AI is not just reshaping who gets hired first, it is changing which experience matters most.
A new Bloomberg report says the leverage is shifting in the AI labor market, and not in the way a lot of startup culture still assumes. More than 40% of CEOs plan to cut junior roles over the next one to two years and move toward mid-level or senior employees, while only 17% want junior roles to make up a bigger share of the workforce, according to a global Oliver Wyman survey cited by Bloomberg. That is almost a mirror image of where sentiment sat a year ago, and it matters because it suggests the winners in an AI-heavy company may be the people who already know how messy the work is.
That is a sharp contrast with the prevailing startup story. The usual narrative says AI tools, especially coding agents, lower the barrier for young developers, making them faster, more productive, and more attractive to founders chasing speed. There is truth in that, but it is only half the picture. The Bloomberg data points to a different reality inside companies that actually have to ship, sell, and support complex products, where judgment, process memory, and the ability to translate business needs into working systems may be becoming more valuable than raw coding output alone.
Bloomberg quoted Oliver Wyman's John Romeo saying the junior level is finding it harder to enter the workforce, while mid- and senior-level employees are what CEOs now want to drive productivity. That tracks with the way AI agents are being used in practice. They can handle chunks of code, review routine tasks, and accelerate workflows, but they still struggle with the parts of work that depend on history, context, and the accumulated knowledge that lives in someone's head after years in an enterprise environment.
That is why the older-worker angle is more interesting than it first sounds. This is not really a story about age for its own sake. It is a story about experience becoming a competitive asset again because AI is compressing the value of baseline execution. If AI can draft the first version of the code, then the scarce skill becomes knowing what should be built, where it should fit, how it breaks in the real world, and which edge cases matter to customers who are not forgiving.
For B2B startups, that is especially important. Enterprise software is rarely about one clever feature. It is about workflows, exceptions, approvals, compliance, and the people who have lived through enough implementation disasters to spot the hidden traps before launch. A founder building an AI product for finance, logistics, healthcare, or internal operations may still want scrappy engineers, but the company can stall if nobody on the team understands how those workflows actually run on the ground.
Hiring should change too
The practical takeaway is that early-stage hiring needs a reset. Too many startups still screen for speed, stamina, and the ability to ship quickly, which makes sense until the product has to survive contact with a real customer organization. The better filter may be whether a candidate has already operated inside the kind of workflow the startup is trying to automate. Someone who has lived through procurement, implementation, customer success, or internal ops can often spot the frictions that a young engineer would only discover after months of trial and error.
That does not mean founders should stop hiring early-career talent. It means the mix matters more than the mythology. Young engineers bring energy, speed, and a natural fluency with AI tools. Veteran operators bring the institutional memory that keeps the product honest. Put the two together and a startup is more likely to build something that survives beyond demo day. Put only the first group on the team, and you risk creating a product that is technically elegant but operationally thin.
Founders should also redesign the funnel itself. Instead of asking only whether a candidate can code or prompt well, they should test for workflow understanding, customer empathy, and the ability to explain how a process really works when the spreadsheet logic fails. That can mean broader interview panels, more scenario-based exercises, and more weight given to people who have run the exact messy systems the company wants to automate. In AI, the hard part is often not generating output. It is knowing which output can be trusted.
Bloomberg's report is a reminder that AI does not erase the value of experience. In some companies, it may increase it. The startups that understand that early will hire differently, retain more institutional knowledge, and build products that feel less like clever prototypes and more like systems that enterprises can actually depend on.
Also read: Claude's radio station refusal says more about AI reliability than laziness • Mistral's Arthur Mensch warns Europe has two years to stop AI dependence • Read-only finance is the cautious path while agentic money moves faster
