AI in Tax and Finance: Accuracy First, Value at Scale
Tax and finance functions are at an inflection point. Generative and agentic AI can reshape how work gets done, but only if accuracy, trust, and data quality lead the way. The 2025 EY TFO survey shows the tension clearly: 86% of leaders rank data, AI, and technology as top priorities, yet 80% still face foundational data issues.
That gap matters because these teams operate under strict regulatory scrutiny. If outputs aren't accurate, the risk isn't theoretical-it's compliance, audit, and reputation.
What CEOs Expect Now
Top-performing CEOs are treating today's volatile, interconnected environment as a catalyst for action. They're asking tax and finance to move beyond compliance and deliver insight into strategy-scenario planning, deals, and supply chain shifts. In fact, 79% identify this as a top priority over the next two years.
The opportunity: use trusted data and AI to create clarity, inform faster decisions, and keep the business ahead of change. That means treating transformation as an ongoing operating discipline, not a one-off project.
Deterministic vs. Probabilistic: Why It Matters
Most tax and finance processes rely on deterministic models that return the same answer from the same input-ideal for compliance. GenAI and agentic AI are probabilistic; they interpret ambiguity and incomplete data, which can lead to different outputs for the same input. That's useful for analysis and drafting, but it raises accuracy concerns.
The upside is real: teams expect up to a 30% effectiveness boost and 23% more budget freed for high-value work. The blocker is data. Eighty percent cite insufficient AI-ready data, only 17% (tax) and 13% (finance) rate themselves "very effective" at managing data, and 91% say data lives in too many silos-often on local drives.
Leaders are centralizing and standardizing data on modern AI data clouds, making it accessible across systems. They also integrate solutions across processes instead of stacking point tools.
Adoption, In Practice
Most teams are early. Seventy-five percent are in the first phases of deploying GenAI, only 21% find it easy to build tax tech apps, and 39% feel moderately or fully prepared to deploy AI agents. As a result, 78% are working with third-party providers investing at scale.
The focus is shifting from single tools to orchestrating AI agents into end-to-end processes that collaborate, escalate, and execute with clear controls and audit trails.
Agentic AI: Where It Lands First
Near-term impact is expected in data acquisition and cleansing (81%), automating tax compliance (79%), and analytics plus scenario planning (70%). Fewer leaders see immediate benefits in modeling legislation or delivering deeper strategic insights (27%), and only 10% expect big gains in closing the talent gap-for now.
Expect rapid movement as confidence grows. Some leaders already use AI to triage daily tax and regulatory updates, accelerate memo prep, and reduce manual review with rule-based checks plus anomaly detection-while keeping human review for accuracy and compliance.
External Pressures Are Rising
Change outside the function is accelerating. Eighty-one percent plan moderate to significant shifts to how they run their business over the next two years, including supply chain changes, due to geopolitical pressure. Twenty-six percent expect significant changes-more than twice the prior year.
Pillar Two is front and center: 81% say it's the top regulatory shift potentially affecting their business, yet only 21% feel very prepared for BEPS 2.0 global minimum tax reporting. For reference, see the OECD's Pillar Two overview here.
Tax transparency is also expanding. Two years ago, 37% planned to publicly disclose total taxes paid; now it's 80%. Meanwhile, authorities are leaning into real-time data and analytics, raising the bar for accuracy and alignment across public and private reporting. Forty-five percent cite the lack of a sustainable data, AI, and technology strategy as their top barrier.
Trust in Your People
Technology needs talent to match. Sixty-one percent expect senior tax retirements to hit hard, and 66% believe fewer accountants entering the field will be a detriment. Today, internal teams spend 53% of their time on routine tasks and 16% on highly specialized work.
The goal is a flip: reduce routine to 21% and push specialized up to 34%. Leaders are acting-89% are upskilling, 81% are hiring for skills beyond tax technical, and 62% are redefining roles. Co-sourcing is helping too; 85% say it improves focus on high-value activities.
The Data-First Playbook for Continuous Transformation
- Build a single source of trusted, granular data. Centralize, standardize, and govern data so it's accessible, auditable, and AI-ready across systems.
- Equip your people to work with AI and data. Upskill on prompt strategy, data literacy, agent orchestration, and human-in-the-loop review.
- Set clear standards and governance. Define controls, testing, documentation, and model risk guidelines that support fast iteration without losing accuracy.
- Tighten cross-functional collaboration. Align tax, finance, IT, legal, and operations so processes flow end to end-and the tech stack reflects that.
- Partner to accelerate. Use providers with proven AI agents and integration capabilities; focus on outcomes, adoption, and measurable ROI.
90-Day Starter Plan
- Data inventory and prioritization: Identify top 3 domains (e.g., indirect tax, transfer pricing, close). Map sources, owners, and gaps. Eliminate local-drive storage.
- Quick wins: Pilot AI for document search/classification, compliance memo drafts, and anomaly detection with human review checkpoints.
- Agent pilot with a partner: Orchestrate an AI agent for data intake → validation → calculation → draft workpapers. Track accuracy, cycle time, and rework.
- Governance sprint: Stand up AI usage policies, approval workflows, and lineage tracking. Decide what's in-scope vs. out-of-scope for GenAI.
- Upskill core team: Run short workshops on prompts, evaluation, and error analysis. Assign owners for agent maintenance and metrics.
Where This Goes Next
As data quality improves and governance matures, expect broader use of agents across compliance, controversy support, scenario modeling, and reporting alignment. The result: faster cycles, fewer manual touches, and more time on advisory work where judgment matters.
If your team is building practical skills for AI in finance, here's a curated list of tools that can help pressure-test use cases: AI tools for finance.
Your membership also unlocks:
