# Age Predicts Who Adopts AI. Skill Predicts Who Benefits.

Published: 2026-07-08T13:00:00.000Z · Updated: 2026-08-01T15:07:06.000Z · Author: Rod Amora · Canonical URL: https://rodamora.com/blog/age-predicts-who-adopts-ai-skill-predicts-who-benefits/

> AI lifts the least experienced workers most, while AI expertise peaks mid-career. That inversion breaks the service-firm pyramid. The fix I keep seeing work: sort work by risk, not seniority, and change what juniors and seniors are for.

When researchers gave a generative AI assistant to [5,179 customer support agents](https://www.nber.org/papers/w31161?ref=g.rodamora.com), the least experienced agents resolved [about 35% more chats per hour](https://mitsloan.mit.edu/ideas-made-to-matter/workers-less-experience-gain-most-generative-ai?ref=g.rodamora.com). The most skilled agents gained almost nothing. That 2023 study, by Erik Brynjolfsson, Danielle Li, and Lindsey Raymond, is the most important org-chart finding of the AI era, and most service firms have not absorbed it yet.

The people most likely to adopt AI are young. The people most likely to benefit from it are the least skilled at the job. And the people who understand it best are neither: self-reported AI expertise peaks in the late 30s and early 40s. Age predicts adoption. Skill predicts benefit. The traditional service-firm pyramid assumes those two things point the same direction, and they no longer do.

![Two arrows diverge from a box marked the traditional service-firm pyramid: age predicts adoption and skill predicts benefit](https://rodamora-uploads-347628392068-sa-east-1-an.s3.sa-east-1.amazonaws.com/illustrations/age-predicts-who-adopts-ai-skill-predicts-who-benefits/fig-diagram-1.jpg "Adoption and benefit no longer point the same direction")

So what does the org chart look like when that assumption breaks? I have watched hundreds of service firms adopt AI through a franchise network's delivery data, and the firms getting this right are not flattening the pyramid or cutting the bottom of it. They are re-sorting the work: low-risk work goes to juniors working with AI, and high-stakes work stays attached to a senior.

## The pyramid's math just broke

The classic professional-services pyramid is a pricing structure. Juniors produce work cheaply, seniors review it, and the client pays for the gap between junior cost and senior-quality output. Every staffing ratio, every review layer, every billing rate rests on that spread.

![Cutaway drawing of a stepped pyramid building with workers at desks on the wide lower floor, fewer figures conferring on a middle floor, and one figure alone at the peak](https://rodamora-uploads-347628392068-sa-east-1-an.s3.sa-east-1.amazonaws.com/illustrations/age-predicts-who-adopts-ai-skill-predicts-who-benefits/fig-illus-1.jpg "The pricing spread the pyramid was built to protect")

AI collapses the spread at the production layer. If a first-year analyst with a good AI workflow produces something close to senior-shaped output, the firm can no longer price the difference the old way. The MIT study puts a number on it: roughly 35% productivity gains for novices, [flat gains for experts](https://mitsloan.mit.edu/ideas-made-to-matter/workers-less-experience-gain-most-generative-ai?ref=g.rodamora.com) on the same tasks.

The labor market has already responded, and the response is crude. [2025 follow-up work by Brynjolfsson and colleagues](https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/?ref=g.rodamora.com) found AI-era job losses hitting workers aged 22 to 25 hardest in AI-exposed occupations, while [employment grew for workers 30 and older](https://generations.asaging.org/not-left-behind-older-workers-artificial-intelligence-and-the-data-behind-adoption-and-adaptation/?ref=g.rodamora.com), including those over 50. Firms are cutting the people AI helps most.

That is the wrong lesson from the data. The gap AI closed is production. The gap it left open is judgment, and juniors were never priced on judgment. Cutting them saves salary and destroys the one layer of the firm that gains the most from the technology.

## Sort work by risk, not seniority

Across the firms I have watched, the durable pattern is a re-sort of work rather than a re-sort of people. The question stops being "who is senior enough for this?" and becomes "what happens if this goes wrong?" That question is the whole distance between Stage 0 and Stage 1 on the [Delivery Model Ladder](https://rodamora.com/delivery-model-ladder/?ref=g.rodamora.com): individuals already use AI on their own, and the firm's job is to sanction it and decide who does what by risk.

