What Software’s AI Shift Reveals About the Future of Tax Work

AI-driven automation is arriving in tax at a moment when accuracy, defensibility, and professional judgment have never been more consequential. As artificial intelligence accelerates preparation and analysis work, execution no longer is the primary constraint.

The new challenge for tax professionals is how responsibility shifts when work moves away from execution and toward judgment and accountability. This shift doesn’t signal a breakdown in professional judgment, but a change in where and how that judgment must be applied.

Software development offers one of the earliest and clearest views into what happens when AI compresses execution-heavy professional work at scale. AI assistants have moved quickly from concept to operational reality for developers, embedded directly inside day-to-day workflows and materially accelerating creation and iteration.

The lesson for the tax and accounting profession is about how responsibility shifts when execution is augmented and automated, not about the coding tools themselves.

Software’s AI Adoption

Software isn’t the perfect blueprint for tax or accounting, but many similarities can be observed from early signals of AI adoptions.

In software development, AI coding assistants can now generate functional code from a short prompt and move it into a pull request within minutes, shifting attention from writing code to reviewing whether it meets clearly defined acceptance criteria and scalable architecture. When those criteria are vague, teams experience rapid cycles of rework—because the intent was never precisely specified, not because the code can’t be produced.

A comparable dynamic appears in tax functions building AI-driven transaction classification tools. A model can quickly categorize thousands of ledger entries for capitalization, tax incentive eligibility, or other tax-sensitive treatments, but inconsistencies surface when capitalization policies, materiality thresholds, or exception rules were never explicitly defined. The technical build moves quickly; the ambiguity in rules becomes the constraint. In both cases, execution accelerates faster than the clarity required to govern it.

As the speed of execution increased, teams discovered that producing code no longer was the constraint. Instead, deciding whether the code was correct and ready to release was the new bottleneck.

At the same time, a second pressure became clear. AI systems execute based on the context they’re given. When requirements lack clarity, AI can generate outputs quickly but incorrectly.

This gap creates a new level of rework that many experience when they begin to adopt AI-enabled workflows. Software teams have reacted by focusing on clearer specifications and acceptance criteria. This shift is a prerequisite to making execution speed useful.

These dynamics have made human value more visible at both ends of the process. Upstream, teams get explicit about what facts and constraints must be present before work begins. Downstream, teams get explicit about the evidence and rationale that must exist before a position is approved.

Context Is Critical

Tax teams already handle large amounts of structured inputs: trial balances, transaction data, forms, returns, and workpapers. But the most influential and contextual inputs are qualitative in nature:

• What is changing in the business?
• What is material?
• What is the risk posture?
• What outcomes need to be optimized?

As AI accelerates preparation tasks—including extracting source documents, tagging and mapping tax data, synthesizing research, drafting analyses, and performing analytics—context becomes more critical. Faster execution raises the bar on defining the problem clearly before work begins.

Strong context helps generate fast execution and better insights. With weak context, speed simply amplifies ambiguity. In this environment, expertise concentrates where context and judgment matter most.

Technical Review Importance

The importance of technical review counterbalances the need for strong inputs. Historically, review has required a combination of technical expertise and a significant amount of human effort.

This is where tax leaders can learn by observing the software industry. As execution accelerates, technical review becomes the point where judgment is formally exercised and owned. In the pre-AI world, judgment and insight were applied continuously throughout the process, with multiple layers of human review as the safeguard.

If you’ve spent time in a tax department, this experience may feel familiar: The same workpaper gets reviewed by five or six people, each acting in good faith, and yet an error still makes it through. Sometimes it’s a small classification issue.

Sometimes it’s a sign error with real consequences: Negative $1 million becomes positive $1 million in a workpaper. This type of occurrence typically isn’t a competence issue. It is a reminder that tax quality has long depended on overlapping human attention and many hours of effort.

As execution scales faster, adding more reviewers applies old logic to a new problem. AI changes this dynamic by making it possible to document decisions, assumptions, and analyses in an automated way rather than requiring human reviewers who must infer the human preparer’s reasoning.

When judgment is documented, review can focus on accountability. In an AI-enabled tax function, the role of review shifts toward supporting decisions in full context, not solely catching mistakes.

This problem also becomes more nuanced for tax professionals due to the interpretive nature of tax law and the absence of a single objectively correct outcome. Unlike software, where outputs can be validated through tests, tax positions often require weighing competing authorities, risk tolerance, and scope of impact. The quality of the result depends on how well those trade-offs are understood, documented, and ultimately defended—not just on technical accuracy.

Viewing This Shift

Tax leaders don’t need to re-create their teams in the mold of a software development team to take something useful from the industry’s experience with AI. What matters is recognizing where roles shift as AI accelerates execution. The most impactful decisions move away from preparation and put higher importance on clearly articulated upfront framing and ultimate ownership of conclusions.

For tax leaders, AI can speed up familiar work. The new challenge becomes how to decide where human judgment should remain intentional, explicit, and documented. How assumptions are surfaced, how risks are evaluated, and how positions are supported become more important as automation expands.

As AI continues to accelerate execution, tax leaders should be asking different questions:

• Where does judgment need to remain intentional?
• Which decisions require explicit ownership?
• How is accountability reinforced as automation expands?

The rulebook for these questions is still being written and depends on each organization’s risk tolerance, priorities, and regulatory landscape.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law, Bloomberg Tax, and Bloomberg Government, or its owners.

Author Information

Matt Paparella is partner and tax AI leader at Crowe.

Matthew Worley is a CPA and senior manager at Crowe.

Write for Us: Author Guidelines