Today we're publishing the CFO AI Leverage Report, Edition 1 — the first release of the Enterprise AI Buying & Budget Index, Open Future Forum's research program on how executives who hold AI budgets are actually buying and funding AI.
The report reads one question above all others: are finance leaders using AI to grow output without growing the team? That's the CFO AI Leverage Index. Edition 1 is an early instrument — we're building the methodology and publishing the baseline before the story is settled.
What's in it
The report draws on data from 200+ executive registrations across 20+ Open Future Forum events in 2026. Finance and security are the two deepest operating-executive lanes we run, and the finance lane has been consistent across the whole year — not a single spike. That consistency is what made it worth studying properly.
Five theses sit at the center of the report. Enterprise GenAI spend hit $37 billion in 2025 and the share funded from experimental innovation budgets fell from 25% to 7% in a single year — AI has moved from experiment to permanent budget line. Finance is becoming the sign-off seat for AI purchases, not IT. Founders haven't figured out the finance buyer yet; not one of the sixteen AI founders we spoke to named finance as their buyer. Outcome-based pricing — the model that makes finance the natural buyer — is already where most AI founders are going. And the headcount-leverage thesis is the strongest draw in our finance rooms.
All figures show their base. Early reads are labeled as directional. A small, honest claim beats a large, shaky one.
Why we built this
Most AI market research reads the market from above — analyst estimates, vendor surveys, top-down projections. We're in the room. We run the dinners. We hear what finance leaders say off the record about where they're spending, what's working, and what their boards are asking.
The Index is designed to track this over time. Each edition covers one executive function in depth. Finance is first. Security and growth lanes follow. The URL stays stable so the data can be cited and compared edition to edition.