Better Tools, Higher Stakes: What AI Actually Demands From CFOs

A top Wall Street law firm submitted a court filing full of fabricated case citations last month. The culprit wasn’t a careless associate. It was AI, and nobody caught it before it reached a federal judge.

Sullivan & Cromwell, one of the most prestigious firms in the country, apologized to a bankruptcy court in Manhattan after an emergency motion filed on April 9 was found to contain AI-generated errors: invented citations, misquoted statutes, and legal sources that don’t exist. The firm had policies to prevent exactly this. Those policies weren’t followed.

It’s a story about AI failure. But more precisely, it’s a story about what happens when the humans responsible for oversight stop acting like it.

AI Didn’t Lower the Standard. It Exposed Who Wasn’t Meeting It.

The Sullivan & Cromwell incident isn’t an argument against AI. It’s an argument for understanding what AI actually does well, and what it doesn’t.

AI is fast. It’s generative. It produces output that looks authoritative whether or not it is. And that quality, the convincing surface, is exactly what makes unsupervised AI dangerous in high-stakes environments.

The same week Sullivan & Cromwell was correcting its filing, Spirit Airlines was dissolving in bankruptcy court. The airline had filed for bankruptcy protection twice in under a year. Its attorneys told the court that a fuel cost spike of roughly $100 million over just two months had “engulfed Spirit entirely,” draining liquidity faster than the business could respond.

Two very different stories. Same underlying failure: the people responsible for oversight weren’t operating with a clear enough picture of what was actually happening, fast enough to act.

The Real AI Opportunity Isn’t Fewer People. It’s Better Visibility.

There’s a version of the AI story that finance leaders hear constantly: automation will reduce headcount, simplify operations, and make the finance function leaner.

That version is incomplete.

When AI reduces the time it takes to run analysis, model scenarios, or generate a forecast, organizations don’t slow down. They ask more questions. They run more scenarios. They move faster. The pace of decision-making accelerates, and the demand for high-quality financial judgment rises with it.

This is what economists call the Jevons Paradox: efficiency improvements tend to increase consumption of a resource, not decrease it. Better tools produce more output, not less work.

For CFOs and treasury teams, the implication is straightforward. AI doesn’t remove the need for real-time cash visibility. It makes that visibility more urgent. When decisions happen faster, the cost of being wrong about your cash position goes up, not down.

Spirit Airlines ran out of runway in weeks. A business with live, connected cash flow data doesn’t eliminate that risk, but it sees it coming.

What the Bar Actually Looks Like Now

The finance teams that will struggle in the next few years aren’t the ones that adopt AI too slowly. They’re the ones that adopt it without raising their own standards to match.

AI produces outputs. Finance leaders are accountable for outcomes. Those are different things, and the gap between them is where judgment lives.

That means:

Data quality matters more, not less. AI is only as reliable as what it’s trained on or connected to. Forecasts built on live bank feeds and ERP data are fundamentally different from outputs generated without grounded source data. Sullivan & Cromwell’s AI hallucinated because it had no connection to reality to check itself against. The same dynamic applies in finance.

Oversight isn’t optional. The Sullivan & Cromwell filing wasn’t caught internally. It was caught by opposing counsel. In finance, the equivalent scenario is a CFO presenting a board with a forecast that nobody stress-tested. The speed at which AI produces outputs can create the illusion of rigor where none exists.

The cadence of insight has to match the pace of the business. When GameStop makes a $56 billion unsolicited bid for eBay on a Sunday night, the treasury teams of every company watching that deal need to be thinking about what it means for their own capital position, their competitive set, and their planning assumptions. That kind of responsiveness requires infrastructure, not just instinct.

The Bottom Line

AI is raising the bar for finance, not by replacing judgment, but by accelerating everything around it.

The firms and finance teams that will benefit most are the ones that treat AI as a reason to sharpen their oversight, deepen their data connections, and move their planning cadence closer to real time.

The ones that treat it as a reason to check out will find out the hard way: in a courtroom, in a bankruptcy filing, or in a boardroom where nobody saw the liquidity problem coming until it was already too late.

The tool got better. The standard got higher. That’s the story.