These days, PwC’s lower floors are filled with a certain kind of unease that isn’t evident in press releases or quarterly updates. It appears in conversations in the hallway, in Slack threads that are discreetly screenshotted and shared, and in the way junior employees describe seeing a piece of software finish in nine minutes when it used to take them the majority of a workday. It is now impossible to ignore the discrepancy between what the company claims about artificial intelligence and what it actually pays employees to do, according to a number of current and former employees.
The figures contribute to the explanation. Beginning in 2023, PwC invested about $1 billion in generative AI, growing to become OpenAI’s biggest enterprise client and even reselling ChatGPT Enterprise to its own customers. That is a significant wager, and it appears to be a transformation on paper. However, the underlying billing model hasn’t advanced nearly as quickly. Like the majority of the Big Four, PwC still charges primarily by the hour for audit and advisory work, which creates an odd incentive: a tool that completes the task more quickly doesn’t necessarily result in a lower client bill, but it does make it more difficult to justify a room full of first- and second-year associates logging time on tasks the software now handles.
It’s important to keep in mind how this previously appeared in a different form. Only 154.5 hours had been recorded on an audit for a fee of £355,000, according to the 2018 Financial Reporting Council report on PwC’s audit of BHS. This represents an average charge of more than £2,000 per hour, and the majority of the work was completed by junior employees with less than a year of experience. AI was not the subject of that report. It concerned a system in which there was no way for a client to confirm that the hours, fees, and actual work matched. The tension that exists now seems to be an echo of the same structural issue, but instead of a twenty-five-year-old who is overworked, software is handling the work.
The junior employees themselves are becoming aware of the change, and some of them are expressing it. The company’s own AI tools, which are designed to draft memos, identify anomalies, and summarize filings—tasks that were once the mainstay of entry-level training—have reportedly caused internal frustration. There is a perception that younger workers are being assessed based on the same utilization goals that have guided consulting careers for decades, even though they are being asked to bill time for work that a model can complete in a fraction of it. It’s unclear yet if this annoyance turns into something more structured or simply drives people away.
In this, PwC is not by itself. Leadership at the Big Four has discussed a “whole new level” of AI integration, supercharging tens of thousands of workers, and saving consultants a third of their research time. Similar remarks are made by McKinsey regarding Lilli, an internal tool that is currently utilized by most of its employees. Every company uses the same language: AI is an accelerant, not a replacement. Accelerants, however, continue to alter what is paid for and by whom.
Additionally, there is a more subdued implication that affects both employee morale and client trust. The question of what clients are really paying for becomes more acute rather than less so if software can finish audit tasks that junior staff used to bill hours for. One consulting firm’s head, Paul Griggs, has already publicly stated that senior employees who aren’t “paranoid about being AI-first” are likely to be replaced. This statement is directed at leadership, but when junior employees interpret it as a warning about themselves, it takes on a different meaning.
This does not necessarily indicate that PwC is experiencing a crisis. The firm’s leadership doesn’t appear to be giving up on the pyramid structure that has characterized consulting for a century, and audit revenue hasn’t collapsed. However, there is actual friction, and it originates within the building rather than from a rival company or a regulator. As this develops, it’s difficult to avoid wondering if AI will break the billable hour—the foundation upon which consulting was built.
