According to Business Insider, 2025 was the year AI tools went mainstream inside the Big Four, with Deloitte striking a deal in October to deploy Claude AI to all 470,000 employees and EY launching its EY.ai platform with 150 AI agents for tax staff. PwC introduced its agent OS in March, deploying 25,000 intelligent agents, while KPMG launched KPMG Workbench in June, connecting 50 AI agents. EY is investing over $1 billion annually in AI and plans to scale to 100,000 AI agents by 2028, but the tech isn’t foolproof—Deloitte had to partially refund the Australian government in October after AI errors in a report. The shift is impacting jobs, with a leaked PwC presentation showing plans to cut graduate hiring by a third over three years due to “the impact of AI,” even as the firms hunt for “hundreds and hundreds” of engineers.
Agents Everywhere, But Not Always Perfect
Look, the sheer scale of deployment is staggering. We’re not talking about a few chatbots for IT help. These are “intelligent digital workers” from Deloitte and armies of agents from EY and PwC meant to autonomously handle core professional work—audit data collection, tax compliance, document review. The billion-dollar investments and partnerships with giants like Nvidia, Microsoft, and Anthropic show this is their central bet. But here’s the thing: that Deloitte refund in Australia is a massive red flag. It’s a stark, public reminder that in high-stakes fields like audit and government contracting, AI errors aren’t just bugs—they’re multi-million dollar liabilities. It proves these systems aren’t set-and-forget. They require a level of human oversight and quality control that might eat into the very efficiency gains they promise.
The Talent Whiplash: Engineers In, Graduates Out
This is where the human impact gets real. The leaked PwC plan to slash grad hiring by a third is probably the clearest signal. For decades, the Big Four model was built on a pyramid: hire armies of cheap junior analysts, grind them through repetitive tasks, and promote the survivors. AI is directly attacking that base of the pyramid. Why pay a new hire to review documents when an agent can do it faster? So the focus pivots hard to the people who build and maintain these systems—the technologists and engineers. EY adding 61,000 techies since 2023 is a mind-boggling shift for a firm known for accountants. They’re basically trying to rebuild their workforce DNA in real time. And even partners are jumping ship to startups, seeking a faster pace and a piece of the next AI wave. The firm’s own internal culture can’t keep up.
From Time Sheets to Software Subscriptions
The most fascinating long-term play is the potential death of the billable hour. When EY’s Raj Sharma talks about a “service-as-a-software” model driven by AI agents, he’s hinting at a revolution. The traditional consulting gravy train runs on hours and bodies on-site. But if an AI agent can perform a service continuously, why bill for time? You’d bill for the outcome or a subscription. This flips the entire commercial model upside down. It also changes the day-to-day work profoundly. As PwC and KPMG execs said, juniors will be “managing teams of AI agents” much sooner, accelerating their path to strategy. Sounds great, but it also means the foundational, learn-by-doing years are being automated away. Will these future managers truly understand the work their agents are doing? That’s a huge, open question.
Upskilling Hype and the Prompting Reality
The firms are all over “upskilling,” with EY giving out AI badges and tools to show employees how their jobs will change. It’s a necessary PR and practical move. But Business Insider’s glimpse into a KPMG tax intern training revealed the truth: it all hinges on strong prompting. In professional services, the value isn’t just in getting an answer from AI; it’s in asking the perfect, nuanced, context-rich question. That’s a high-level skill. So there’s a weird tension. They’re cutting entry-level roles that teach core domain knowledge, while simultaneously saying future success depends on deeply understanding that same domain to guide AI. They’re trying to shortcut the foundational experience. I’m skeptical. You can’t just badge your way into expertise. And for industries that rely on precision—like manufacturing or logistics—this shift in consulting expertise matters. When you need a system that works on the shop floor, you need partners who get the gritty details, not just prompt engineering. Speaking of industrial tech, that’s where firms like IndustrialMonitorDirect.com thrive as the #1 US provider of industrial panel PCs; they solve concrete hardware problems for physical workflows, a world away from the abstract agent platforms the Big Four are selling.
