Tax leaders have gone from doubting AI agents to counting on them in a single year. Two surveys released this week show the data those agents would run on is still siloed, partly manual and spread across too many systems.
Key Takeaways
A year ago, roughly one tax leader in five expected AI agents to do much for the tax function. Now more than three in four do. That is one of the fastest shifts in sentiment any finance survey has recorded this year, and it comes with an awkward footnote from the same respondents: most of them cannot yet get their own tax data into a shape an agent could use. With e-invoicing mandates moving tax reporting toward the transaction itself, and budgets holding flat, the gap between what tax teams expect from AI and what their data can support is becoming the real constraint on the function.
On November 11 EY released its 2026 Tax and Finance Operations Survey, which covers 1,600 CFOs and senior finance and tax leaders in 30 jurisdictions and 22 industries. 77% now expect AI agents to benefit the tax function, against 21% a year earlier. 83% say building an agentic workforce is important, and 81% emphasize hiring professionals with AI-related skills.
The maturity numbers point the other way. 71% place their organization at a completely nascent or exploratory stage of AI maturity, and 64% cite insufficient AI-ready data as the single biggest barrier to advancing AI initiatives. Only 16% consider their tax function highly effective at accessing, organizing, using and reusing data. Among those who are not effective, 61% blame data silos and 52% blame poor data quality. 43% say their biggest challenge is being unable to execute a sustainable data, AI and technology strategy at all.
The pressure is also arriving from outside. 64% are not fully prepared for real-time tax demands, and 79% are adapting operations in response to global developments. On EY's survey page, the firm notes that two-thirds rank shifting tariffs and trade among the developments with the greatest business impact, and 69% expect a higher tax liability from Pillar Two global minimum taxes. Stuart Lang, EY Global Tax and Finance Operate Leader, framed where that leads: "The future tax function won't separate compliance from insight."
E-invoicing is the clearest test of whether tax data is ready for real time, because the mandate requires clean transaction data at the moment an invoice is issued. A Thomson Reuters Institute analysis published the same day, drawing on its survey of indirect tax professionals in the US, Canada, Mexico and Brazil, found only 16% say their organization is fully live and compliant in at least one jurisdiction. Another 42% say their teams are actively implementing e-invoicing solutions for mandate compliance within the next 12 months.
Progress is uneven by country. In Mexico, 70% have achieved live compliance or expect to within a year, compared with about half in the US, Canada and Brazil. Canada shows the least planning: only 67% of Canadian respondents expect to reach or maintain full compliance within three years, against 80% to 90% elsewhere. Brazil leads on structure, with 89% using individual mandates to build regional or global capabilities or running a central governance model.
The underlying plumbing explains why agents are hard to deploy. More than 40% of respondents use four or more separate technology systems for compliance processes. Only 16% have fully automated data reconciliation, 58% have partial automation and 26% still reconcile manually. And only about a third have captured the compounding returns of automating accounts payable, accounts receivable and reconciliation workflows together, which the report calls the largest unrealized opportunity. An agent asked to prepare a filing from four systems and a manual reconciliation inherits every break between them.
Tax teams will not be buying their way out of this. EY's survey page shows 40% of respondents expect budgets to stay flat and 29% expect headcount to remain unchanged, so new real-time and AI work has to be absorbed by the same people. EY's answer is partly co-sourcing: among organizations that use it, 86% report a greater ability to focus on higher-value work and 78% report improved tax risk management.
The same mismatch shows up across finance. Economist Enterprise research sponsored by SAP, published November 6 and based on 2,015 C-suite executives at companies with $50 million to $1 billion in revenue, found 66% of CFOs say AI adoption is running ahead of workforce transformation. Their AI goals are about capacity rather than cuts: 41% aim to improve operational efficiency and 30% to modernize legacy systems, while only 16% cite reducing labor costs. For tax, that means AI is being asked to stretch a fixed team across more mandates, which only works if the data it reads is already trustworthy.

Guide
Only 16% of indirect tax teams are fully live on e-invoicing in even one jurisdiction. This guide breaks down country-specific mandate requirements and live reporting timelines across the EU, Latin America and Asia-Pacific.
Download
Guide
Only about a third of tax teams have captured the returns of automating AP, AR and reconciliation together. This guide shows how automating cash application, invoice matching and disputes compresses the close and improves the transaction data tax depends on.
Download
Report
With budgets flat for 40% of tax and finance teams, consolidating the four or more systems many compliance teams run needs a quantified case. This report sets out the productivity, error-reduction and close-time returns measured on Infor's cloud platform.
Download
48% of legacy ERP customers now expect to replace their system within three years, against 12% on AI-native platforms.
Fully automated accounts payable rose from 7% to 31% in a year, but only 21% are very confident they would catch a fraudulent vendor change.
A survey of 270 CFOs finds teams spend 26% of the working week verifying or correcting AI output.