KPMG is testing a new tool to train staff as AI takes over more of the grunt work of taxes

At KPMG US, workers might no longer need years of hands-on, repetitive work to get good at tax prep.

The company is testing an AI-powered simulation tool designed to help employees build the skills they once gained through years of pencil pushing.

The goal is to allow junior staffers, in particular, to learn the ins and outs of tax prep as AI begins to take over tasks, said Brad Brown, the company’s chief digital officer for tax.

“You’re not going to get as many repetitions of doing that task as you would have in the past,” he said. “So we needed something to fill that void.”

Before the AI era, Brown said, an early-career tax professional at the company typically spent about four years preparing returns for clients, one after another. The repetition built judgment.

As staffers gained experience, they could become more like advisors, offering guidance on the interplay between business decisions, regulations, and tax consequences, he said.

The software tool, called TaxSIM, is designed to help workers develop skills “much more rapidly” through high-volume, high-speed simulation, without having to prepare as many returns from scratch, Brown said.

The software will be available to the company’s 10,000 tax staffers later in the year.

KPMG’s approach is an attempt to solve a problem that’s hanging over large swaths of white-collar work: What happens when AI takes over the rote tasks that help people develop their skills?

The idea behind the AI training is similar to how the game “Gran Turismo” simulates zipping around a Formula 1 course, said Kes Sampanthar, cofounder and CEO of Centaurian AI, which developed the tool with KPMG.

“It’s like the top athlete who gets better and better if they can get the right feedback,” he said.

‘Incredible acceleration’

Instead of waiting perhaps years to encounter different client situations, tax workers at the Big Four firm will be able to cycle through simulated scenarios. The tool presents different results based on users’ decision-making.

“That just gives us incredible acceleration to create those skills,” Brown said.

He said that while some early-career workers are expressing concern about how they’ll build their skills as AI takes on more work, the need to continually learn “cuts across every rung of the ladder.”

More experienced workers who use the tool can dig into broad concepts, including changing economic conditions, such as how tariffs ripple through organizations and countries, said Sampanthar, who’s also a KPMG alum.

“These are things you normally don’t get feedback on,” he said.

Sampanthar said that, for any user, the system is designed to prompt them to work through their own reasoning before turning to AI for help. Some simulations can’t be done with AI alone.

“Learning happens when things are hard, not when things are easy,” he said.

Helping junior workers move up

Brown said he expects the tool to accelerate the development of workers’ analytical skills, helping some employees move up more quickly.

In some cases, that might entail managing a team of AI agents and less time spent “working these tasks yourself through the midnight hour,” Brown said.

Ultimately, he said, it’s a reminder that workers who can use AI alongside their judgment and analytic skills will be most successful.

That doesn’t mean AI simulations will replace every task, Brown said.

A first-year tax analyst recently told him about wanting to build their first valuation model by hand before learning through simulation.

The person’s thinking, Brown said, was, “It’s OK to do one or two, but I don’t need to do four years of these to get to that level of skill.”

Do you have a story to share about your career? Contact this reporter at tparadis@businessinsider.com.