KPMG made the boldest model bet of the three: a global alliance with Anthropic that puts Claude inside its KPMG Digital Gateway for all 276,000 of its people. While rivals talk about pilots, KPMG wired a frontier model into the workbench its staff use every day.

The sharper KPMG idea, though, is about the bill, not the build. As companies move from toy AI demos to real usage, the cost of running these models, measured in usage charges it calls tokens, is starting to rival the cost of staff. KPMG is turning that shock into a product: governing how much AI gets used, proving every agent action is logged and accountable, and selling discipline rather than just capability.

The engine, though, is borrowed. KPMG runs on Claude from Anthropic and on the Microsoft stack, so it is a premier tenant rather than an owner, and the promise to govern your AI is also a way to bill for the new complexity AI creates. Selling the cure for a problem you help introduce is a comfortable place for a consultancy to sit.

The read: KPMG is the first of these three to treat the AI bill as the real executive pain and to sell governance as the answer. That is a genuinely sharp read of where clients hurt, but the governance still wraps models KPMG does not control.