Beyond the Grunt Work: How KPMG’s AI-Powered TaxSIM Is Rewriting the Rules of Professional Development
The old model of learning-by-doing is dead. Here’s how KPMG plans to build judgment in an era where AI handles the repetition.
For decades, the path to becoming a tax expert followed a predictable, if painful, trajectory. You spent your first four years in the trenches—preparing return after return, staring down spreadsheets, learning the nuance of regulations through brute force and repetition. The grunt work wasn’t just a rite of passage; it was the teacher.
But what happens when artificial intelligence takes over those repetitive tasks? How do you build the judgment, the strategic thinking, the “aha” moments that used to come from two thousand hours of manual labor?
KPMG US is betting a new simulation tool called TaxSIM has the answer. And it could reshape how entire industries approach talent development in an AI-crowded world.
The Problem AI Created (That AI Now Has to Solve)
Brad Brown, KPMG’s chief digital officer for tax, put it bluntly: “You’re not going to get as many repetitions of doing that task as you would have in the past. So we needed something to fill that void.”
Let’s unpack that. In the pre-AI era, an early-career tax professional at the Big Four firm spent roughly four years preparing returns for clients—one after another. That repetition wasn’t busywork. It was the forge where professional judgment was hammered into shape. Each return taught something: how a specific deduction interacts with state law, how a merger structure changes tax liability, how a client’s business decisions ripple through compliance.
But as AI tools take over more of the “pencil pushing” (data entry, compliance checks, basic return preparation), junior staffers are losing those reps. They’re being asked to move from task-doer to strategic advisor faster—but with fewer foundational experiences to draw from.
That’s the tension hanging over white-collar work across industries. When machines handle the rote, how do humans build the instinct?
TaxSIM: The “Gran Turismo” Approach to Professional Learning
KPMG’s solution looks less like a training module and more like a video game. The firm is testing TaxSIM, an AI-powered simulation tool developed in partnership with Centaurian AI, founded by Kes Sampanthar.
Here’s how it works: Instead of waiting months or years to encounter different client scenarios (a small business sale, an international tax filing, a restructuring), tax workers cycle through high-volume, high-speed simulations. The software presents realistic tax situations, lets users make decisions, and then delivers different results based on those choices. It’s repeatable, scalable, and—crucially—immediate.
“That just gives us incredible acceleration,” Brown said.
Sampanthar draws a direct parallel to racing simulations like Gran Turismo. “It’s like the top athlete who gets better and better if they can get the right feedback,” he explained. In a racing sim, you don’t need to total a real car to learn the limits of a hairpin turn. In TaxSIM, you don’t need to bungle a real client’s return to learn the consequences of a wrong classification.
The tool is set to roll out to all 10,000 KPMG tax staffers later this year.
Why This Matters Beyond Tax
At first glance, this is a niche story about how one consulting firm trains its people. But zoom out, and it’s a playbook for the future of professional development in any knowledge-intensive industry.
Think about:
- Law firms where AI handles document review. How do junior associates learn case strategy?
- Accounting firms where AI automates bookkeeping. How do new CPAs develop audit judgment?
- Consulting where AI drafts slide decks. How do analysts learn to structure a narrative?
- Software engineering where AI writes boilerplate code. How do junior devs learn architecture?
Every industry that depends on cumulative repetition to build expertise is about to face the same void KPMG identified. And the answer isn’t to fight AI or slow its adoption. It’s to build deliberate practice environments that compress years of experience into weeks.
The Mechanics of “Incredible Acceleration”
Let’s get tactical. How does TaxSIM actually create faster skill development?
1. Volume Without Risk
In the real world, making a mistake on a client’s return has consequences—financial, legal, reputational. So junior staffers are often shielded from complex decisions. They handle the easy stuff (or now, the stuff AI handles). TaxSIM removes that friction. A user can make 50 decisions in one hour, see the outcomes, and adjust. That’s 50 mini-learning cycles instead of one.
2. Decision Trees That Mimic Reality
The tool doesn’t just ask multiple-choice questions. It presents consequence-based scenarios. Choose a certain tax strategy for a client? Here’s how the audit risk changes. Miss a filing deadline? Here’s the penalty. Structure a deal differently? Here’s the cash flow impact. That cause-and-effect feedback loop is what builds genuine judgment.
