How CFOs Drive Real-Time Decisions with Trusted Agentic AI

For most of its existence, the CFO role centered on looking backward—managing budgets, reviewing audits, and closing the books. The role was built around historical accuracy, not forward motion.

That's no longer the case.

Today's CFO sits at the right hand of the CEO, guiding strategy in real time. One of the biggest challenges CFOs face is that financial data has exploded in volume, while the tools to synthesize it have lagged. Manual close processes and fragmented spreadsheets still bottleneck decisions that need to happen in minutes, not weeks.

This tension is exactly what the Financial Times explored in recent digital coverage and on its podcast, The Next Five, featuring finance leaders navigating the shift to a real-time enterprise. The consensus from those conversations points to three non-negotiables for any CFO trying to build an autonomous, real-time finance function:

• Unshakable data integrity

• Trusted, auditable governance

• Continuous human-in-the-loop control

The Shift Beyond Automation

Robotic Process Automation (RPA) streamlined static, rules-based tasks. Agentic AI is a different technology entirely. It analyzes, recommends, and increasingly executes across entire financial workflows, from record-to-report (R2R) to invoice-to-cash (I2C).

Adoption is accelerating fast. Research from Deloitte cited by the FT found that 48% of finance professionals who see themselves as playing a leading role in strategy have already deployed AI agents, compared with just 18% of those in more supportive roles. AI, which was recently seen as merely a productivity tool for finance, is now becoming a competitive differentiator.

Don’t Skip Clean Data

While AI holds great promise, generic AI tools currently fail in finance because financial data is messy. Disconnected spreadsheets, siloed ERPs, and inconsistent chart-of-accounts structures slow agents down and corrupt their output.

As BlackLine CTO Jeremy Ung put it, building trust in agentic AI requires more than buying a generic model off the shelf. It involves a framework that gives agents real operational context about how a specific business runs. Without that context, he notes, "you can't have human operators manually checking the work of potentially thousands of agents." At scale, that's the whole point: "You need a framework that delivers deterministic control."

That starts with a unified financial data model—a single source of truth agents can trust.

None of this replaces human judgment. Humans still own the critical control points like reviewing agent outputs, catching anomalies, and making the calls that carry real business consequences.

The Buck Still Stops with the CFO

No matter how sophisticated the agent, accountability doesn't move. As Gina Gutzeit, Senior Managing Director at FTI Consulting, said on The Next Five, "the buck stops here” because the CFO's name is on the financial statements.

BlackLine CFO Patrick Villanova echoed the same point from the inside: "CFOs are inherently risk averse. We have to be." That caution isn't necessarily resistance to AI, but it is the reason agentic AI in finance has to be engineered differently than agentic AI anywhere else in the business.

Deterministic vs. Probabilistic: A Line Finance Can't Cross

This is where finance parts ways with generic, black box AI. Most generative AI models are probabilistic in that they produce likely answers that can vary each time a question is asked. That's unacceptable in finance, where every number needs to be reproducible and auditable.

Villanova is blunt about the stakes: "95% right is 100% wrong in accounting and finance." A 95%-accurate agent might sound impressive elsewhere in the business, but in finance, it's a liability. SEC oversight and reporting standards leave almost no room for error.

Opaque, untraceable algorithms have no place in a function that regulators, auditors, and boards all need to interrogate. Auditable governance means every transaction match, journal entry, and agent decision has a clear trail that users can trust and compliance teams can verify. Finance requires glass box AI.

Beyond the Batched Close

For decades, finance has run on a batch cycle: close the books, report, repeat next month. Agentic financial AI, built on trusted, governed data, breaks that cadence. Variance analysis, cash forecasting, and risk monitoring shift from backward-looking exercises done after the fact to active, continuous capabilities.

Villanova has lived this shift firsthand. Reflecting on how the role has changed, he described a relationship that's grown tighter than ever: "The partnership between the CFO and the CIO has never been closer." That closeness is what real-time finance demands, since data accuracy and data speed are now inseparable from the CFO's core job.

The Strategic Co-Pilot, Realized

When cash positions, reconciliations, and risk signals are visible in real time instead of weeks later, the CFO's role shifts from reporting on the past to shaping what happens next. That's the practical payoff of all the groundwork: clean data and strong governance make real-time, board-level decision-making possible in the first place.

The three threads we introduced at the beginning tie this together:

• Clean, unified data as the launchpad

• Trusted, auditable governance as the guardrail

• Continuous human oversight as the control that makes autonomy safe

Together, they're what separates glass box agentic AI that finance teams can rely on from black box AI that just adds another layer of risk.

The CFOs and CTOs quoted across the Financial Times’ coverage are describing work already underway. The question for every finance leader now isn't whether to engage with agentic AI, but whether the data and governance foundations are solid enough to support it.

Take stock of your current data architecture. Ask where the audit trail breaks down. That assessment is the real starting point for Agentic Financial Operations.