The most useful statistic published this week in enterprise AI is a pair: 74% of finance teams now use AI to reduce manual work, but only 14% say their investments have achieved genuine "enterprise value and decision impact." That gap — adoption without returns — is the finding of OneStream's inaugural Forward Finance Index, released October 13, 2026, based on a global study of 1,300 CFOs and finance leaders.
If your organization is pouring budget into AI and struggling to point at outcomes, this study is the mirror. And the diagnosis is not "buy a better model." It is data foundations, workforce skills, and operating capabilities — in that order.
What the Index found
The Forward Finance Index benchmarks CFOs against the Forward Finance Blueprint, OneStream's proposed AI operating model for the finance function. The headline numbers:
| Measure | Result |
|---|---|
| Finance teams using AI to reduce manual work | 74% |
| Teams whose AI investments achieved genuine enterprise value and decision impact | 14% |
| Organizations with fully transparent financial data | 40% |
The 74%-to-14% conversion is the story. Roughly six in seven adopters cannot yet demonstrate that their AI investments mattered. Asked why, the study points to critical lags in three areas: data foundations, workforce skills, and operating capabilities.
The transparency figure is the one that should stop readers mid-scroll. Only 40% of these large organizations have fully transparent financial data — and finance is the function whose entire charter is trustworthy numbers. If AI in finance is being built on data that is not fully transparent in 60% of enterprises, the "decision impact" shortfall is not surprising; it is arithmetic.
The independent commentary in the release lands the same point. Robert Kugel, Executive Director, Business Research at ISG Research: "The potential for AI in Finance has never been greater, but adoption alone doesn't create enterprise value. Most organizations are scaling AI faster than the data foundation it relies on – and in Finance, that foundation is everything. As AI takes on more consequential work..."
Two caveats, stated plainly. First, this is vendor research: OneStream sells the AI operating system for finance, and the Blueprint the Index benchmarks against is its own framework. Second, "enterprise value and decision impact" is the study's own threshold — ambitious by design. Even discounted for both, a 74/14 split across 1,300 global finance leaders is a data point nobody planning an AI budget should ignore.
Why the gap persists: three lags, not one problem
1. The data foundation lag
AI amplifies whatever it is fed. Where financial data is fragmented across ERP modules, spreadsheets and legacy extracts, agents produce confident answers built on inconsistent foundations — which in finance is worse than no answer at all. The study's 40% transparency figure says most enterprises have not finished the unglamorous work: reconciled data, governed definitions, clean hierarchies, accessible history.
2. The skills lag
The Index identifies workforce skills as a binding constraint. Finance teams built around process execution are being asked to supervise systems that execute processes. That is a different job — closer to exception management, model oversight, and judgment under uncertainty — and most organizations are training for it after deploying, not before.
3. The operating-model lag
Tools do not produce value; operating models do. The 14% who report real enterprise value almost certainly share a structural trait: they changed how decisions get made — where AI sits in the close, the forecast, the reconciliation cycle — rather than bolting AI onto unchanged processes. Automating a broken process at higher speed has never once fixed the process.
Read it as a readiness audit, not a finance story
Strip out the industry specifics and the Index describes every function now being sold agents — marketing, HR, support, sales, operations. Three percentages, three questions:
- What is the state of your data? If your source-of-truth records are inconsistent or opaque, no agent layer will compensate. Fix transparency before funding another tool.
- Who supervises the machine? If nobody's role explicitly includes oversight of automated decisions, adoption is importing unmanaged risk.
- What changes when AI works? If the workflow diagram looks identical post-AI, the technology is a cost, not a capability.
Run those three questions as an audit and the 74/14 pattern becomes a design choice, not a fate.
The small-organization advantage
Here is the counterintuitive encouragement for smaller teams: the enterprises in this study are failing at the unglamorous foundation work precisely because scale makes it expensive — dozens of systems, decades of accreted definitions, political ownership of data. A 30-person company can reconcile its data in weeks, define its metrics in a meeting, and retrain a handful of people in a month. The readiness gap that takes a multinational years to close is a one-quarter project for an SMB. Read the Index's failure modes closely and you'll notice they are coordination failures, not capability failures: too many systems, too many owners of data, too little agreement on what the numbers mean. Coordination is precisely what small teams have by default.
Which means the right conclusion from OneStream's Index is not "AI is failing." It is "the winners will be the ones who did their homework first." Foundation-first is equally valid advice for the enterprise CFO and the ten-person startup.
If you want an outside eye on your own readiness — data foundations, the human oversight role, or designing an AI-assisted workflow that actually changes the diagram — that's the kind of engagement we take on. Reach out through our contact page.
Read next:
