Thomson Reuters integrates AI into its professional tax and auditing software

Thomson Reuters Corp. today launches a revamped version of its tax, audit and accounting platform CoCounsel that not only answers questions but also schedules tasks, performs calculations and creates documents for professional review.

The advanced platform combines artificial intelligence agents with customer information, tax research and code execution to handle complex tasks such as creating repayment schedules, preparing spreadsheets and creating memos. The company said the new platform is available to all US customers.

The paper targets a challenge in applying generative AI to professional services: a mathematically correct answer can still be wrong if the system uses the wrong tax rule, ignores the customer’s circumstances, or fails to explain its reasoning.

“They expect near-perfection from every tool they use,” Chief Product Officer David Wong said at a briefing on accountants.

Unlike its predecessor, which was focused on individual tasks, the rebuilt system can break complex requests into steps and select the tools and information needed to complete them, Wong said. Experts still need to check and verify the results. The system provides quotes, assumptions and calculations to help them evaluate their conclusions before approving work.

The research database is Checkpoint, Thomson Reuters’ collection of more than 12 million primary source and analytical documents curated by hundreds of experts. The platform supports multiple AI models.

In a demonstration, David Ragones, product lead for direct tax and audit, showed how CoCounsel can use a commercial loan agreement and a company’s spreadsheet template to create a repayment plan for a $5 million loan. The example included changing interest rates, a lump sum payment, and a balloon payment at maturity.

The resulting workbook included live formulas, supporting assumptions, and references. Users can select individual cells and ask CoCounsel to explain the calculations and their sources.

The demo also showed how the system recalculated spreadsheet results using Python routines to check their accuracy. “A lot of times you show that the work is actually getting done,” said Lucas Adams, vice president of engineering at Thomson Reuters.

Security controls restrict agents to files permitted by the user’s access and workspace. Adams said those limitations also extend to the underlying infrastructure. Internet access is limited to approved government websites adjacent to Checkpoint.

A shared client context layer preserves uploaded documents and completed work across conversations, reducing the need to repeatedly provide background information. Integrations include UltraTax, Engagement Manager, and Guided Assurance from Thomson Reuters.

Model Context Protocol support enables connections to other systems and access to cited tax and accounting responses via third-party AI tools, including Anthropic PBC’s Claude and Microsoft Corp’s Copilot.

Wong said staffing shortages are helping to drive demand as companies try to manage workloads without expanding seasonal hiring. Thomson Reuters also sees an opportunity to release specialists for consulting work.

Executives acknowledged that the practical test is whether reviewing AI output requires less effort than creating it manually. “If reviewing the work takes as much time as the work itself, AI isn’t really helpful,” Adams said.

Photo: Unsplash