Entry-level jobs are often the first thing to be cut once AI can do the routine parts of the work. Amy Edmondson, the Harvard Business School researcher best known for her work on psychological safety and teaming, argues that this trade is shortsighted in ways most organizations have not thought through.
The jobs organizations cut first
When budgets tighten or AI tools mature, entry-level roles are usually the first line of items on the chopping block. The reasoning seems sound on the surface: junior work is often repetitive, AI can do a version of it faster, and the savings show up immediately on a spreadsheet. Amy Edmondson, Novartis Professor of Leadership and Management at Harvard Business School, has been making the opposite case, as she thinks that removing these roles trades a short-term efficiency gain for a much larger long-term cost.
What entry-level jobs are really for
Edmondson’s argument is that junior roles do far more than produce output. They are where future leaders learn the business from the ground up, where fresh perspectives feed into innovation before people become fully socialized into how things are done, and where organizational culture gets passed on in ways that are difficult to formalize into a manual. Cut that layer out and a company is not just saving on entry-level salaries, it is quietly cutting off the pipeline that produces its next generation of managers.
The fix isn’t no AI, it’s better delegation
The recommendation is not to protect entry-level jobs by keeping AI out of them. It is to redesign the roles deliberately: hand the routine, repetitive pieces of the job to AI, and redirect the human time that frees up toward judgment, creativity, and the kind of collaboration junior employees can only learn by doing. Work, in Edmondson’s framing, should stay a site of growth, resilience, and shared human achievement, not just a set of tasks waiting to be automated away.
Why the AI conversation itself needs to change
Edmondson has also argued, that the way most organizations talk about AI is part of the problem. She draws a sharp line between debate, which produces a winner and a loser, and dialogue, which produces shared understanding. A debate mindset is what let engineers’ safety concerns get overruled ahead of the NASA’s Challenger launch in 1986: one side won, and the conversation ended before the real risk was understood. She sees a similar pattern in how companies handle AI today, swinging from enthusiastic adoption to abrupt restriction and back again, reacting to headlines instead of thinking the decision through.
Her alternative is a small set of questions leaders should ask together before any AI rollout, including an entry-level hiring freeze: what problem we are actually solving, which decisions should stay human, and what does success look like. Asked in dialogue rather than argued in debate, those questions tend to produce very different answers, including, often, a decision to keep junior roles and redesign them rather than eliminate them.
Where this conversation is already happening
That question, which decisions should stay human, is at the heart of “AI, Analytics & Human Judgment in HR Decisions,” one of the Table-Top discussions at Horizon Summit 2026, held November 10 to 11 at the Hilton Amsterdam Schiphol. Edmondson’s argument around entry-level work raises a broader HR question: when AI can take over parts of a role, should organizations automate the role itself, or use the technology to redesign it around distinctly human capabilities such as judgment, creativity, learning and collaboration? At Horizon Summit 2026, senior HR leaders will use the Table-Top Peer Discussion format to explore how AI and analytics should shape HR decisions without allowing efficiency metrics to overshadow the longer-term human and organizational consequences.
