Finance Teams Are Spending More on AI in 2026 — But Only 1 in 5 Call It a Success

Auditoria.AI released its seventh annual State of AI Automation in the Finance Office report this week, surveying around 300 finance, accounting, and finance-technology professionals. The headline finding: investment is climbing steadily, results aren't.

The gap between spending and success

66.5% of finance organizations are increasing their AI investment this year, and 24.2% now treat it as a top budget priority. But when asked whether their AI initiatives actually worked, only 21.0% reported meaningful, measurable results. 64.8% described mixed or unsuccessful outcomes, and another 14.2% haven't attempted an AI initiative at all.

Maturity tells the same story. 58.4% of finance teams are still exploring or piloting AI rather than running it in production, and only 12.8% have reached the optimizing or autonomous stage. Just 2.5% describe their finance operations as fully autonomous. The average team now touches 2.43 different finance functions with AI, up 36% from last year — adoption is spreading even where it isn't yet paying off.

The number that stands out: manual follow-up got worse, not better

This is the counterintuitive part. The share of finance teams spending 11 or more hours a week chasing vendors, customers, and internal stakeholders held in the 17.6%-29.7% range every year from 2022 through 2025. In 2026 it jumped to 44.5% — the single largest year-over-year change anywhere in the report. The share spending more than 30 hours a week on follow-up alone more than doubled, from 5.6% to 13.4%.

Data entry and correction, repetitive manual tasks, and pulling numbers out of documents are the three things people complain about most. And the most common daily frustration, named by 22.1% of respondents, is simply dealing with vendor or customer records that are wrong or half-filled-in.

What's actually blocking progress

For years, the standard explanation for slow AI adoption was that people didn't trust it. That's no longer the top barrier. Getting new AI tools to work with existing systems now ranks first, cited by 36.7% of respondents — just ahead of upfront cost (36.3%), security and privacy worries (34.5%), and trouble finding a tool that actually fits (33.8%). Resistance to adopting the technology in the first place has slid all the way down to seventh place.

Asked directly why a specific AI project underdelivered, 29.2% of all respondents — and 36.6% of those who reported a shortfall — pointed to the technology itself failing to perform as promised. Getting buy-in from the team came in well behind that, at 21.7%.

Where control is actually shifting

The report also tracked how much decision-making finance teams are willing to hand over. A third of respondents, 33.1%, still sign off on every single AI-recommended action individually. Just over a quarter, 26.3%, let AI run the routine work on its own and only step in when something looks off. A smaller group, 13.5%, has gone a step further into what the report calls "governed autonomy" — AI carrying a task through to completion inside pre-set rules and compliance limits, without a person checking each individual step.

What this means if you're choosing finance AI tools right now

The barrier has moved from "will my team trust this" to "will this actually plug into what we already run." For a bookkeeping, month-end close, or financial reporting tool, that makes ERP-integration depth (Workday, Oracle, SAP, NetSuite, and similar) a more useful question to ask a vendor than raw AI-capability claims. The tools worth evaluating closely are the ones that can show a defined, auditable path from "AI drafts it" to "AI executes it under rules a controller actually set" — because that gap, more than any feature list, is where this year's survey says most finance AI projects are getting stuck.

Sources:

• GlobeNewswire — "New Report Says Finance Teams Are Speeding Up Tasks with AI, But Failing to Fix The Cash Cycle" (Auditoria.AI, Sept. 16, 2026)