In October 2024, Donald King stood up in front of thousands of colleagues at PwC and presented a fleet of AI agents that automated manual tasks. He had just won first place in a firm-wide OpenAI hackathon with more than 3,000 contestants. Hours later, two, by his account, he was on a call being told his role was eliminated. He recorded it. The video went viral on TikTok (over 2 million views). He was 25 or 26, depending on the telling, and he had been working 60-to-80-hour weeks in the firm's AI factory, building tools for Fortune 500 clients.

The sequence is almost too neat for the moral. King had pitched himself into the AI team after PwC announced a $1 billion investment. He coded, managed onshore and offshore teams, ran internal knowledge-sharing sessions that drew hundreds, and described the work as his dream job. Then the same category of systems he helped create, agents that could mimic people on Microsoft Teams, automate workflows, and take over repetitive knowledge work, became part of the broader efficiency story at client companies. King later told Fortune he saw the connection as "100%": consulting had always been a "hatchet-man" role, but this felt different. "It was actually like I was just feeding myself into the AI meat grinder," he said in one account. He did not claim his own specific job was automated by the hackathon entry; the firm pointed to overhiring and shrinking consulting demand. The timing still landed as humiliation.

This is the classic "digging one's own grave" pattern with advancing AI, sharpened by agentic systems. Knowledge workers, especially those close to the technology, are asked to make the machine better at the tasks that previously justified headcount. In professional services the feedback loop is short: the consultant's deliverable is often a recommendation or a tool that lets the client do more with fewer people. When the tool gets good enough, the client's next budget conversation includes fewer external consultants and fewer internal roles. Entry-level and routine white-collar work is the most exposed first: data entry, basic analysis, document processing, customer-support scripting, junior research. Broader data from 2024–2026 show elevated AI exposure correlating with slower hiring or outright declines in some of those categories, even while overall employment and senior roles have been more resilient. Companies cite AI alongside cost-cutting; the technology is rarely the sole cause, but it expands the menu of what "lean" can look like.

King himself felt the moral friction while building. A late-night call among the team produced the uneasy question of what, exactly, they were constructing when they treated agents like humans. Agents are still limited; King compared them at one point to the reasoning of a five-year-old that nonetheless has access to the world's information, but they are already good enough at the menial loops that used to fill junior calendars. "If your job is doing those menial types of things, if you're just emailing a spreadsheet back and forth, you can kiss your job goodbye," he said. That is not universal prophecy; many jobs contain irreducible judgment, relationship, or physical elements. It is a description of the parts of knowledge work that are most easily specified and verified.

Historical parallels exist, weavers and the power loom, clerks and the spreadsheet, typesetters and desktop publishing, but the speed and the generality of current models compress the transition. Previous waves often created new demand in adjacent or higher-value activities while the old ones shrank. AI agents do the same in principle: oversight, exception handling, new product design, and entirely new categories of work. The friction appears in the gap between those new opportunities and the people whose current tasks are being compressed. Young workers and those in highly exposed roles have felt hiring freezes or reductions first in several datasets. Firms that over-hired in the prior boom, then layered AI on top, have had more room (and incentive) to cut.

King's own next chapter undercuts pure fatalism. After a period of disorientation he founded a marketing agency (AMDK) aimed at older executives who need help with personal branding on platforms like TikTok. The viral layoff video became distribution. He has described the layoff, in retrospect, as the best thing that happened to him even while it felt like the worst in the moment, more ownership, less "cog in the machine." He still works with AI, including for clients, but under conditions he controls. That is one adaptive path: treat the disruption as forced optionality rather than pure loss. Not everyone has the same network, savings, or temperament for entrepreneurship, and not every displaced role maps cleanly onto a new one.

The larger pattern remains. Organisations that reward the people who most effectively automate work will keep doing so. The people who succeed at that automation sometimes discover that the success reduces the organisational need for their prior form of labor. That is not a conspiracy; it is what efficiency looks like when the technology improves. Individuals close to the tools can see the outlines earlier and can sometimes reposition, toward the parts of the work that are harder to specify, toward ownership of outcomes rather than execution of steps, or toward the new demand the tools create. Waiting for the old job description to stay stable is the riskier bet. The grave-digging feeling is real when the shovel is in your hands and the hole is the shape of your old role. The more interesting question is what you do with the dirt once it is out.