Which Side of the AI Divide Is Your Company On?

Stanford tracked employment for workers aged 22–25 in AI-exposed roles and found a 16% decline since late 2022, accelerating month by month. The same data shows that companies in the top quartile of AI spending grew total headcount 10.2% over the same period and increased entry-level hiring 12%. Same technology. Opposite outcomes. The gap is not explained by industry or company size. It comes down to one question: is AI being used to cut costs, or to build capacity?

What happened and why it matters

The data comes from a Stanford study cited in The Neuron and TLDR on June 29. It captures two distinct AI adoption strategies running in parallel across the same economy.

Companies using AI to automate existing work cheaper — fewer people needed to produce the same output — are shrinking entry-level roles. Companies using AI to expand what each person can deliver — more clients, faster turnaround, new services — are hiring. Both groups are using the same tools. One sees AI as a substitute for headcount. The other sees it as a multiplier.

The –16% number describes one strategy. The +12% is the other. Most career anxiety about AI is focused on the first. Most of the career opportunity is in the second.

What to do differently on Monday

If your company is using AI primarily as a cost tool, the risk is real — and defensive. You become harder to cut when you are doing the part AI cannot: judgment calls, quality review, client relationships, decisions that require organisational context the AI doesn't have.

If your company is on the other side of the divide, the risk looks different: can you absorb the expanded scope without letting output quality slip? The professionals who hold their own here are the ones who learned early to frame tasks for AI precisely, review what comes back critically, and know when not to use it.

Both situations require the same underlying skill — knowing when AI is the right tool for a specific task and when it isn't.

Jamie: the coordinator deciding what the AI can do without her

Jamie is a content coordinator at a 50-person digital agency. Her company adopted AI to accelerate client content production — the sell to clients is faster turnaround at current rates. The internal result: each piece of work requires fewer human hours. Jamie's team has one fewer coordinator than it did eighteen months ago.

Jamie has not been cut. But her job has changed in a way nobody announced. She used to manage the production of content. Now she is most useful when she catches what the AI produces that would embarrass the client — the wrong tone, a claim the client can't back up, a draft that's technically correct but misses the brief.

That judgment — knowing when the output is ready and when it isn't — is what the agency can't automate. It is also something Jamie has never had to name or develop deliberately, because it used to be one step among many.

Marcus: the analyst at the company that's adding headcount

Marcus is a junior analyst at a 280-person professional services firm that recently committed to taking on 40% more client engagements per analyst — without immediately growing the team, but with plans to hire as the model proves out. His role is not under threat. It is expanding.

His firm is on the +12% side of the data. What that requires from him is different from what Jamie's situation requires. The risk is not being replaced. The risk is becoming the analyst who uses AI to produce volume without producing insight — whose outputs are fast but unreliable, requiring senior staff to fully rework them before use.

The professionals who hold their value when a company scales on AI are the ones who frame the first task correctly: specific enough that the output is usable, constrained enough that reviewing it takes minutes rather than another round of prompting.

The one-sentence version

The AI jobs conversation defaults to the –16%. The number that's actually useful for your career is the +12%.

<BlogPracticeSection />