The test is short. High-risk work is work where an error costs money you can't claw back, damages the firm's name, or draws a regulator. Everything below that line can go to juniors working with AI. Everything above it stays attached to a senior, no matter how good the tooling gets. One amendment: work only belongs below the line if [a reviewer can catch an error](https://rodamora.com/blog/delegate-the-inputs-own-the-outputs/?ref=g.rodamora.com) by inspecting the output. If a mistake is invisible in review (bad data cleanup is the usual culprit), it stays above the line no matter how routine it looks.

In practice the split looks like this. Internal research, first drafts, meeting follow-ups, and routine client deliverables move below the line. Juniors run them with AI, output volume goes up, and review gets faster because it checks work against a known standard rather than teaching someone to produce it. Some firms hire more juniors for this tier, not fewer, because each one now gets far more done than the old model allowed.

Above the line sits everything that fails the test. Pricing decisions, final recommendations, anything that ships under the firm's name to a board. That work keeps a senior attached, and the senior uses AI too, just differently.

Below the line

Above the line

The test

An error is cheap to fix and visible in review

An error costs money you can't claw back, the firm's name, or a regulator's attention

Typical work

Internal research, first drafts, meeting follow-ups, routine deliverables

Pricing, final recommendations, anything a board sees

Who produces it

Juniors working with AI

A senior, using AI for research and artifacts

What review does

Checks output against a known standard

Makes the judgment call, and teaches it

If you run a firm, the applied version takes an afternoon. List your deliverables. Mark each one against the three-part test: unrecoverable money, the firm's name, or a regulator. Then look at who actually produces each item today. Most owners find seniors spending hours below the line and juniors locked out of tools that would multiply their output.

## Seniors stop being producers and start being multipliers

The flat expert gain in the MIT study is easy to misread. It measured production speed on routine tasks, and by that measure AI does little for someone who already knows the answers. It says nothing about what happens when a senior points AI at the hard problems.

What I have seen is that seniors attached to complex work deliver far more impact when they lean into the tools: deeper research in the same billable window, better artifacts, workflows that used to need a support team. The senior's job shifts from producing the work to multiplying it. They make the risk calls, set the quality bar, hold the client relationship, and use AI to go further on the problems that justify their rate.

![Drawing of a branching structural truss rising from a single base, with figures standing at each fork directing the branches outward](https://rodamora-uploads-347628392068-sa-east-1-an.s3.sa-east-1.amazonaws.com/illustrations/age-predicts-who-adopts-ai-skill-predicts-who-benefits/fig-illus-2.jpg "From producing the work to multiplying it")

The broader data points the same direction. Older workers already sit in judgment-heavy roles that AI helps rather than replaces: [49.4% of workers over 50 hold AI-insulated positions, against 42.2% of younger workers](https://generations.asaging.org/not-left-behind-older-workers-artificial-intelligence-and-the-data-behind-adoption-and-adaptation/?ref=g.rodamora.com). They use it for practical tasks, not experiments: 67% to find information and 40% to analyze data. That is what judgment work with AI looks like from the inside.

This changes what clients pay for. What a client buys from a senior is no longer production capacity. It is the risk call, the taste, and the accountability. Those got scarcer relative to output, which means they got more valuable.

## Innovation now flows up the chart

The pyramid used to move two things: knowledge flowed down through apprenticeship, and work flowed up through review. AI adds a third flow, and it runs bottom-up.

![Stepped pyramid with three arrows: knowledge flowed down, work flowed up, and innovation now flows up the chart](https://rodamora-uploads-347628392068-sa-east-1-an.s3.sa-east-1.amazonaws.com/illustrations/age-predicts-who-adopts-ai-skill-predicts-who-benefits/fig-diagram-2.jpg "The third flow the pyramid never had")

Juniors [adopt the tools fastest](https://rodamora.com/blog/shadow-ai/?ref=g.rodamora.com) and become the firm's evangelists. [74% of both Gen Z and Millennials](https://www.deloitte.com/global/en/issues/work/genz-millennial-survey.html?ref=g.rodamora.com) already use AI in their day-to-day work, per Deloitte's 2026 global survey. In the firms I have observed, juniors routinely find faster or better ways to do things before anyone senior does, and that contrast pushes the more open-minded seniors to rethink their own process.

