Is AI Replacing Engineers? What the Headcount Data Shows

#Don’t learn to code?

In February 2024, Nvidia CEO Jensen Huang took the stage in Dubai and told the audience to stop telling kids to learn to code. For years, tech leaders called computer science a vital skill. He now believed that AI would handle the coding, so young people should study biology or farming instead. The headlines read “The death of coding.” “Don’t learn to code.”

Enrollment in computer and information science programs at four-year universities fell 8.1% in 2025-2026, the steepest one-year decline of any major since at least 2020 (National Student Clearinghouse). The computer science major specifically fell 11.2%.

The story shifted from CS being a valuable school to the idea that AI writing code means computer science isn’t worth learning. I wanted to know whether the headcount numbers at major tech companies actually backed up this story. If software engineers were being replaced, it should show in headcount, and it should be impacting the roles right next to them too, the designers and product managers who build software alongside engineers.

But it isn’t. Instead, engineering headcount has grown.

Background

I set out to see whether the headcount growth matched the “AI is taking the engineering jobs” narrative.

Historically, Software engineering in tech companies has been a sought after role. These roles pay well, offer a lot of autonomy and responsibility for features and products used by millions of people worldwide. Software engineers are key members of the product development team, alongside designers and product managers.

So if software engineers are being gobbled up by AI, what about the adjacent product roles? Are they affected too? To find out more, I wanted to look at AI exposure scores against LinkedIn headcount growth data to see whether exposure correlated with shrinking roles.

Defining AI exposure

One of the two main data sources for this analysis is the Yale Budget Lab, which blends six distinct exposure metrics into a single, PCA-weighted composite score.

AI exposure metrics measure what share of a job’s tasks could be performed by AI or assisted with AI. High exposure doesn’t mean the job disappears, but it can mean AI replaces some of the tasks workers perform. In fact, AI could be augmenting the task, rather than replacing the worker.

The Yale analysis found that the metrics mostly agree on who is not exposed, for example, plumbers being the least exposed. A separate problem appears in the coding field itself.

The underlying job taxonomy

The government SOC (Standard Occupational Classification) taxonomy splits the role into Computer Programmers (5.21) and Software Developers (1.01). The industry has largely abandoned the term “computer programmer,” but this is what shows up in the official taxonomy and is used to measure AI exposure.

Seeing these charts without context causes understandable career anxiety. No one wants to see their job listed as AI-exposed. To see the reality behind the metrics, I used the Budget Lab data to recreate the baseline exposure framework.

Measuring headcount growth with LinkedIn

To layer real world activity over these exposure scores, I pulled LinkedIn headcount data. This data measures the one year net change in headcount based on the number of members listing a specific title, giving a read on whether a role grew or shrank, not a count of hires.

LinkedIn data is useful because it allows us to look at certain roles within specific companies, which provided the detailed level I needed to compare against role by role exposure scores. The trade-off is that LinkedIn reflects self-reported profiles from a handful of companies over one year, making this directional rather than definitive.

The Yale occupations use the SOC classification, the standard government occupational taxonomy that the underlying O*NET metrics map to, so the names don’t match LinkedIn’s job titles. I built a crosswalk to map LinkedIn titles onto the Yale occupations.

amazon = pd.read_excel('amazon.xlsx', sheet_name='Titles')
amazon = amazon.rename(columns={'1y growth': 'growth'})
# growth = net one-year headcount change per title (not gross hires)

crosswalk = {
    'Software Engineer': 'Software Developers',
    'Product Designer': 'Web and Digital Interface Designers',
    'Senior Product Manager': 'Computer Occupations, All Other',
    # ...35 title-to-occupation mappings total
}

Note on Product Management: Modern product management roles don’t have a dedicated classification for a tech product manager, so PM titles fall into a generic “Computer Occupations, All Other” bucket. That means the exposure score attached to PMs isn’t really measuring product management work, so I treat anything in that category with caution. More on this in the caveats.

The three company snapshot

I worked with the available LinkedIn data. The initial pull covered Amazon, Google, Shopify, Klarna, and Meta, but cleaning left three files. Meta was a duplicate and was dropped. Klarna has only a few hundred employees, mostly in sales and operations, and almost no design or product management headcount, making a three-role comparison difficult.

That left Amazon and Google. and Shopify. This small sample is an illustrative case study rather than an industry census, but it offers an interesting contrast between companies. Two incumbents against Shopify, a company that has publicly embraced AI.

In April 2025, CEO Tobi Lütke made AI use a baseline expectation, asking teams to prove a job couldn’t be done by AI before requesting new headcount.

