In a recent finance training session, an FP&A analyst took the prompting method we had covered twenty minutes earlier and applied it to a monthly variance report spanning four business units. The report normally took one and a half to two hours. It was finished in about two minutes, live, in front of the room. Every leader watching did the same math: if experienced people can work like that, why hire the junior analyst in next year’s budget?
That question sits under most of the early AI jobs story, which has been dominated by fear: fewer junior openings, thinner entry-level ladders and senior employees using AI to absorb work that used to train the next generation.
There is evidence behind that concern. Researchers at the Stanford Digital Economy Lab found that employment among workers aged 22 to 25 had fallen in occupations with high exposure to generative AI. The effect was concentrated in jobs where AI could automate tasks. Employment held up better where companies used AI to support employees.
The companies furthest ahead on AI appear to be doing the opposite, according to recent research from Revelio Labs and Ramp reported by the Financial Times. Among nearly 22,000 U.S. companies, the heaviest AI adopters grew white-collar headcount by 10.2% over two years, with entry-level roles up 12%.
The study does not prove that AI caused the hiring. AI adopters were already larger, more engineering-focused and more likely to be growing, but the findings still matter. Companies making serious investments in AI were expanding their workforces, including their junior ranks. Companies making serious investments in AI were expanding their workforces, including their junior ranks. The difference is team design. AI productivity gains have a shelf life: within a year or two the new pace becomes the baseline, and companies that cut their junior layer to capture those gains will have no one ready to step up.
In my previous CIO article, I argued that successful AI adoption requires companies to change how they operate. Giving employees new tools while leaving existing workflows and decision-making structures in place produces scattered experiments and limited gains. Companies need to give the people closest to the work more authority to test AI, improve processes and teach others what works.
Team design is the next part of that transformation. Once AI changes how work gets done, companies also need to change how teams are built and how employees develop. That includes the junior roles where people gain the experience required to become future managers and specialists.
Cutting junior roles creates a capacity problem
The mistake is easy to make. A company trains its existing staff on AI. Suddenly, experienced employees are 30% to 60% more productive. Junior hires start to look like overhead. For a while, that logic feels rational.
The problem is that this kind of approach has an expiry date. A year or two from now, those productivity gains will be absorbed into the business. The new pace will become normal. Clients will expect more. Internal teams will take on more. Senior employees will be stretched again; except this time there won’t be a trained junior layer underneath them.
The company then has limited choices. It can overwork its experienced staff, compete for expensive external hires or promote people before they are ready. Each option carries a cost.
This is a pipeline problem as much as a hiring problem. Companies cannot produce experienced employees on demand. Judgment develops through repeated exposure to real decisions, feedback and consequences.
Research on AI and customer service offers a clue about how companies can approach this. A study of more than 5,000 support agents found that access to a generative AI assistant increased productivity by 15% on average. Less experienced workers benefited the most. The system helped them apply some of the practices used by higher-performing colleagues.
That finding should interest any leader thinking about early-career development. AI can help junior employees reach competence faster when the tool contains useful organizational knowledge and employees receive guidance on how to use it.
Building teams like multi-agent systems
When we build multi-agent AI systems, we usually build them in a hub-and-spoke model. There’s an orchestrator agent in the middle, then smaller agents handling specific tasks around it. In a marketing function, for example, one agent might handle social media, another might handle SEO, another might handle analytics and another might handle reporting. The user works through the orchestrator, which coordinates the system.
I think this model offers a useful template for team design.
The senior employee acts as the hub, with junior employees and AI agents working as the spokes around them. Instead of simply delegating tasks, that senior person shapes how the work gets done: teaching the AI systems and juniors what good looks like, and helping junior employees build judgment by working inside the same system. The junior employee learns from the AI by seeing how it approaches research, drafting, analysis and problem-solving; the AI improves because the junior is constantly correcting, refining, rejecting and escalating its outputs.
That’s where the training model changes. Historically, juniors learned through repetition, writing the first draft, building the spreadsheet, checking the details, sending the recap and slowly absorbing the standards of the job. AI will take over parts of that repetitive work, but it won’t remove the need for training. It changes where the learning happens: juniors are no longer just learning by doing the task from scratch, they’re learning by directing the task, judging the output and feeding better decisions back into the system.
A junior employee now needs to learn how to question an AI output, spot weak reasoning, understand what a client or stakeholder actually needs, and know when to escalate. This kind of judgment is built like a pilot’s instincts: not by watching someone fly, but by spending time in the cockpit as experienced pilots navigate real decisions.
What companies should hire for now
This changes what hiring managers should look for in junior talent.
Curiosity has always been a crucial quality to look for, but it matters even more now. Juniors need to be willing to learn the tools and understand how AI changes the work. But the stronger signals are emotional intelligence, good interpersonal instincts, discernment, judgment and the ability to explain why they made a decision.
The World Economic Forum’s Future of Jobs Report 2025 supports this broader view of skills. Employers expect AI literacy and technological capability to grow in importance. They also continue to value analytical thinking, curiosity, leadership and social skills. The report found that 77% of surveyed employers planned to upskill workers in response to AI disruption.
That’s also why senior employees need to work differently. Engineers need to spend more time with business teams, so they understand how the company actually operates. Business teams need to spend more time with technical teams so they understand what the tools can and can’t do.
If companies want to keep moving faster, they still need to hire junior talent. But the apprenticeship has changed. Instead of learning by taking repetitive work off someone else’s plate, juniors now join AI-enabled teams where they can see how experienced colleagues think, learn from the systems those colleagues build, and develop judgment through decisions they’re gradually trusted to make. We’re all managers now, some will manage humans, some will manage machines and a select few will build the skills to manage both.