There's a gap opening up inside large finance organizations, and Deloitte just put a number on it. In its second-quarter 2026 CFO Signals survey of 200 chief financial officers at North American companies with at least $1 billion in revenue, 93% said their organizations now use AI extensively or modestly across multiple functions. But when those same CFOs were asked how confident they are in their organization's AI governance, only 43% said they felt confident. More than half, 53.5%, felt only somewhat confident.
Read those two numbers together, and the picture is clear: AI adoption has sprinted ahead of the ability to control it. Three years ago, most finance leaders were still reading about generative AI or running small experiments. Now it's embedded in daily operations, and the guardrails haven't caught up. For any finance leader, that gap between deployment and confidence is where the real risk lives.
The speed of the shift is the story. In a CFO Signals survey conducted less than three years ago, 66% of respondents said their companies were still experimenting with generative AI or simply talking about it. In the latest survey, 93% report active use across key functions. That's a near-total reversal in under three years, and governance frameworks built for occasional experimentation simply weren't designed for enterprise-wide deployment.
The confidence numbers reflect that mismatch. When fewer than half of CFOs at billion-dollar companies trust their own AI governance, the problem isn't a lack of policy documents. It's a lack of operational capacity to enforce whatever policies exist. Governance in practice means someone reviewing AI outputs, catching errors before they propagate, monitoring how tools are actually used, and maintaining the judgment layer that separates a controlled process from an uncontrolled one. That work requires people, and specifically it requires high-quality finance talent with the technical depth to know when an AI output is wrong.
This is the part that gets lost in the governance conversation. A framework on paper doesn't govern anything. What governs AI is a qualified professional applying judgment to its output. As adoption climbs toward universal, the constraint shifts from "do we have a policy" to "do we have enough capable people to make the policy real." Finance leaders who treat governance as a documentation exercise rather than a staffing one are the ones most likely to land in the 53.5% who feel only somewhat confident.
When CFOs were asked to name the single biggest challenge to building an effective enterprise-wide AI governance framework, 59% pointed to the same thing: balancing the pressure to deploy AI quickly against the need to manage its risks. That tension defines the moment. Boards and competitors are pushing for fast adoption, while the risks of moving too fast – incorrect outputs, unpredictable behavior, compliance exposure – are real and rising.
The resolution to that tension is capacity. The reason "deploy fast versus manage risk" feels like a tradeoff is that most teams don't have enough skilled people to do both at once. A thin finance team has to choose: move fast and skip the review, or review carefully and move slowly. A team with sufficient high-quality finance talent doesn't face the same forced choice, because it has the review capacity to deploy quickly and still catch what matters. The tradeoff softens exactly in proportion to how much qualified human oversight you can bring to bear.
This reframes AI governance as a talent problem wearing a policy costume. The companies that will deploy AI both quickly and safely aren't the ones with the thickest governance binders. They're the ones with enough capable professionals to supervise AI-assisted work at the speed the business demands. And sourcing that capacity is where an AI talent marketplace like MAVI becomes genuinely useful, because the specific profile – technically strong, judgment-capable, comfortable reviewing automated output – is hard to find and slow to hire through conventional channels.
If the governance confidence gap is really a capacity gap, the fix follows directly. You close it by putting more qualified people in the seats where AI output gets reviewed and signed off. That's easier said than done in a tight domestic market, which is why finance leaders are increasingly looking wider for the talent that makes governance real.
An AI talent marketplace built to vet for exactly this profile changes the math. Rather than posting a role and hoping the right blend of technical depth and AI fluency turns up, you can draw from pre-vetted accounting talent already screened for US GAAP proficiency, ERP experience, and the judgment to supervise automated work. Instead of choosing between fast deployment and careful control, you add the review capacity that lets you have both. When your governance is only as strong as the people enforcing it, pre-vetted accounting talent sourced through an AI talent marketplace is one of the most direct ways to move from "somewhat confident" to genuinely in control. The Deloitte data shows adoption isn't slowing down. The question is whether your team can keep pace with it, and that comes down to who's on the team.
The survey of 200 CFOs at North American companies with $1 billion or more in revenue found that 93% now use AI extensively or modestly across functions, but only 43% feel confident in their AI governance. More than half, 53.5%, feel only somewhat confident, revealing a gap between adoption and control.
Adoption moved extremely fast. Less than three years ago, 66% of companies were still experimenting with or just discussing generative AI. Now 93% use it across key functions. Governance frameworks built for occasional experimentation weren't designed for enterprise-wide deployment, and the operational capacity to enforce them hasn't kept pace.
59% of CFOs cited balancing the pressure to deploy AI quickly against the need to manage risks as their top governance challenge. This tension is largely a capacity issue: teams without enough skilled people are forced to choose between speed and careful oversight.
A policy only governs AI if qualified people enforce it by reviewing outputs, catching errors, and applying judgment. As adoption approaches universal, the constraint shifts from having a policy to having enough high-quality finance talent to make it operational, which makes governance fundamentally a staffing question.
An AI talent marketplace can supply pre-vetted accounting talent already screened for technical depth, US GAAP proficiency, and the judgment to supervise AI-assisted work. This adds the review capacity that lets companies deploy AI quickly while still managing risk, rather than being forced to trade one for the other.