One example from the delivery data has stayed with me. A junior consultant built an agent workflow on their own initiative. It reads the day's meeting transcripts, saves every agreed task into the project management software, and sends a summary of what was decided to both the consultant and the client. Nobody asked for it. It removed a chore that had quietly eaten senior hours for years, and it became standard process.

That kind of contribution never appears on a traditional org chart, because the chart assumes process improvements come from the top. An [AI-first firm](https://rodamora.com/blog/what-is-an-ai-native-service-business/?ref=g.rodamora.com) treats its junior layer as a working R&D function. If your tooling ideas need senior sign-off before anyone can try them, or your juniors are never in the room when delivery gets discussed, you are paying for your fastest adopters and then blocking them.

## Who runs the redesign, and the pipeline you must not break

The instinct is to hand the AI agenda to the youngest person in the room. The data says otherwise. In McKinsey's January 2025 Superagency report, [62% of employees aged 35 to 44 reported high AI expertise](https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work?ref=g.rodamora.com), against 50% of Gen Z respondents aged 18 to 24 and 22% of Boomers over 65. Slack's 2025 Workforce Index found the same shape: workers 28 to 43 are the [most likely to use AI daily, at 33%](https://slack.com/blog/news/the-new-ai-advantage?ref=g.rodamora.com), edging out Gen Z at 28%.

That band is exactly who runs delivery in a $1M to $20M firm. Your delivery lead combines enough hands-on fluency to judge the tools and enough experience to judge the work. The re-sort described above is theirs to own. The intern evangelizes; the delivery lead decides.

![Drawing of a raised middle platform with a figure at a control panel, connected by a channel to a worker at machinery on one side and a distant figure on a stepped platform on the other](https://rodamora-uploads-347628392068-sa-east-1-an.s3.sa-east-1.amazonaws.com/illustrations/age-predicts-who-adopts-ai-skill-predicts-who-benefits/fig-illus-3.jpg "The middle band that runs the sort")

One pipeline needs deliberate protection while all this happens. Nearly half of Gen Z workers say they [turn to ChatGPT before asking their manager](https://builtin.com/articles/ai-generational-divide-work?ref=g.rodamora.com) a question, and in an October 2025 Resume.org survey [45% said ChatGPT knows them better than their boss](https://finance.yahoo.com/news/resume-org-survey-majority-gen-183200689.html?ref=g.rodamora.com). [The apprenticeship loop](https://rodamora.com/production-gap/?ref=g.rodamora.com) that used to run on a thousand small questions is re-routing through a chatbot.

That loop is how firms manufacture seniors. A junior can learn production from AI. They cannot learn judgment from it, because the model knows nothing about your clients, your pricing, or your risk line. So the AI-first org chart schedules what used to happen by osmosis. Seniors run reviews as [teaching moments rather than pure QA](https://rodamora.com/blog/ai-anxiety-and-the-output-validator-workflow/?ref=g.rodamora.com). Juniors rotate onto above-the-line work with a senior attached, on purpose, even when it costs some speed.

## Keep the juniors. Change the job.

The market is cutting workers in their early 20s at the exact moment the evidence says they gain the most from AI. For a service firm owner, that is a hiring bargain. Keep the junior layer, and point it at generating more value and bringing in more business, so the cost justifies itself instead of being defended.

The redefined roles fit on an index card. Juniors own below-the-line delivery volume, tool experimentation, and apprenticeship on judgment. Seniors own above-the-line work, quality, and client trust, amplified by AI rather than threatened by it. The delivery lead owns the sort.

Run the test on your own service list this week. Mark every deliverable by what an error would cost: money you can't claw back, the firm's name, a regulator's attention, or none of the above. Then compare that against who does the work today. The distance between those two lists is your org-chart redesign, already written.

If you run that exercise and the result surprises you, I would like to hear about it. Reply and tell me what moved.