5 Risks CFOs Take with AI Adoption

CFOs take significant risks with AI adoption when they implement AI without a clear strategy for data quality, governance, security, and business processes. While AI can improve forecasting, reporting, and financial analysis, its outputs are only as reliable as the financial data and controls behind them. Successful AI adoption in finance requires trusted data, human oversight, and strong governance to reduce compliance, operational, and decision-making risks while delivering measurable business value.

Most of what gets written about AI in finance focuses on the upside: faster closes, sharper forecasts, leaner teams. Less gets said about the risks of AI adoption for CFOs, who are expected to fund these projects while still owning the numbers if they go wrong.

5 Risks CFOs Take with AI Adoption

Before signing off on the next AI initiative, here are five risks CFOs take with AI adoption, and what a more responsible AI strategy looks like in practice.

Risk #1: Unclear ROI Makes AI Adoption in Finance a Gamble

One of the most common AI risks in finance is simple: the payoff often doesn't show up when expected. PwC's latest Global CEO Survey of more than 4,400 chief executives found that 56% saw no increase in revenue or reduction in costs from AI investments in the past year. Only around three in ten reported higher revenue, and roughly a quarter reported lower costs.

EY's research tells a similar story, specifically from the finance seat. Most CFOs surveyed said standard ROI metrics don't hold up well against AI and other emerging technology, since traditional financial frameworks weren't built to capture benefits that are indirect, delayed, or hard to quantify, such as faster decisions or better forecasting accuracy. The CFOs who struggled most to justify AI budgets said proving return on investment upfront was their single biggest obstacle.

That said, the payoff is real for companies that get it right. Firms that concentrate their AI efforts on a small number of business areas, rather than spreading investment thin, have reported meaningfully faster paths to breakeven and stronger earnings growth. For CFOs, the lesson is to treat early AI investment as a learning cost, use pilot projects to contain exposure, and where possible negotiate vendor contracts that tie payment to measurable outcomes rather than upfront licensing fees.

Risk #2: Weak Governance Opens the Door to Shadow AI Agents

Among the challenges of AI adoption in finance, ungoverned deployment may be the most underestimated. A recent Cloud Security Alliance survey found that the large majority of enterprises had discovered AI agents running inside their environment that leadership didn't know existed. Each one represents spending nobody approved, data access nobody reviewed, and potential compliance exposure nobody flagged.

This tends to happen when individual teams adopt AI tools independently, connect them to internal systems or customer data, and skip the usual approval process because the tools are easy to access and quick to set up. For a CFO, that turns into unbudgeted financial commitments and operational risk hiding in plain sight.

Reducing this exposure starts with basic governance questions before any tool goes live:

• Is it compatible with core financial systems?

• What company data can it touch?

• Do the people using it have the training to catch mistakes?

• Does it introduce new cybersecurity exposure?

A standing review group that includes security, compliance, technology, and finance can catch most of these issues before they become expensive.

Risk #3: AI-Ready Data Is Harder to Build Than It Looks

AI is only as useful as the data behind it, and this is where many finance functions are furthest behind. EY's research found that data quality and bias are the top barrier CFOs cite when trying to secure investment in new AI tools, with roughly six in ten calling it a significant obstacle. Unclear or long-term benefits and a shortage of in-house skills followed close behind.

The complication is that useful AI in finance rarely runs on finance data alone. Solving a supply chain problem, for example, might require pulling in manufacturing, tax, customs, and legal contract data that was never built to talk to each other. Aggregating and cleaning that information is often more expensive and time-consuming than the AI tool itself.

CFOs building a responsible AI strategy should treat data architecture as a prerequisite for scaling AI, not something to patch after the fact. That means budgeting for data cleanup and integration work before, not after, committing to a wider rollout.

Risk #4: Losing Institutional Knowledge and Trusting the Black Box

Two related risks of AI adoption for CFOs show up once tools move from pilot to daily use. The first is the quiet erosion of institutional knowledge. When AI takes over forecasting or analysis that used to force a team to dig into the underlying business, the team stops building the judgment that comes from doing that work by hand. Automating a forecast is easy; replacing the understanding a finance team gained while building it manually is not.

The second is the black box” problem: AI-generated insights that come without a clear, traceable explanation of how they were reached. When users, customers, or auditors can't see the reasoning behind a recommendation, trust breaks down fast, and finance leaders end up double-checking the output anyway, which erases the time savings.

Forecasting is a particularly high-stakes place for this to go wrong. Because it directly shapes decisions on budgets, hiring, and capital allocation, most finance leaders agree it shouldn't be handed to AI without a human validating the result. AI can speed up the data gathering and produce a useful first draft, but a person still needs to confirm the underlying numbers are real and the logic holds up.

Risk #5: Regulatory Complexity and Employee Pushback

Rounding out the list of AI risks in finance are two people-and-policy problems that are just as costly as the technical ones. On the compliance side, data security and privacy consistently rank as the top concern finance executives raise about AI adoption, according to Protiviti's research. Regulations covering data privacy, cybersecurity, and intellectual property vary by jurisdiction and sometimes conflict outright, which makes compliance a moving target for any company operating across borders.

On the people side, employees tend to respond to AI adoption in one of two ways: as a threat to their job, or as a chance to build new skills. When leadership rolls out AI top-down without input from the people who run daily finance operations, the result is often automation layered onto broken workflows, plus tools nobody trusts enough to use. That shows up on the balance sheet as sunk implementation costs and stalled adoption.

The fix for both issues is involvement. Legal, compliance, and cybersecurity specialists should be part of AI decisions from the start, not brought in after a tool is already live. And employees at every level need a chance to raise concerns and understand how the rollout affects their role, rather than finding out after the fact.

How CFOs Can Reduce AI Risks and Build a Responsible AI Strategy

Treat Governance as a Prerequisite, Not an Add-On

An AI strategy for CFOs should start with a governance framework, not a tool. Define who approves new AI use cases, what data they can access, and how spending gets tracked, before teams start experimenting on their own.

Redefine What Success Looks Like

Because standard ROI metrics undersell AI's value, pair financial measures with qualitative ones, such as time freed up for higher-value work or improvements in forecast accuracy, to get a fuller picture of impact.

Fund Data Readiness Before Scaling

Budget for data cleanup and integration as its own line item. AI-ready data is the foundation on which every other part of a CFO AI strategy depends.

Keep Judgment in the Loop

Use AI to speed up research, drafting, and data-gathering, but keep a person accountable for high-stakes calls like forecasts, payments, and anything customer-facing.

Make the Rollout a Two-Way Conversation

Bring finance staff into the planning process early. Understanding their concerns and addressing them directly does more for adoption than any dashboard or training video.

None of this is a case against AI adoption in finance. It's a case for going in with eyes open. The CFOs who get the most out of AI tend to be the ones who treat governance, data readiness, and employee buy-in as seriously as the technology itself. Avoiding AI implementation failures isn't about moving slower than everyone else; it's about pairing the same ambition with real discipline, so the investment actually pays off instead of becoming another line item nobody can explain.