The form of the deliverable can, in turn, change how the work is organized.
This is also how I understand the “one-person company”: if a company is designed from the beginning to have agents do a great deal of execution, it may be able to bypass many traditional divisions of labor and layers of coordination between people.
Information gathering, organization, analysis, production, and checking that once had to be assigned to different people can be carried out under a single goal, with agents calling tools, advancing the work, and responding to feedback.
In such an organization, the demand for human coordination devoted to assigning tasks, chasing progress, and forwarding information may fall substantially. Real-world relationships, authorization, and responsibility still need to be handled, but not every deliverable requires assembling a human team first.
A person might not even be good at organizing people in the traditional sense. If they can see the goal clearly, understand the key trade-offs, know what form the result could take, express their intent clearly, and judge the quality of delivery, they may be able to drive work that previously required several people to collaborate.
In “The Gentle Singularity,” Sam Altman also made this prediction about individual productive capacity:
“the ability for one person to get much more done in 2030 than they could in 2020 will be a striking change”
His point is that the amount one person can accomplish in 2030, compared with 2020, will change dramatically. Original: Sam Altman’s personal blog ↗
That passage is about expanding individual productive capacity. Following that direction, I believe businesses will have an opportunity to redesign how many people they need, what kinds of people they need, and which work can be handed directly to agents.
It also means that general entry-level office skills, along with many intermediate professional skills that depend on proficient execution, may quickly lose the scarcity that once made them valuable.
Writing emails, making spreadsheets, organizing material, and applying templates used to be enough to sustain a position. Knowing a set of rules and producing a certain kind of professional document according to established practice used to command a higher salary.
As these execution tasks become easier for agents to perform, organizations will recalculate: how many people are actually needed to deliver the same result?
Experience may still be valuable, but that value must show up in judgment. Can you notice that the task’s fundamental purpose has been misunderstood? Can you identify a result that looks complete but cannot actually be used? Can you decide which conditions must be satisfied and which allow compromise?
Simply repeating “I’ve been doing this for years” does little to answer those questions.
So I want to give a clear warning to office workers whose main work is processing language in this broad sense:
You may already be running out of time to change the structure of your capabilities.
Others have expressed this urgency even more explicitly than I have.
In his January 2026 essay “The Adolescence of Technology,” Anthropic co-founder Dario Amodei reiterated an earlier prediction:
“AI could displace half of all entry-level white collar jobs in the next 1–5 years”
He was warning that AI could replace half of entry-level white-collar positions within one to five years. Original: Dario Amodei, “The Adolescence of Technology” ↗
He also discussed a question office workers should take seriously: although professions such as finance, consulting, and law require different specialist knowledge, their entry-level work relies on similar general cognitive abilities. If all these fields are disrupted at once, moving into a neighboring profession may not be enough to escape the change.
He acknowledged that businesses take time to adopt new technology. But he immediately added:
“But diffusion effects merely buy us time.”
The delay in the spread of technology, in other words, only gives us more time. The same original essay ↗
That sentence is well suited to anyone who assumes they can keep working as before simply because their own company has not changed yet.
My judgment is that, over the next year or two, people who still cannot move from proficient execution to defining goals, judging trade-offs, and accepting deliverables will begin to feel real pressure. My timeline is more urgent than Amodei’s range. It is my own judgment.
Change will not happen simultaneously across every industry and company. But you may first notice fewer new positions, work that once needed several people being assigned to one, and someone else delivering in far less time a result that takes you days.
Being displaced may not begin with a dismissal notice. It may begin with your work becoming worth less and less.
The market will not stop comparing delivery speed, quality, and cost because you believe your business is special. Nor will a company forever pay for steps that can already be performed more efficiently simply because you work hard every day.
You can test yourself against a real task right now: Can I explain its fundamental purpose? Can I judge which requirements must be maintained and where compromise is acceptable? Can I imagine a more suitable form of delivery? Can I use actual results to specify what should improve next?
If you cannot answer those questions, you should address that as soon as possible.
Because the next time you confidently declare, “My work is too complicated. AI can’t do it,” someone else may already have handed the goal to an agent, let it execute, verify, and adjust, and received the very result you are still producing by hand.