I spent nearly fifteen years at McKinsey and Deloitte watching technology programs succeed or fail, and the pattern that predicts failure most reliably has nothing to do with the technology. It’s who’s in the room when the buying decision gets made.
In most companies I work with, the AI conversation is owned by a combination of IT, operations, and whoever on the senior team is most personally interested in the topic. The CFO is invited to review the budget. That’s the wrong sequencing, and it’s costing companies real money.
The CFO arrives too late
Here is how the typical AI buying process works in a mid-market company. A business unit leader or IT director identifies an opportunity. They evaluate tools. They run a pilot. They build a business case. They present the business case to the leadership team. The CFO reviews the financial projections in the business case, asks some questions about payback period, and either approves or defers.
By the time the CFO sees the business case, every important decision has already been made. The workflow was selected. The vendor was chosen. The success metrics were defined. The pilot was designed and executed. The financial projections were built by the people who want the project funded, using assumptions the finance team did not validate.
Deloitte’s Q4 2025 CFO Signals Survey found that technology transformation emerged as a top strategic priority for CFOs in 2026. But Bain & Company’s survey of senior finance executives found that only 15 to 25 percent of CFOs have fully scaled AI in their own departments. CFOs care about AI. They’re just not driving it. That gap is where the value disappears.
Finance speaks a different language than technology
The case for putting the CFO in charge has nothing to do with CFOs being smarter or more strategic than CIOs. It comes down to language. CFOs speak the one that boards understand and investors require.
When a technology leader builds an AI business case, the success metrics tend to be operational: processing speed, accuracy improvement, user adoption rate, tickets deflected. These are legitimate metrics. They’re also invisible to the people who allocate capital.
When a CFO builds a business case, the success metrics are financial: cost reduction, revenue impact, margin improvement, cash flow effect. These are the metrics that appear in board materials, that inform valuation conversations, and that justify the next investment. That difference in vocabulary decides whether an AI program reads to the board as activity or as return, and only one of those gets funded twice.
The CFO is the most undertapped AI leader in your company, and the reason is structural.
Oliver Wyman’s 2026 CFO research, covering finance leaders at public companies worth roughly 12 percent of global market capitalization, found that 80 percent of them rank data, automation, and AI among their top three priorities for transforming finance. The most sophisticated finance leaders have already claimed AI as their own capability to build.
The measurement problem starts at design
In my diagnostic framework, the dimension that scores lowest across most assessments is whether AI results show up in the financials, meaning the ability to convert AI outputs into dollars the business can act on. These companies aren’t bad at finance. They score low because the finance team wasn’t in the room when the AI initiative was designed.
This matters at the pilot stage, before any scaling conversation happens. If the finance team isn’t in the room when the pilot is designed, three things don’t get built.
First, the dollar baseline does not get captured properly. The operational team knows the workflow takes too long. The finance team knows what “too long” costs. Without the finance team, the pilot measures improvement in minutes saved. With the finance team, the pilot measures improvement in dollars recovered.
Second, the cost structure does not get modeled correctly. AI tools have a cost profile that looks different from traditional software: usage-based pricing, token consumption, compute scaling. A technology leader evaluates these costs against the tool budget. A CFO evaluates them against the fully loaded cost of the workflow they are replacing, including the opportunity cost of the headcount that could be redeployed.
Third, the payback timeline does not get pressure-tested. Most AI business cases use straight-line projections: “If we save X hours per week at Y cost per hour, the annual savings are Z.” Finance teams know that straight-line projections ignore adoption curves, integration costs, change management drag, and the discount rate on uncertain future savings. Without that pressure test, the business case is optimistic to the point of fiction.
What the CFO-led model looks like
The companies I’ve seen get the most value from AI share a specific pattern: the CFO or a senior finance leader is involved before the pilot is designed, not after it concludes.
In practice, the CFO co-authors the success criteria: what “measurable” means, the financial baseline the pilot has to beat, and the evidence that would justify a production budget. That threshold gets named before the pilot starts, so the team knows exactly what they are proving.
The Consero Global 2026 CFO Survey found that fully embedded AI in finance nearly doubled in a year, from 22 percent of firms to 42 percent. And Deloitte’s Q4 2025 survey found that 87 percent of CFOs now expect AI to be extremely or very important to their finance department. The finance leaders moving fastest have learned where the value actually sits, and it sits in the measurement discipline around the tool, which is exactly what a CFO brings.
The mid-market advantage nobody uses
In a Fortune 100 company, getting the CFO involved in an AI pilot is a political act. It requires navigating matrix reporting, crossing P&L boundaries, and coordinating calendars that are booked eight weeks out. The structural overhead is real.
In a mid-market company, the CFO is often sitting in the next office. The senior team eats lunch together. The quarterly planning conversation happens in a room with six people, and the CFO is one of them. The structural advantage is proximity and simplicity.
And yet, in most of the mid-market companies I assess, the CFO’s involvement in AI decisions follows the exact same late-stage pattern as the Fortune 100: review the budget, ask about payback, approve or defer. The mid-market is copying enterprise process when it could be using its own structural advantage to do something better.
The better model: the next time someone proposes an AI pilot, ask the CFO to sit in on the design meeting. Not the approval meeting. The design meeting. Ask the CFO to name the number the pilot has to move, the financial baseline it has to beat, and the threshold that would justify a production budget. Do this before the vendor is selected, before the timeline is set, before anyone builds a slide deck. The CFO’s involvement at the design stage changes the entire shape of the initiative, and it takes one meeting.
One Meeting
Get the CFO in the room when the pilot is designed, not when the budget is reviewed. Before the vendor is selected. Before the timeline is set. Before anyone builds a slide deck.
One conversation this quarter
If you are a CEO or COO reading this, here is the move. Before your next board meeting, sit down with your CFO and ask one question: “Of the AI spending we have approved in the past twelve months, how much of it has a measurable financial return that you would put in front of the board?”
If the answer makes you uncomfortable, look at who was in the room when that spending was designed. Your most financially literate leader was almost certainly brought in too late to build the measurement. That’s fixable, and it starts with the next pilot.
How RLK Can Help
My AI Diagnostic measures whether your AI results are showing up in the P&L, and the results frequently reveal that the finance function is disconnected from AI decision-making. The AI Business Case engagement is specifically designed to build the financial framework, with CFO involvement from day one, that converts AI activity into board-ready numbers. Start the conversation.