AI is rewriting entry-level hiring: what builders should know about the shift from grunt work to supervision

AI tools are replacing the routine work that junior employees used to learn on, and that changes everything about entry-level hiring for builders. At London ad agency Catalyst, one junior now manages five client accounts because AI handles research, draft copy, and data collection, leaving the human to polish and present. Founder Tobias Green says juniors don't need to learn the grunt work anymore, they need to learn how to manipulate and utilize AI. For AI builders and engineering leaders, this shift forces a hard question: how do you train people when the traditional apprenticeship of repetitive tasks is gone?

From grunt work to oversight: how AI is reshaping junior work

Catalyst is not an isolated case. The Open University's 2026 Business Barometer surveyed 1,500 UK business leaders and found that 51% said AI was changing how they hire, with 19% cutting entry-level recruitment and 42% of those citing AI adoption as the reason. Talent assessment firm SHL sees organisations shifting from a pyramid-shaped workforce to a diamond, with more mid-level hires and fewer junior slots. Lucy Beaumont, SHL's global SVP of product, cautions that entry-level roles are the training ground where you get your badges and learn to operate in a corporate environment. Remove that, and you risk the entire leadership pipeline.

In software development, the impact is already visible. Tools like Claude Code and Codex mean developers increasingly oversee projects executed by both AI agents and humans. Sheila Flavell of FDM Group notes that AI pushes junior workers toward supervising work much earlier in their careers. The skillset shifts from executing tasks to managing AI-enabled workflows.

Why this matters for AI builders: the judgement gap

Cheney Hamilton of research firm Bloor puts the core problem simply: AI raises the floor of what a junior can produce, but it doesn't give them the judgement to know when the output is wrong. That judgement historically came from repetition and consequence. If you automate the repetition, you also automate away the learning mechanism.

For product teams building AI tools for internal use or for customers, this matters directly. If your tool replaces the grunt work that used to build context, you need to decide whether you are creating output or creating exposure. A junior who only polishes AI drafts without understanding why certain choices were made will struggle to develop the strategic thinking needed for senior roles.

Redesigning training for an AI-native workforce

Some companies are already experimenting with new models. FDM Group trains people by presenting real business problems and asking them to find solutions using agentic engineering, rather than just teaching tool usage. Hamilton recommends treating entry-level work as exposure, not output. Let juniors check and correct AI-generated work while senior colleagues deliberately teach the judgement behind it.

A Klarus survey found that 45% of mid-market leaders said AI was being used to help juniors work better and faster, and nearly a quarter said it created new roles and opportunities. The jobs are not disappearing. They are being redefined.

What's still unclear

The data in this coverage is heavily UK-focused and relies on a single CNBC article with related industry commentary. Broader cross-industry numbers, especially for software engineering specifically, are thin. It is not yet clear how many companies will invest in deliberate judgement training versus simply expecting AI fluency to emerge. The risk that Hamilton identifies, a generation with AI-enabled breadth and no depth, remains a warning rather than a measured outcome.

The real task for AI builders is not deciding whether to replace juniors. It is designing tools and workflows that let juniors build judgement even as the work itself is automated. If your platform strips away the mistakes and consequences that build expertise, you need to intentionally build them back in.