“There will be a lot of trauma and disruption along the way,” Elon Musk told Joe Rogan, framing artificial intelligence as a “supersonic tsunami” moving through the digital economy faster than earlier waves of automation.
The metaphor is impactful since the boundary is uncharacteristically distinct. Musk distinguishes between work that controls information and work that controls matter. Office work exists within a program that can already be read, recoded, and run by AI; a lot of muscular work still needs quick robotics, dependable senses and that kind of improvisation that is hard to automate. A disproportionate shock is the outcome: those jobs with the most text, number, ticket, form, code, and slide deck inputs and outputs get disproportionately impacted.
The magnitude of the possible earthquake has been measured in a manner that renders the warning less of a science fiction and more of a risk accounting. According to one of the most frequently quoted measurement estimates, over 30 percent of the workers might experience at least one in every five jobs being interrupted by generative AI, and office workers and administrators are some of the most vulnerable. That is significant as task disruption does not necessitate the disappearance of a whole job, but simply sufficient routinised bits to be excised to leave the other job smaller, cheaper, or restructured into fewer jobs. It is also the reason why early change usually appears as a form of “simplification” rather than a clean replacement narrative: smaller number of junior employees, less support levels, greater output per manager and increasingly high demands on a single employee to perform the drafting, summarising and first pass analysis that used to be delivered by a small group of junior staff.
The timeline has been brought out clearly by the Anthropic Dario Amodei who cautions that a large portion of repetitive-but-variable professional labor will vanish in the next few years. The technical motivation is not smarter chatbots, but drift towards systems which are “agentic” which take tasks to completion, such as pulling documents, applying rules, writing drafts, finding inconsistencies, and reworking until the final product meets a target. When the work is packaged into a workflow, firms are then able to redesign around it.
The unequal biting is evidenced in professional services. An advisory framework of leadership makes the case that AI will not dismantle the sector, but it will divide it into tiers, based on whether the value derives out of public information processing or will be based on trust and confidential context. In that perspective, AI is gradually sucking the moat around the work of “knowledge experts”: research, compliance, standardised analysis, and preserves the work of a true advisor based on relationships, judgement and discretion. The biggest area falls into a hybrid the human being oversees machines, automating the “what” to allow him to spend time on the “so what.”
An example of this theory to operating model shift is with accounting. In an industry survey in 2025, 21% of tax companies were already using GenAI, and 53% were planning or considering it, which is a move towards becoming a default and not an experiment. According to the same report, most practitioners use general-purpose tools: 52% of GenAI users in tax firms were using open-source instead of profession-specific and niche platforms. That fact counts: as tools are inexpensive, well-known and run by employees, diffusion accelerates, and the staff of governance, checking, and accountability to management transforms into a core managerial issue.
The executives have also defined AI not as an aid but as a replacement of mid-level product especially in the software industry. With that expectation disseminated, the first job positions to be compressed will be those where many people can be trained as they climb the ladder: junior analysts, paralegal support, basic coding, etc., as companies can afford to just buy the speed.
The longer-term vision of Musk is moving in the other way, towards abundance. He has made the case that AI and robotics will render work optional and contribute to a “universal high income,” but only after a phase of dislocation. The engineering dilemma lies that institutions are slower matched to the capability of capability going up. The benefits in productivity are short-run, whereas the labour market adaptation, retraining potential, and plausible redistribution mechanisms are long-run. Between the speeds, the “supersonic tsunami” transforms into less of a wave than a re-engineering of the way white-collar organisations are manned, trained and measured.