3. Immediate, Specific Feedback
The old model had a major flaw: feedback was slow. You’d prepare a return, hand it off to a manager, wait days or weeks for review, and then receive corrections. By then, the context was cold. TaxSIM provides instant feedback—what you did right, what you missed, what a better approach would have been. That’s the same principle behind the best training in sports, music, and military simulation.
4. Repetition Without Boredom
The “grunt work” criticism was always that repetition was valuable but dull. TaxSIM gamifies the process by varying scenarios, increasing complexity, and letting users compete against their own performance. High-volume, high-speed, high-engagement.
What This Means for Revenue Teams and Go-to-Market Professionals
You might be reading this thinking, “I’m not in tax. How does any of this apply to my SaaS sales or marketing role?”
More than you’d think.
The same structural problem exists in go-to-market roles. SDRs used to learn objection handling by making 500 cold calls. Marketers used to learn campaign strategy by running 50 failed A/B tests. Sales leaders used to learn deal negotiation by blowing a dozen quota-carrying quarters.
AI is now handling the volume piece. Outreach sequences are automated. Ad creative is generated by models. CRM data is enriched by algorithms. The “reps” that once built expertise are vanishing.
So how do you build a SalesSIM or a MarketingSIM?
- Simulate real objections in a no-risk environment, with an AI playing the voice of a skeptical CFO.
- Run high-volume deal scenarios where the outcome changes based on pricing, timing, or competitor action.
- Test campaign strategies with synthetic audiences before spending real ad dollars.
The principle is identical: compress learning cycles so that “experience” is earned through deliberate simulation, not accidental exposure.
The Bigger Picture: Judgment as a Premium Skill
KPMG’s move signals something deeper about the future of professional services. As AI commoditizes the “what” (data processing, compliance checks, return preparation), the value shifts entirely to the “why” and “what if” (strategic advice, regulatory interplay, business context).
Brown noted that as staffers gained experience in the old model, they became more like advisors—offering guidance on the interplay between business decisions, regulations, and tax consequences. That’s the judgment that can’t be automated. But it also can’t be taught through a lecture. It has to be practiced.
TaxSIM is an attempt to build that practice muscle faster than the four-year grind allowed.
The Challenges Ahead
Of course, simulation isn’t a silver bullet. There are real questions any organization adopting this approach will need to face:
- Fidelity: How close can a simulation get to the messiness of real client work? Sampanthar’s comparison to Gran Turismo is apt—but even a top-tier racing sim can’t replicate the feel of G-forces or the fear of a crash.
- Feedback loops: The tool is only as good as the scenarios and evaluation logic built into it. Garbage in, garbage out.
- Human mentorship: Simulation can build pattern recognition, but can it build the trust, empathy, and nuance that come from real human relationships with clients? Probably not entirely.
- Adoption: Will senior partners embrace this as a replacement for “paying your dues,” or will they distrust it as a shortcut?
KPMG seems to be betting that the acceleration outweighs the risks. And given the competitive pressure to develop talent faster, inertia is riskier than experimentation.
What to Watch For
If you’re following this trend, here are three signals to track:
- How do junior staffers respond? If the first cohort of TaxSIM-trained professionals outperforms traditionally trained peers in advisory work, the model will spread fast.
- Does the simulation expand to other domains? KPMG could build similar tools for audit, advisory, or consulting practices. Watch for announcements about new “SIM” platforms.
- Do competitors follow? Deloitte, PwC, and EY face the same talent development challenge. If TaxSIM delivers results, expect a wave of simulation-based training across the Big Four.
The Takeaway
The AI era doesn’t just change what work gets done. It changes how people learn the craft. The old model of learning through grunt work—years of low-level repetition—is fading. But the judgment that repetition once produced is more valuable than ever.
KPMG’s TaxSIM is an early, bold attempt to bridge that gap. It’s a recognition that if AI takes the drudgery, humans need a new way to build mastery.
For any leader building a team in 2025 and beyond, the lesson is clear: Experience isn’t measured in hours worked. It’s measured in decisions made, feedback received, and patterns recognized.
Simulation might just be the most efficient way to manufacture that experience at scale.
The tool hits KPMG’s 10,000 tax staffers later this year. Expect the ripple effects to hit your industry soon.