On October 28, 2026, OpenAI published the second essay in its "next economy" series on its website. Two young managers who started leading their own teams at twenty-four drew three org charts: AI growing beneath employees, sitting beside employees, and wedged between people. The authors state at the top that these are their personal views, not those of OpenAI or their colleagues. None of the three scenarios will play out exactly as described, but each is already showing early signs.
Three places, three different problems
AI beneath employees means a "basement" of AI labor grows under each person, and everyone becomes a manager. The advantage is a smooth transition: no restructuring is needed, since existing employees first get assistants, then chiefs of staff, and eventually a small team of their own. The cost falls on newcomers: the lowest rung of grunt work is taken over by AI, and young workers lose the apprenticeship path of learning from that work, so judgment, taste, and management skill have to be trained another way.
AI beside employees joins the existing structure as a colleague, reporting directly to a human manager, moving from taking on tasks, to taking on roles, and eventually to taking on a whole function. The more complete its capability, the easier it is to hand over, but a half-ready "colleague" creates many seams at every handoff: each additional AI colleague adds another interface to align, and the friction of adoption may grow before it shrinks.
AI between people is the most counterintuitive: below executives and above employees sits a layer of AI. It reads work records across the company for executives, proactively spots a mismatch between what sales promised and what engineering is building, and makes quietly diligent employees visible. It is the easiest to adopt, because executives need no technical skill and can simply ask. But it is also the most likely to pull power toward the top, turning the information advantage closest to the executive into a new instrument of centralization.
The real bet is not layoffs, but who decides
The essay's judgment is that none of the three scenarios necessarily means less total employment: there may be more companies, each with fewer people. Cheaper prototyping will let "build it and see" win over layers of approval, and employees' room for initiative may grow, or may be tightened from above by an ever-present "AI consultancy." What decides the direction is not the model itself, but whether a company uses AI as scaffolding or as reins: the former reports genuinely major exceptions upward and leaves everyday decisions with employees; the latter uses it to enforce an executive's preferences all the way down. The right way to read this essay is as a map to be verified, not as an org chart that has already arrived.
