AI Agents in Finance: Use Jumps to 41% in Zapier Survey

Zapier’s 2026 survey shows AI agents in finance rising from 24% to 41% of teams, the fastest gain of any department.

AI agents in finance: Zapier survey

Use of AI agents in finance rose to 41% of enterprise finance teams in 2026, up from 24% a year earlier, according to Zapier’s 2026 AI Agents survey of more than 500 enterprise leaders. The 17-point gain is the largest of any department in the survey.

Finance was the department least likely to use agents in 2025. It has now overtaken sales, at 37%, and marketing, at 39%, and sits just behind human resources, at 43%, and DevOps and engineering, at 44%.

The shift comes as companies move agents out of trials and into daily work. Across all departments, 71% of leaders said their organisation runs AI agents in production, up from 40% in 2025.

Accountio.The go-to weekly newsletter, straight to your inbox.Weekly newsletter, straight to your inbox.

What Zapier’s survey found about AI agents in finance

Zapier, the workflow automation company, ran the survey for the second year running. It asked enterprise leaders which departments use AI agents and how far those agents have moved beyond testing.

Finance recorded the biggest increase of the seven departments tracked:

  • Finance: 24% in 2025, 41% in 2026
  • Human resources: 30%, then 43%
  • Operations: 47%, then 59%
  • Customer support: 49%, then 57%
  • DevOps and engineering: 35%, then 44%
  • Marketing: 31%, then 39%
  • Sales: 26%, then 37%

Operations and customer support remain the heaviest users, at 59% and 57%. Both departments handle high volumes of repeat requests, which made them early candidates for automation.

Overall, 88% of organisations now use or test AI agents, up from 72%. The share still piloting fell from 32% to 17% as more teams moved into production.

Investment is set to keep rising. Nine in ten leaders said their organisation is certain or likely to increase spending on AI agents over the next year, up from 84% in the 2025 survey.

Why finance moved fastest

Zapier’s chief financial officer, Ryan Roccon, put the jump down to better models and closer links with the spreadsheets finance teams already use, such as Microsoft Excel and Google Sheets. He said the models available a year ago were not accurate or reliable enough for finance work.

Accounting also suits automation. Many finance processes follow fixed rules, and their output can be checked against a ledger, a bank statement or a supplier invoice. That makes errors easier to find than in most other departments.

Roccon said companies that hold back from the technology put themselves at a disadvantage. He did not see reports of AI models acting unpredictably as a reason for CFOs to pause. At Zapier, many month-end processes now run through an AI system.

Finance also started from the lowest base. With less than a quarter of finance teams using agents in 2025, there was more room to grow than in customer support or operations, where about half already did. The 2026 figures show that gap closing quickly.

Where finance teams use AI agents today

The most common uses across all departments are document analysis and summarisation, and data extraction and entry. Both led the list in 2025 and again in 2026.

In finance, those map directly onto daily work. Agents read supplier invoices, pull figures from contracts, match payments to remittances and prepare reconciliations for review. Accountio’s look at which AI accounting tools are in production and which are still in pilots shows the same pattern, with document work and reconciliation furthest ahead.

Human review remains part of the process. Only 30% of leaders said their agents complete more than 60% of assigned tasks without a person stepping in. Zapier applies the same rule to its own books.

Nothing hits the general ledger ever without a human clicking approved.

Ryan Roccon, Chief Financial Officer, Zapier

Privacy, compliance and audit trails

Data privacy, security and regulatory compliance were the most cited barrier to adoption, named by 27% of leaders. Accuracy, reliability and errors came next at 24%.

Both weigh heavily in finance. Finance teams hold payroll, banking and customer data, and their output is audited. In the UK, the US and the EU, that data falls under different protection rules, so a workflow built for one office may need changes before another can use it.

Under UK and EU GDPR, payroll records and supplier contact details processed by an agent are personal data. Finance teams need to know where an agent sends that data and which provider processes it.

Governance is still catching up. A Censuswide survey for Avalara of more than 1,500 CFOs and senior finance leaders found that 46% had AI incident response plans that were untested or still in development. In the US the figure was 38%.

Audit trails are the other priority. In a Datarails survey, 75% of respondents named auditability as the top obstacle to trusting AI, as covered in Accountio’s report on AI auditability in finance. An agent that records every step it takes makes the reviewer’s sign-off faster and the auditor’s work simpler.

How to start with AI agents in finance

Roccon’s advice is to start with work that is low risk and contained. For most finance teams, that points to three first projects:

  • Extracting data from supplier invoices and receipts into the ledger or AP system
  • Summarising contracts, leases and customer agreements for review
  • Preparing bank and balance sheet reconciliations for a person to approve

Each one has a clear check. The extracted invoice matches the PDF, the summary matches the contract, and the reconciliation either balances or it does not.

A budget request for AI agents in finance should name who reviews each output, what happens when the agent gets something wrong, and where company or employee data is processed. Accountio’s guide to agentic AI in accounting, its capabilities and limitations sets out where current systems perform well and where a person still needs to sign off.

Measuring AI agents in finance

The survey’s autonomy figure gives finance leaders a useful benchmark. If fewer than a third of organisations see agents finish most tasks unaided, a finance team should track how often its own agents need a person to step in, and on which tasks.

Four measures cover most of what a controller needs to know:

  • The share of items an agent completes without a correction
  • The number and type of errors found at review
  • Hours saved at month-end close, compared with the same close a year earlier
  • The exceptions an agent passes to a person, and how quickly they are cleared

These figures show which processes are ready for more automation and which still need closer review. They also give the audit committee and external auditors a clear record of how the agents are being controlled.

Tracking them from the first project makes the second one easier to approve. A team that can show a falling error rate on invoice capture has a stronger case for moving agents onto reconciliations or accruals.

What the rise means for finance software buyers

Finance teams now have two routes to agents. General automation platforms such as Zapier connect existing applications and add AI steps between them. Accounting and close software vendors are building agents directly into their own products.

General platforms suit teams that run many separate applications and want one tool to connect them. Built-in agents suit teams that already do most of their work in one accounting or close system, since the agent sees the same data and permissions as the people using it.

The right choice depends on company size and on how many systems the finance team runs. Accountio’s comparison of month-end close software by company size shows how the specialist tools line up, and the Accountio directory lists vendors by category.

With 90% of leaders planning to increase spending, AI agents in finance are moving from early trials to standard tools. Finance has gone from last place to one of the fastest-growing departments in a single year.