A Tax Primer on Agentic AI

In case you haven’t heard, agentic AI has arrived, and many parts of the organization are putting these largely autonomous tools through their paces. Tax leaders and their teams should consider following suit. Fortunately, plenty of helpful articles and reports can help you get started.

A definition provides a good starting point: Agentic AI differs from generative AI based on its ability to operate with greater autonomy, which typically involves making decisions and taking action with minimal human intervention. AI agents “move AI from passive information retrieval to proactive execution and decision-making,” according to the KPMG report, Creating Value with AI Agents.

Microsoft catalogs hundreds of AI use cases and points to IDC survey research showing that 1) three-quarters of companies deployed some form of agentic AI last year; and 2) for every $1 organizations invest in generative AI, they realize an average of $3.70 in return.

Given the collaborative nature of the finance-tax dynamic, it also helps to consider how CFOs and their teams use agentic AI. Finance groups now leverage AI agents to provide real-time data analyses supporting the financial close (including moving toward “zero-day” closes) as well as a range of financial planning and analysis (FP&A) activities. Plus, most ERP systems at the heart of finance operations have already incorporated AI tools (generative and agentic) or will integrate them soon.

“Tasks and processes throughout the procure-to-pay (P2P) and order-to-cash (O2C) cycles along with the financial close and consolidation present prime targets for agentic AI deployment,” Protiviti Managing Director, Jim DeLoach writes in Forbes. “... Finance teams deploy AI agents to conduct forecasting and modeling of supply chain operations. Agents can also analyze the root causes of anomalies and exceptions that occurred in the previous quarter.”

This CFO Dive article features insights from an EY expert, who describes the new mind shifts, organizational designs and ROI calculations leaders need to optimize agentic AI investments.

And this CFO.com article highlights how PwC’s tax team leverages agentic AI. U.S. AI Tax Leader, Dom Megna, says building trust with tax departments when using agentic AI tools remains crucial. His group accomplishes this by “starting with governance,” clearly identifying risk tolerances and keeping humans in the loop: “We don’t just push a button and accept the result,” he tells CFO.com. “Our agents still operate within a human-reviewed, multi-eye process. Especially for complex outputs, like tax calculations, we maintain strong oversight. For simpler diagnostics, you might not need three layers of review, but you still keep that human in the loop.”

Megna also encourages his groups to experiment with AI applications. “No single tool solves everything,” he adds. “So, we want teams trying new tools, giving feedback and applying them thoughtfully.”

In another post, I share considerations and actions tax groups can implement when trying out new AI tools. At the core, Agentic AI invites us to reflect on and evolve current processes or ways of working – opening the opportunity to reimagine how work is done with Agentic AI as an enabler.