Tax AI push stalls as 41% used no GenAI in 2025: EY

EY says 78% expect outside AI providers to help in 2 years, as GenAI use stays mostly exploratory.

AI has become a top operating-model priority for tax and finance leaders, although most functions remain at an early stage of adoption, according to a report from EY.

In its report, the firm said 78% of leaders rank increasing internal automation using data, AI, and technology as their top priority, with growing interest in generative and agentic AI to support day-to-day compliance and operational tasks.

Despite this, adoption remains limited: about 75% of functions are still in the earliest phases of deploying generative AI. In 2025, 41% reported no GenAI use at all, and a further 23% said they were still in exploratory stages, with only a small minority reporting integrated or transformative use.

Building AI capabilities internally remains a challenge. Citing survey results and external research, the report said only 21% of respondents find it significantly easy to build technology applications for their tax function, and just 39% said they are moderately or fully prepared to deploy AI agents.

It added that vendor-aligned “buy” approaches tend to deliver results more consistently than in-house builds, as many organisations struggle to demonstrate return on investment from proprietary AI development.

As a result, third-party partnerships are emerging as a key route to scaling AI. About 78% of respondents said working with an external provider with deep AI capabilities over the next two years would moderately or significantly benefit their tax function, with interviewees citing faster access to technology and quicker value creation.

Current use cases remain narrow and practical, such as filtering large volumes of regulatory updates or automating documentation-heavy compliance work.

However, leaders expect rapid change. Interviewees said AI is not yet widely used for filings or submissions, but several anticipate AI agents supporting most tax filings within the next five years.

The firm described agentic AI as the next step-change in automation, defining it as groups of AI agents that can operate independently on tasks whilst coordinating workflows, resolving data issues, and adapting decisions under human oversight.

It argued that this model could significantly accelerate automation across tax and finance processes.