How Will AI Transform Accounting and Tax Work?

TL;DR

AI can automate much of the repetitive document review, data transfer, arithmetic, and evidence collection performed by accounting firms. Its most important advantage is the ability to interpret words and documents, helping systems substantiate tax and audit outcomes rather than merely calculate them, while human expertise remains valuable for nuanced cases and reliable results.

Transcript

From my perspective, I understand why it's hard to imagine a post big four, a post accounting firm world. It is a foreign concept. Conversely, I don't actually think they've earned the right to maintain the reverence that they've they've been granted for all of these decades, centuries sometimes. [music] And so, if this keeps going the way that I t... Read More

Key Insights

• Accounting work is often a document-processing workflow in which junior employees copy data between records, enter it into spreadsheets, and perform formulas. This structure makes many routine tasks suitable for automation, especially when the required calculations and transformation rules can be described clearly.
• The difficult part of an R&D tax credit is substantiation, not arithmetic. A company must support claims about qualifying employee activity with contemporaneous documentation, which may include Git issues, Jira tickets, interviews, or other records showing how workers spent their time.
• R&D tax incentives encourage organizations to conduct research domestically, including creating or improving products, processes, or techniques. Eligibility analysis therefore requires identifying qualifying activities, estimating or determining employee time, and connecting those conclusions to evidence that can withstand scrutiny.
• Modern AI works beyond earlier task automation because it can interpret words and read varied documents. That capability helps automate workflows involving unstructured evidence, although the difficulty differs by business because software, manufacturing, architecture, and engineering companies may track employee activities in different ways.
• Frontier models are described by Dominic Vitucci as already outperforming junior and mid-level accounting workers in some tasks. He also believes they are approaching or exceeding certain senior technical experts, while emphasizing that current systems will continue to improve beyond their present capabilities.
• Onshore is designed to automate accounting-firm services through AI models, beginning with areas such as corporate tax and expanding toward audit and advisory work. Its stated purpose is to remove repetitive tasks while ensuring customers still receive the expertise and outcomes for which they are paying.
• Large accounting firms may mistake technology purchasing for operational transformation. Examples in the discussion include spending on robotic process automation tools, employee certification programs, and millions of dollars in Copilot licenses without establishing useful workflows or demonstrating that the tools materially changed daily accounting work.
• Software engineering capacity is essential for meaningful AI transformation inside accounting firms. The discussion describes technology initiatives with no software engineers and recounts how Grant Thornton had only one employee with a software background across its United States workforce before forming a small tax innovation team.

Questions & Answers

Q: What parts of accounting work can AI automate?

AI can automate repetitive accounting activities such as reading documents, moving data into spreadsheets, applying formulas, reviewing records, and collecting supporting evidence. Onshore focuses on work performed by accounting firms in tax and aims to extend its approach to audit and advisory services. The intended result is less time spent on superfluous execution and greater reliance on expertise where judgment remains necessary.

Q: Why is R&D tax credit work difficult to automate?

R&D tax credit work is straightforward arithmetically, but it is difficult because the claimed activity must be substantiated. The accounting provider needs evidence showing how employees spent their time and whether their work involved creating or improving a product, process, or technique. Relevant support can include Git issues, Jira tickets, interviews, and other contemporaneous documentation maintained by the company.

Q: How does AI improve on older accounting automation tools?

Older automation could handle repetitive, predefined tasks, but modern AI can also understand words and interpret documents. That makes it more useful for accounting workflows built around varied records and narrative evidence. Earlier efforts discussed in the interview included robotic process automation, Alteryx licenses, and Automation Anywhere certificates, while frontier models can address work that previously required more human reading and interpretation.

Q: What does Onshore use AI to do?

Onshore uses AI models to automate work currently delivered by accounting firms. Its activities include corporate accounting and tax, with R&D tax credits presented as a flagship product, while audit and advisory services are identified as possible future areas. The company aims to eliminate repetitive tasks so customers can obtain needed expertise and outcomes without paying for unnecessary manual processing.

Q: How can companies substantiate R&D tax credit claims?

Companies can substantiate R&D tax credit claims by connecting employee activity to contemporaneous documentation. For a software business, that evidence may include Git issues and Jira tickets that demonstrate time spent on qualifying work. Manufacturing, architecture, and engineering companies may require different records because their activity tracking can be less direct, making evidence collection and interpretation more nuanced.

Q: Can AI replace junior and mid-level accounting staff?

Dominic Vitucci believes current frontier models are already as capable as, or more capable than, junior and mid-level accounting staff for relevant tasks. He also argues that they are approaching or exceeding some senior technical experts. The claim is tied to workflows involving document interpretation, repetitive processing, arithmetic, and evidence review, rather than a statement that every accounting responsibility has already been automated.

Q: Why do large accounting firms struggle with AI transformation?

Large accounting firms can struggle because they may invest in licenses, certifications, and public initiatives without building useful workflows or employing enough software engineers. The discussion cites a large firm's purchase of millions of dollars in Copilot licenses and an AI team with zero software engineers. These examples suggest that buying technology does not by itself produce operational change or effective accounting automation.

Q: How could AI change the role of accounting firms?

AI could shift accounting firms away from labor-intensive intermediation by automating document processing, calculations, interviews, and evidence organization. Customers could receive tax, audit, or advisory outcomes through technology while using human experts for nuanced judgment. Dominic Vitucci describes this as a functional and tectonic change in how outcomes are provided, potentially weakening the traditional position of accountants as required middlemen.

Summary & Key Takeaways

• Corporate accounting work often begins with repetitive execution: employees transfer information between documents and spreadsheets, apply formulas, interview client personnel, and organize supporting records. The R&D tax credit example shows that the arithmetic can be simple, while the difficult and valuable part is proving that qualifying research work actually occurred.

• Onshore uses AI models to automate work currently performed by accounting firms, including tax services and, potentially, audit and advisory services. Its approach targets repetitive and superfluous tasks while preserving access to expertise for nuanced judgments, with the goal of delivering outcomes without relying as heavily on traditional intermediaries and labor-intensive workflows.

• Dominic Vitucci argues that frontier AI models can already rival or outperform some junior and mid-level accounting staff, while approaching certain senior technical capabilities. He also contends that established firms struggle to transform because technology investments, licenses, certificates, and public announcements do not necessarily create functioning products or genuine software engineering capacity.