After AI, which layer of you loses value first?

The part of you that loses value first after AI is the part you can write on a resume. The parts that only show up under pressure are gaining.

I once interviewed a candidate with a packed resume. Owned this, led that. I asked for one specific event: the situation, what he chose, how it turned out. He stalled. Looking back, that interview was a preview.

I look at people in four layers. The core is what you want and what you will never do. Patterns are who you become under stress. Equipment is skills, experience, track record. Context is whether you add or drain when placed with this company, this manager, this team. I used the same model in the exit survey post.

These four layers don't lose value at the same speed. Here is how I see it.

After AI, the equipment layer, what fits on a resume, loses value first

The four layers I use to read people (equipment, context, patterns, core), in the order I expect them to lose value. My judgment, not a measurement. Evidence as cited in the article: Stanford SIEPR (August 2025) and Deming (NBER, US, 1980 to 2012)

1. 01 Falls firstEquipmentKnowledge, templates, coding speed: whatever can be written down. Stanford: in the most AI-exposed jobs, employment for ages 22 to 25 fell 13% in relative terms. Now: write it down yourself first and make it your own method
2. 02 Half fallsContextA new hire learns in a week what inside knowledge used to buy. Fit with the manager and team stays. Now: pick jobs on team size, the manager, and whether you two work under stress
3. 03 Holds, more visiblePatternsDeming: the share of jobs needing heavy social interaction rose nearly 12 points. Now: prepare three to five events: situation, judgment, result and cost
4. 04 Holds mostCoreWhen tools let you do more, choosing what to do becomes the constraint. Now: write out the real reasons behind your last three job changes

Equipment stays as the entry ticket, but it no longer sets you apart. In interviews, bring events, not a list of what you owned.

Equipment falls first, especially what can be written down

Inside the equipment layer, whatever can be written down loses value first. Knowledge, playbooks, a language's syntax, speed at writing code. If it can be written down, a model can learn it, and a junior with good tools can get to 70 or 80 percent of it.

A Stanford paper from August 2025 used data from the largest US payroll software provider to see whose jobs moved first after generative AI spread. In the most AI-exposed occupations, employment for 22 to 25 year olds fell 13% in relative terms. More experienced workers in the same occupations held steady or grew. The drop was concentrated where AI tends to replace people rather than help them.

Why the young first? My guess is that a new graduate's equipment is almost all writable. Coursework, tutorials, practice. A senior person's equipment has other things mixed in, a feel for situations and lessons from things that went wrong. Those are hard to write down, so they don't move as easily. I wrote about experience being unbundled in When experienced stopped being a selling point.

Equipment doesn't go to zero. It's still the entry ticket. It just stops setting you apart.

Context: half of it falls

Part of your context advantage is just information. You know the internal processes, how a system is configured, who can approve things. That edge is getting thin. With documents, search and AI on top of internal knowledge, a new hire can learn in a week what used to take six months.

The other half holds. Do you work well with this manager and this team? That matters more as teams shrink. One person out of step in a team of five does far more damage than in a team of fifty.

Patterns hold, and get more visible

Patterns are you under stress. When a project is about to fail, do you tighten control or go quiet? When you're challenged, do you explain or push back?

AI can't stand in for this layer. And it is being weighted more. The economist David Deming found that from 1980 to 2012, the share of US jobs requiring heavy social interaction rose by nearly 12 percentage points, while math-heavy jobs with little social interaction fell by 3.3 points. That predates generative AI. It points the same way, though. The more machines can do, the more how people work together is worth.

In an interview, the only way to see this layer is through events. As I wrote in what I look for in an interview, interviewers write down events, not adjectives. That candidate stalled on exactly this layer. The equipment words were all there. There was no event that showed his patterns.

Core holds most, and is hardest to prove

Core is what you want and what you won't do. Once tools let you do far more, the hard part becomes choosing what to do. That choice comes from your core.

The trouble is that it's the hardest layer to show on a resume or in an interview. It only shows in past choices. Why you left, why you stayed, what you gave up. Write out the real reasons for your last three job changes side by side. I described how in the four-layer self-portrait.

What to do now

Keep building equipment, but don't hoard it. Write down yourself whatever can be written down, and turn it into your own method.

Shift your interview prep. Stop listing what you owned. Prepare three to five specific events, each with the situation, how you judged it, the result and the cost. In each one, be ready to say how you behaved under pressure.

When choosing your next job, look hard at the half of context that holds: team size, who the manager is, whether the two of you are likely to work under stress. Those will matter more than the company's name.

If you're just starting out, the entry door really is narrower. Find a role, even a small one, where you make calls and own the outcome. Patterns and core only grow in real events.

The Stanford data is from the US. Japan still hires graduates in one annual intake, so the effect may come later here, or in another form, like fewer graduate slots. There's no Japanese data of the same precision yet. I'm waiting for it.

The order of the four layers is my judgment, not a measurement. If you're seeing something different, tell me.

Which layer of yours do you most want to protect?

Parental leave is one case where the equipment and context layers decay on a schedule; I wrote about keeping your promotion track after parental leave in Japan.

What this means if you are changing jobs past 40 is in a post on selling judgment after 40.