EY is the odd one out of the Big Four: it has no deal with a frontier AI lab. Instead of betting on access to the newest model, it is betting on trust, governance and proof. Its pitch is that it ran AI on itself first, calling itself client zero, and only then sells the same transformation to clients.

That means EY.ai, its in-house assistant EYQ rolled out to hundreds of thousands of staff, and agents aimed squarely at the hardest regulated work: tax, audit and assurance, where being right and being able to show your work matters more than being fast. The framing is human-led and AI-powered, with people kept firmly in charge of judgement.

The catch is twofold. First, that human-judgement message conveniently protects a headcount-heavy business; CEO Janet Truncale says EY can double in size with the people it has rather than cut. Second, EY does not own its AI engine: it builds on Microsoft and NVIDIA, so the trust it sells sits on top of technology it rents.

In sum, EY has the most credible we-did-it-to-ourselves story and the deepest push into regulated, audit-grade AI. But putting AI into audit invites the regulators in, and a story built on human oversight is also a story built on still needing all those humans.