microsoft's Security Blog reported a campaign detected August 3 to 5 that sent more than a million emails, 87.7% to U.S. recipients, impersonating CEOs to push AP staff toward an ACH payment of about $50,000. Microsoft says the attackers used generative AI to build the email templates. Each email carried a fabricated invoice branded as ServiceNow (which was not involved) and a fake "forwarded" thread between the spoofed CEO and a spoofed ServiceNow executive, built to make the request look pre-approved before AP ever saw it.
Why it matters: the campaign stacked executive impersonation, vendor branding, a fake invoice and a fake approval trail into one story. It was built to pass the "does this look legitimate" test, which is still how most AP teams judge a request under deadline pressure.
What to do this week: check the tools and the practice. Tools: confirm your email domain enforces SPF, DKIM and DMARC, and that your email filter can pull a phishing message back after delivery (both Microsoft recommendations). Practice: any payment request that arrives with its own supporting "conversation" gets verified through a contact you already had on file, never by replying to the email. Make it a written AP rule.
Source: Microsoft Security Blog, "Protecting organizations from AI-assisted executive impersonation and invoice fraud," published Sept. 10, 2026 (campaign observed Aug. 3–5, 2026); older than two weeks, not a this-week finding
Story 3: Audits will increasingly focus on AI accountability
Gartner released its 2027 Audit Plan Hot Spots on September 28, based on a May–June 2026 survey of 190 audit leaders. The top theme for internal audit next year is pressure to realize AI value, ahead of IT governance and operational resilience. Issue 6 covered the other side of this: Gartner found 93% of audit functions already use AI and 54% of chief audit executives have not started measuring what it delivers. Next year's audit plans are aimed at that gap.
Why it matters: AI went into finance and audit before the accountability structure to govern it. Auditors are now being told to close that gap in 2027.
What to do this week: assume your auditors will ask where AI touches your numbers and who answers for it. Have three answers ready before the next audit cycle: which finance processes use AI, who reviews each output before it reaches a filing, board deck or close entry, and what each tool has delivered. If you cannot answer one of them today, that is the work.
Source: Gartner, 2027 Audit Plan Hot Spots, released Sept. 28, 2026 (190 audit leaders, fielded May–June 2026); audit AI-use figures from Gartner, Sept. 10, 2026, covered in Issue 6
Story 4: Smaller firms twice as likely to be held back by financing
Duke University's Fuqua School and the Federal Reserve Banks of Richmond and Atlanta released their third-quarter CFO Survey on September 23 (roughly 500 firms, fielded August 17 to September 4). Large firms grew more optimistic this quarter; small and financially constrained firms grew less so. 19% of firms say the cost of or access to financing has constrained their investment or spending plans. Firms with fewer than 500 employees report that constraint at around double the rate of larger firms. Among constrained firms, over 60% say it kept them from pursuing new business opportunities. Among unconstrained firms, nearly half say sufficient cash has made them less reliant on external financing.
Why it matters: the divide is cash on hand. For a smaller firm, it decides whether the next opportunity gets taken. Story 1 shows the pressure on it: far fewer customers are paying early.
What to do this week: list the opportunities your company deferred this year for lack of funding. Then size the cash tied up in receivables past their due date. If faster collection would have funded one of them, start there.
Source: Duke University Fuqua School of Business and Federal Reserve Banks of Richmond and Atlanta, The CFO Survey, Q3 2026, released Sept. 23, 2026 (roughly 500 respondents, fielded Aug. 17–Sept. 4, 2026)
Story 5: Leaders may not want AI in human-centric decisions like hiring
A dated benchmark, not a this-week finding. ACCA's Global Talent Trends 2026 survey of 11,389 finance and accounting professionals in 160 countries (report published May 2026, re-released by ACCA September 17) found that 48% have reservations about AI algorithms in hiring. Among board-level leaders, 54% express doubts about the growing reliance on AI to select talent. The top concern was losing the human touch, including rejected candidates getting no human feedback, alongside bias and data privacy.
Why it matters: the most skeptical group is the one that approves the technology, and the leading objection is about the human element, not accuracy. That could point to something broader: leaders may accept AI for transactions and analysis but not for decisions about people.
What to do this week: before approving AI in any people decision (hiring, performance, promotion), ask where a human makes the final call and whether a rejected candidate gets feedback from a person. If the answer to either is no, expect resistance from your own leadership.
Source: ACCA, Global Talent Trends 2026, published May 2026 (11,389 respondents, 160 countries, fielded Oct. 2025–Feb. 2026); hiring findings re-released by ACCA Sept. 17, 2026; older benchmark, not a this-week finding
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.