Hemant Kapadia is the CFO at Anaplan, a leading AI-driven scenario planning and analysis platform designed to optimize decision-making.
For decades, building a finance organization followed a predictable formula: define the necessary subfunctions (FP&A, controllership, revenue accounting, treasury, internal audit and business unit finance), then determine how many people and how much software each required. Inputs divided by outputs established a unit-cost baseline that CFOs could optimize, benchmark and defend to the board.
Agentic AI is rewriting the golden ratio. In my conversations with peers, the divide is no longer between believers and skeptics; few CFOs doubt that AI will materially shape their organizations. It’s between those running pilots and those redesigning the finance function. The former may capture incremental productivity gains; the latter can build something structurally faster, leaner and more strategically valuable.
The decisions CFOs make now, including how we define unit costs, structure subfunctions and prepare talent, will shape organizational health, growth and competitive advantage.
The Evolving Definition Of Unit Cost
Historically, the cost equation was simple: headcount plus software platforms equaled output cost. Time was the less visible variable: hours spent closing the books, days waiting on reconciliations, weeks consumed by planning cycles. That time carries a cost when it keeps organizations reacting to last quarter instead of planning for the next one.
Agentic AI introduces another variable: tokens. Every agent action, from scenario analyses to anomaly flags to board commentary, consumes tokens. Unlike headcount, token costs are elastic, usage-based and variable, even for the same prompt.
A well-designed agent workflow executes tasks at a fraction of the cost and time of manual processes. A poorly governed deployment accumulates token spend without delivering accountable outputs or meaningful time savings. CFOs already scrutinize cost per transaction; that discipline now needs to be applied to AI usage.
The architecture behind the AI strategy is therefore critical. Structured computation and data management, including calculations, allocations and scenario modeling, should remain within deterministic planning systems, while agents orchestrate and interpret on top of that foundation. Using large language models only where their capabilities are required makes the token economics more sustainable.
Here’s the diagnostic I would put to any CFO evaluating an agentic transformation: Are you seeing, or credibly projecting, unit-cost reductions of 30% to 40% against baseline, including token costs? At 10% to 15%, you likely have a productivity project, worthwhile perhaps, but not enough to justify redesigning subfunctions, retraining teams and migrating workflows.
Cost reduction is only part of the return. When a close compresses from weeks to days or a forecast updates in hours, finance can operate on current information instead of stale results. Know the savings, but don’t lose sight of the value of acting sooner.
Subfunctions Reimagined
Agentic AI also raises a more fundamental question: Do the traditional subfunctions of finance still make sense?
Consider deal desk and revenue accounting: typically separate functions connected by structured handoffs that reflect the coordination cost of moving work between teams. When an agentic workflow can apply recognition logic, flag edge cases and reconcile transactions in real time, the question isn’t whether to automate the handoff. It’s whether the handoff should exist at all.
The same logic applies to business unit finance: A centralized, AI-augmented FP&A team delivering faster, more consistent support makes the embedded BU model harder to justify on economics alone.
Every function within the office of the CFO deserves this scrutiny. Some will consolidate, some will shrink, others will grow more strategically important. The job is determining which structures still serve the business, and which persist only from historical habit.
People Still Matter
These structural changes will affect finance professionals. Roles will change, and some may disappear. CFOs who pretend otherwise aren’t being straight with their teams. But the reductive “AI replaces headcount” narrative misses an important nuance: The goal isn’t to hollow out the finance function for profit, it’s to redeploy capacity toward the judgement-driven work only people can do.
As close cycles compress and reporting accelerates, finance professionals can spend less time on reconciliation and number crunching and more time on analysis and business partnership. That balance will vary by organization, but it must be intentional.
The CFOs I’ve spoken with who are furthest ahead have made AI fluency an expectation. A senior analyst may need to construct and evaluate an agentic variance-analysis workflow; a controller may need to interrogate an AI-generated reconciliation rather than accept it.
Those CFOs also give teams clear direction on what will change, where the organization is investing and how leadership will manage the transition. Organizations falling behind often have the AI tools already; what they lack is clarity about how those tools will change the work.
The Time For ROI Is Now
Organizations have spent heavily on token allocations, pilots and experiments without always defining the return, and that tolerance is expiring. For many, the ROI deadline has arrived.
The next phase of enterprise AI adoption requires the same rigor applied to any other capital decision: credible business cases, measurable outcomes and a clear accounting of costs and results. No function is better positioned to set that standard than the office of the CFO.
Finance’s golden ratio hasn’t disappeared, but its components are changing. The numerator now includes tokens alongside headcount and software. The denominator includes greater speed, capacity and outputs that were once impractical.
The CFOs who lead in this environment will define the new ratio, establish a credible baseline and hold AI investments accountable to it. That work starts now.
Forbes Finance Council is an invitation-only organization for executives in successful accounting, financial planning and wealth management firms. Do I qualify?
