“There will be a lot of trauma and disruption along the way,” Elon Musk told Joe Rogan–pulled out a metaphor that would be found in an engineering laboratory more than a policy meeting in a workplace: a “supersonic tsunami.”
The image is practical as it suggests something definite regarding speed and detection. A tsunami is not very visible on the surface of the sea and destructive when it hits; “supersonic” brings with it the implication that once organizations notice the wave, it is already ahead. The wave does not strike the entire economy across in the framing of Musk. It focuses first in regions where work is completely digital, and in places where it can be represented as text, code, spreadsheets, slides, and tickets, all of which can be sent through a model and assessed by the downstream systems.
That dichotomy–between data manipulation and atom manipulation–is now the most convenient way of expressing near term exposure. Generative AI is mostly “disembodied,” i.e. it is good at information work but poor at those jobs where the physical presence, dexterity and sense of the real world are critical. This establishes a predictable trend in even areas where the adoption of AI has a significant role: a slow shift in the core of the hands-on jobs and a rapid shift in the administrative and documentation layers. The outcome is neither a single automation tale but a redistribution of labor, in which the aspects of a job that were previously the point of entry into the job, drafting, summarizing, triaging, formatting, etc., are the initial to be taken over by the software.
The process of measuring that absorption has also come into maturity. The framework presented by Indeeds Hiring Lab categorizes the skills based on their susceptibility to change due to GenAI usage and does not ask whether the job is on the verge of disappearing. In that strategy, 26% of the positions posted on Indeed may be highly transformed, with the majority being somewhere in the middle of “moderate” change where adoption decisions are as significant as the model capability. It further concludes that 46 percent of skills in an average job advertisement are in the so-called hybrid “transformation” AI has the capability of performing routine tasks, but humans are still responsible of exceptions, ambiguity, and risk. It is a less noisy, more functional image than the more conventional “jobs gained vs. jobs lost” debate, and it is what a lot of organizations are constructing, workflows where AI drafts and offers suggestions and routes and people approve, make corrections and sign off.
Nonetheless, executives continue to outline staffing implications in crude language. Mark Zuckerberg has justified that AI will start to do the jobs of mid level engineers saying, “My view on this is like [in the] future people are just going to be so much more creative and are going to be freed up to do kinda crazy things.” Meanwhile, organizations have started to use AI as an excuse to make their teams flatter and smaller. Marc Benioff, Salesforce CEO, explained that “I’ve reduced it from 9,000 heads to about 5,000, because I need less heads.”
The exposure of rhetoric to tasks is no longer theoretical, which is one of the reasons why the rhetoric lands. Occupational analyses associated with openAI indicate that over 30 percent of the employees may face the risk of at least half of their jobs being disrupted and office and administrative support jobs are some of the most at risk. Brookings introduces a second, structural point, that diffusion is abnormally easy since the “rails” are already there and are in the form of browsers, enterprise software, and cloud tools, such that gains in capability can spread without the hardware footprint being established.
The arc by Musk is longer, and it attempts to eliminate the tension by referring to robotics. Assuming that screen work compresses due to AI, physical work can equally be compressed in this fashion; and with some optimistic assumptions, productivity can become ubiquitous material plenty. The Optimus is within that promise of Tesla and Musk has indicated that Tesla would only sell its humanoid Optimus robot once high reliability and safety is proven. It is not the date that is critical but rather the gating criteria, reliability, safety, and functionality, the same limitations that ensure that “AI replacing physical labor” remains slower than “AI replacing screen labor.”
At this point, the tsunami metaphor is primarily a reconstruction of white-collar processes: there will be less individuals writing first drafts and more first-time auditors, integrators, and result-takers. The disruption does not occur evenly among occupations, but among tasks, in particular the ones that have traditionally been invisible until the point at which a model has been trained to perform them cheaply and in bulk.
