
Behind-The-Scenes of KPMG’s New Tax Tool
Most AI agents struggle to deliver reliable results, but KPMG’s latest development may change that. Their Australian team built an AI system capable of producing comprehensive tax advice in just a day, a process that normally takes two weeks.
Curious how it works and what it means for the future of tax services?
Keep reading 👇
This Week’s Featured Jobs: Find your next career move
Senior Accountant – Auburn, MA | $80k/year
Financial Analyst – Nashua, NH | $80k-$95k/year
Senior Controller – Portsmouth, NH | $150k-$175k/year
Assistant Corporate Controller – Waltham, MA | $150k-$180k/year
Director, Finance Operations – Wilmington, MA | $185k/year
Chief Financial Officer – Lewiston, ME | $225k-$260k/year
Tax Manager – Exeter, NH | $140k-160k/year
Sr. Financial Analyst – Biddeford, ME | $100k-$120k/year
Assistant Controller – Newton, MA | $150k-$180k/year
Controller – Brunswick, ME | $125k/year
Accounting Manager – Andover, MA | $115k-$140k/year
FP&A Manager – Bedford, MA | Up to $140k/year
Senior Accountant – Ashland, MA | $90k/year
Manager of Global Tax – Boston, MA | $150k-$185k/year
From 2 Weeks To 1 Day: KPMG’s AI agent tackles taxes
While most AI agents have fallen short of expectations, KPMG may have found a way to change that, at least when it comes to tax season.
According to The Register, KPMG’s Australian branch has developed a 100-page AI prompt capable of autonomously generating tax advice for clients much faster than humans. The firm built its own AI bots using models from major players such as OpenAI, Microsoft, Google, Anthropic, and Meta, feeding them extensive historical tax data alongside the Australian tax code.
“It is very efficient. It does what our team used to do in about two weeks in a day. It will strip through our documents and the legislation and produce a 25-page document for a client as a first draft.”
-John Munnelly (KPMG Chief Digital Officer)
This progress stands out in a landscape where many AI agents have underperformed. A Carnegie Mellon University study found that even top-performing models completed intended tasks only 30.3% of the time. Earlier this year, AI expert Glenn Hopper noted that AI agents are “not ready for primetime” and largely “vaporware.”
Despite the breakthrough, the technology isn’t quite ready for everyday taxpayers; its complex outputs are still geared toward seasoned tax professionals rather than casual filers. But maybe one day.
