Asher Cohen
Back to posts

Hire Juniors and Train Them Well

Why AI makes junior hiring more important, not less—and how organizations that maintain talent pipelines will outlast those optimizing for short-term efficiency.

I keep coming back to a simple belief: hire juniors and train them well.

Not because it's charitable. Because it protects the future of thinking.

There is a growing narrative that AI will replace the next generation of builders, designers, and engineers. I don't buy it. AI is powerful, yes. But it is still a tool—an amplifier. It processes patterns and probabilities at scale, but direction, judgment, and accountability still come from humans.

This is not just philosophy. It shows up in how leaders and companies are behaving.

Adobe has publicly framed AI as something that expands creativity rather than replacing it, repeatedly positioning generative tools as enabling more people to create and participate in design and media workflows rather than displacing creators. The emphasis is on augmentation, not substitution.

At the enterprise level, IBM has spoken openly about automation reshaping parts of its workforce while continuing to invest in roles that require human judgment, engineering depth, consulting capability, and early-career development. The signal is not "fewer people overall," but "different skills and training pipelines."

Across the industry, the language is consistent:

Microsoft positions AI as "copilots" that assist developers and knowledge workers.

Google describes AI as assistants embedded into workflows across productivity, development, and search.

Amazon frames AI as infrastructure for human innovation, not autonomy.

None of these products are framed as replacements for people. They are framed as force multipliers.

The economic reality reinforces this. Organizations still need:

  • institutional knowledge that accumulates over years
  • human oversight of AI outputs
  • domain judgment built from experience
  • leadership pipelines formed through progression

Seniors do not appear fully formed. They are trained, mentored, and shaped through junior roles. If companies stop hiring juniors, they are not just reducing headcount at the bottom of the ladder—they are removing the foundation of the entire structure.

Training juniors is how organizations:

  • build institutional memory
  • develop judgment that cannot be inferred from datasets alone
  • create future leaders who understand systems, constraints, and trade-offs

There is also a practical cost reality. Hiring experienced talent externally is expensive and volatile. Developing talent internally compounds capability over time and embeds context that AI cannot synthesize.

Yes, AI will automate repetitive tasks. And yes, some traditional entry-level roles will disappear. This is already happening in support, documentation, and routine analysis.

But the response should not be to stop hiring juniors. The response should be to change how we grow them:

  • give real production responsibility earlier
  • pair them with AI as scaffolding rather than a crutch
  • build structured mentorship tied to outcomes
  • measure learning velocity, not just output

Short term, this looks inefficient. A senior engineer with AI assistance can produce more immediately than a junior learning the craft.

Long term, it is the only sustainable path. Organizations that hire only experienced talent eventually lose the ability to grow their own expertise. They become dependent on external hiring markets and lose continuity in how knowledge evolves internally.

AI does not replace human thinking. It raises the ceiling for humans who are trained to think well.

The real risk is not that machines will replace people.

The real risk is that we stop developing the people who know how to use them.

If that happens, the loss is not jobs. It is capability, judgment, and the next generation of problem-solvers.

Additional context: what recent data and leadership signals show

1) Productivity gains are real—but uneven and supervision-heavy

Multiple 2024–2025 field studies across engineering, support, and knowledge work show measurable productivity gains from AI-assisted workflows (often 20–55% depending on task type). However, the largest gains appear in environments with:

  • strong senior oversight
  • defined problem spaces
  • feedback loops that refine prompts and outputs

Organizations without those structures often see rework increase, not decrease. This reinforces that AI effectiveness compounds with experienced humans—not in their absence.

2) Junior hiring slowdowns are already visible—and risky

Industry hiring data since 2023 shows a sharper contraction in entry-level roles than in senior roles across software, analytics, and design. The short-term logic is clear:

  • AI increases senior leverage
  • cost pressures reward immediate output

But the structural risk emerges 3–5 years later:

  • fewer mid-level candidates
  • fragile leadership pipelines
  • higher dependency on external hiring markets

Historically, industries that paused junior intake (post-dotcom, post-2008) experienced talent cliffs later.

3) AI governance demands human maturity

Regulation and enterprise adoption trends reinforce the need for trained professionals, not fewer of them:

  • model risk management requirements
  • auditability and traceability of AI outputs
  • safety, bias, and compliance controls
  • domain accountability (legal, medical, financial, infrastructure)

These functions cannot be outsourced to models. They require people who understand both the system and the context it operates in—skills typically built over years starting in junior roles.

4) Product strategy signals: augmentation is deliberate

Major platforms consistently design AI as embedded assistance:

  • copilots inside IDEs, not autonomous codebases
  • design assistants inside creative tools, not end-to-end brand systems
  • knowledge assistants inside enterprise workflows, not decision authorities

The architecture itself assumes a human-in-the-loop model. This is not transitional messaging; it is a design constraint driven by liability, trust, and accountability.

5) Capability compounds through exposure, not consumption

Juniors who learn with AI develop differently than those trained without it:

  • faster exposure to system-level thinking
  • earlier interaction with production complexity
  • more opportunities to critique outputs rather than produce from scratch

The skill shifts from "generate everything manually" to:

  • evaluate correctness
  • refine intent
  • understand system boundaries
  • manage trade-offs

These are senior skills, acquired earlier—but only if juniors are present in the system.

What this implies for hiring strategy

Organizations optimizing for durability—not quarterly efficiency—tend to do five things:

  1. Maintain a baseline junior intake regardless of automation gains.

  2. Pair juniors with seniors + AI, not AI alone.

  3. Redefine entry-level work from execution to supervised problem-solving.

  4. Instrument learning: measure decision quality, not just deliverables.

  5. Build internal mobility paths so AI-literate juniors become mid-level quickly.

This reframes junior hiring from cost center to infrastructure.

The structural argument

Every mature organization relies on a pipeline:

junior → mid → senior → leadership

AI increases the output of each stage. It does not create the pipeline itself.

If intake stops at the junior level, the system eventually hollows out:

  • fewer mentors
  • fewer architects
  • fewer decision-makers with historical context

AI can accelerate individuals inside the pipeline. It cannot replace the existence of the pipeline.

The long view

The shift underway is not "AI vs. humans." It is "AI + trained humans vs. AI + untrained humans."

The organizations that win will not be the ones that eliminated junior roles fastest. They will be the ones that redesigned them first.