Chief financial officers (CFOs) who see AI primarily as a cost-reduction tool are optimizing a horse while others are building highways. The instinct to measure AI against headcount savings misses the more consequential question entirely: What can the finance function now do that was previously impossible at its scale?
Mark Beltran, a fractional CFO serving software as a service (SaaS), AI, and Web3 companies from Seed to Series C, has built his entire practice around that question. Without AI, the analytical work he delivers across several portfolio companies simultaneously would require a 12-person team. With AI and experienced leadership, it takes a lean, distributed operation moving at a pace that becomes a competitive moat. “Cost reduction is a rear-view mirror play,” Beltran states. “The CFOs who only see cost reduction are optimizing a horse. The ones who see operating leverage are building a highway to visit distant lands.”
AI Does What Spreadsheets Were Never Built to Do
A spreadsheet is a scalpel; precise, powerful, and narrow. It does exactly what it is told with exactly the data it is given. It cannot simultaneously read a cap table, order to cash data, a draft term sheet, and a hiring plan and produce a coherent narrative that connects all four into a capital raise decision. “Spreadsheets model what you know,” Beltran reflects. “AI helps you see what you’re missing. For a founder walking into a Series B, the difference between those two could be existential.”
When Beltran inherited a client’s cash flow model with seven structural failures, reference errors, flat lines beyond 2027, and broken assumptions, a spreadsheet analyst was able to locate the errors. What AI enabled was a rapid, full diagnostic, a clear articulation of why each failure mattered to the investor narrative, and a complete reconstruction path, all in a single working session that would previously have taken days.
In a separate engagement, a client with disorganized financials preparing for an acquisition moved from messy cash-basis books to acquisition-ready in approximately one week, with AI-built agents converting general ledger (GL) data into investor-grade output while simultaneously pattern-matching against what an acquirer’s diligence team would scrutinize. “AI is like a whole surgical team coordinating with purpose in real time,” Beltran notes. The scalpel is still necessary. But the surgical team wins the complex cases.
Finance Gets AI Wrong in Two Ways
The finance function is the slowest major business function to adopt AI, and the reasons are structural rather than incidental. The first problem is conflating precision with perfection. The professional instinct to care about the last decimal place is correct in audit and reporting contexts. Applied to AI adoption, the same instinct leads teams to reject a tool that occasionally produces a 90% first-pass answer because it is not 100% accurate in every output. “The right question isn’t: ‘Is it always right?'” Beltran challenges. “The right question is: ‘Does it get me to the right answer faster with my review as the control?'”
The second problem is treating AI as a tool rather than a workflow redesign opportunity. Finance teams add an AI subscription, use it for one-off questions, and declare it marginally useful. Beltran is unsparing about what that looks like in practice. “That’s like buying a Formula One engine and dropping it into a school bus.” The leverage is not in the tool; it is in rebuilding the workflow around what the tool makes possible. The functions winning with AI right now – sales, marketing, and engineering – did not add AI to existing processes. They redesigned around it. Finance must be willing to do the same within its regulatory and audit obligations, identifying clearly what gets automated, what gets delegated, and what remains the irreplaceable human contribution.
The Finance Function of 2028 and What Will Not Change
By 2028, a well-architected finance function will operate on a continuous, close, rolling, agent-managed basis, with variance flags surfacing in near real time. A chief executive officer (CEO) will ask what happens if the Series B is pushed six months and two more engineers are hired instead, and the chief financial officer (CFO) will have a fully recalculated 18-month model within minutes during the conversation.
Agents will span the entire financial data stack, enterprise resource planning (ERP), revenue recognition, order to cash, payroll, cap table, customer relationship management (CRM) pipeline, synthesizing across all sources without a human manually building Excel bridges between them. “The most stunning thing will not be a single capability,” Beltran reflects. “It will be the compression of the feedback loop between financial signal and business decision.”
Today, a CFO receives a close package 10-15 days after the month-end. The paradigm shifts and the 2028 question is no longer what happened last month, but will be what is happening right now and what will happen if this trend continues. Roughly 80% of close tasks can be automated with agents today. The human role over the next two to three years compresses from doing the work to reviewing output and owning the judgment calls. What will not change is the role of the CFO as the person whose judgment, credibility, and accountability stand behind the numbers. “The 2028 CFO will be more strategically visible, more analytically lethal, and less buried in mechanics,” Beltran states. “Which is frankly what the role was always supposed to be.”
Follow Mark P. Beltran on LinkedIn for more insights on AI in finance, financial planning and analysis (FP&A) transformation, and building the lean, high-leverage finance functions that give growing companies a genuine competitive advantage.
