AI Isn't Replacing Expertise: It's Raising the Bar for New Hires

The short answer

In a July 2026 Harvard Business Review article, "Research: AI Is Changing What Employers Want from New Hires," Jim Doucette and Vishal Gaur report a finding that cuts against the usual entry-level panic: generative AI is raising the bar for new hires, not lowering it. After studying 30 organizations, they conclude that employers want people who can direct a model, not people a model can replace. And each of the three capabilities they say now separates strong hires from weak ones is, underneath, a task-level skill. You cannot demonstrate it, or defend against it, until you can see your work the way these employers now do: task by task.

What the study looked at

Doucette and Gaur studied 30 organizations across three sectors that recruit the largest share of MBA graduates: banking and finance, management consulting, and technology, media, and e-commerce. Their conclusion is that while AI automates a growing pile of lower-level tasks, it makes the surrounding human judgment more valuable, not less. Functional expertise in your own area, they write, "will continue to be essential," but "it will not be sufficient." They group what they learned into three capabilities professionals "will have to possess to be hired and succeed in the AI era," and every one assumes you can break your job into its component tasks.

Capability 1: Take on broader roles

By "reducing the time that professionals must spend on lower-level tasks," AI lets one person cover ground that used to take a team. The authors' example is product management. Products were once built by a "core trio" of product managers, UX designers, and developers. Now basic tasks like ticket tracking, coordination, data collection, and reporting are automated, and AI generating interface designs from prompts is eroding early-career jobs such as building mockups. Executives expect these roles to converge into a single "general technologist" who uses AI to build mockups, write first-pass code, and ship prototypes. With execution handled, product managers are expected to be "more strategic and visionary," to know the product life cycle end to end and "know what can go wrong along the way."

Capability 2: Synthesize knowledge across functions

The second capability is pulling insight across departments to move faster. Doucette and Gaur describe a "premier management consulting company" using AI to speed physical-product development: tools mine internal client data, estimate costs from historical data, "rapidly test concepts on synthetic customer personas," and evaluate procurement strategies. The key point: the most important role for new product designers "is not building technical designs; it is rigorously assessing the quality of inputs, building test scenarios, pressure-testing AI outputs, and enhancing and fine-tuning the design." That requires "a complete picture of the data in their organizations and supply chains," down to whether the synthetic personas "capture the richness and variety of their customers." The model produces; the human is paid to know whether the output can be trusted.

Capability 3: Redesign workflows with embedded AI

The third and highest-leverage capability is restructuring how work gets done around AI agents. The authors report that when agentic tools automate tasks like data collection and reporting, the time to complete them, such as producing versions of an API, falls by 30% to 60%. So the human job changes: instead of gathering information and designing test cases, product managers now "focus on defining product requirements, developing concepts, deciding whether the agentic AI solution makes sense," and judging whether launch risk is acceptable. They name a new task, "watching over AI," because when AI deploys products "small errors can become large production risks," so managers "need to develop guardrails, not just focus on efficiency."

The finding job seekers should notice most

Near the end sits the point that should reframe how you think about your own risk. The executives Doucette and Gaur studied "made it clear that they are investing in training existing mid-level professionals to become proficient in AI rather than replacing them with new AI native hires." That is good news and a challenge at once. Companies are not swapping experienced people for AI-fluent newcomers, but the door for new hires is narrower. As the authors put it, "for MBAs seeking jobs at companies in the three sectors, the bar is higher." The message is not that AI will take your job. It is that AI just raised the price of admission.

Why every one of these is a task-level skill

Look across all three capabilities. You cannot take on a broader role without knowing which of your tasks a model can absorb and which still need you. You cannot synthesize across functions without understanding what each function's work is made of. And you cannot redesign a workflow around an agent, the capability the authors rate highest, without decomposing that workflow task by task and deciding which steps to automate, which to keep, and which to guard. That is the lens we take in every analysis here: jobs do not get automated all at once; individual tasks do. These 30 employers have simply internalized that faster than most job seekers have. If you are thinking longer term, this is also why future-proofing your career against AI starts with a task map, not a mindset.

What employers used to reward vs. what the HBR study says they screen for now

Frequently asked questions

Is AI lowering the bar for new hires?

No. In 2026 Harvard Business Review research, Jim Doucette and Vishal Gaur studied 30 companies and found AI is raising the bar. Employers want people who can direct AI and judge its output, not people a model can replace.

What capabilities do employers want from new hires in the AI era?

The study names three: taking on broader roles, synthesizing knowledge across functions, and redesigning workflows with embedded AI. All three depend on understanding your job at the level of individual tasks.

Map your tasks before your next interview

Doucette and Gaur's research and the task-level view we take here point to the same place: your appeal to an employer, and your exposure to AI, is decided at the level of individual tasks, not job titles. An AI Job Risk Score breaks your role into its real tasks, shows which are most exposed to AI, and hands you the task map this new class of employer expects you to bring. Know which parts of the job a model is already reshaping, and how you would redesign the rest, before your next conversation.