If the “AI displaces engineers” narrative held true anywhere, it should be visible here. Instead, Shopify’s engineering headcount outpaced its peers. While its smaller base means a few key hires shift percentages more sharply than they would at scale giants like Google or Amazon, the headcount tells an interesting story.

Key Insights

With both datasets in hand, I could line up exposure against actual hiring. If the panic were right, the most exposed roles should be the ones shedding people. Here’s every occupation I could match, sorted by exposure.

# join LinkedIn titles to Yale exposure scores via the crosswalk
jobs = crosswalk_df.merge(combined_companies, on='title') \
                   .merge(tech_yale, left_on='yale_occupation', right_on='occupation')

# average exposure and 1-year headcount growth per occupation
chart_data = jobs.groupby('yale_occupation').agg(
    growth=('growth', 'mean'),
    exposure=('ai_score', 'mean'),
).reset_index()

What the data shows (Amazon, Google, Shopify; one year net headcount change):

• Engineering headcount grew. Software Developers sit at the bottom of the exposure range and added people (about +2% weighted by headcount, more on an unweighted basis). For a low-exposure role, growth is what you’d expect.
• High AI Exposure roles mostly shrank. Of the four most-exposed roles, three contracted: Web and Digital Interface Designers (-3%), product management (-2%), and Web Developers (-11%). So “high exposure maps to shrinking teams” holds loosely. It applies to designers and web developers, not to software engineers.
• Some jobs with high AI Exposure grew in headcount. Database Architects are at the top of exposure (7.06) and grew 10%, while Information Security Analysts grew 23% at mid-high exposure. Two strong exceptions in a set of nine drag the overall correlation to about -0.18, weak and noisy. So I can say the most exposed roles tended to contract, not that exposure reliably predicts headcount.
• Two underlying job codes, one job. Remember Computer Programmers at 5.21 versus Software Developers at 1.01? The coding role, “programmers”, is a separate, older SOC category that BLS projects to keep shrinking and that few companies even hire under that title anymore. The role engineers really hold, Software Developers, is rated low in AI exposure and with high disagreement.

Caveats on what the data doesn’t show:

To keep this analysis on the up and up, let’s talk about a few things:

• Correlation isn’t causation. This is a one-year snapshot of self-reported profile data at three companies. I can’t say why adjacent roles contracted while engineering grew.
• No seniority or hiring volume. Net headcount can’t separate slowed hiring from layoffs, and there’s no junior-versus-senior breakdown, so I can’t say whether companies are not hiring junior talent while keeping senior people.
• Taxonomy limits. Product management is poorly cataloged in standard labor data, so the contraction in that category requires a skeptical read.
• Not a read on all of tech. Three companies, with small headcounts in design and PM. It might reflect a broader trend, but this data can’t establish that.

What I think could be going on

TThis is my own hypothesis based on my knowledge of how tech functions operates and the patterns I’m seeing, not a finding.

Open engineering roles are also up, according to Lenny’s Newsletter (Paid), drawing on TrueUp data and referenced by Business Insider, which reported around 67,000 open engineering positions globally as of March 2026, roughly 78% above the early-2023 low, with the increase accelerating into 2026.

If AI automates parts of engineering, it also creates adjacent work, setting up the models, debugging what they produce, and maintaining the systems built around them, and that work still needs engineers. That could explain this phenomenon.

I’m less sure about design and product management. The data does show less design headcount. It could be AI substituting for designers, design getting cut as a nice-to-have under tighter budgets, or design work being absorbed into other roles like PM.

I’ll follow up specifically on the design numbers. Product management is harder to discern since it maps to a catch-all occupation code that doesn’t really describe the role.

Bottom Line

The fear that high AI exposure translates into immediate headcount loss isn’t supported by the headcount patterns at these companies. At these companies, in this window, the most exposed role grew, and the roles that shrank weren’t the most exposed.

A year of headcount at three companies can’t tell you what AI is doing to a profession. But, maybe high exposure looks less like a death sentence for a career and more like a sign that the work is changing. Into what, and how fast, is what we don’t know yet. But this is something we can watch and see how it develops.

References

Business Insider. (2026, April 3). AI isn’t killing software coding jobs — they’re booming, according to TrueUp data. https://www.businessinsider.com/ai-isnt-killing-software-coding-jobs-booming-trueup-2026-4

Gimbel, M., Kendall, J., & Kulsakdinun, R. (2026, February 19). Labor market AI exposure: What do we know? The Budget Lab at Yale. https://budgetlab.yale.edu/research/labor-market-ai-exposure-what-do-we-know

O*NET OnLine. (n.d.). Summary report for: 15-1255.00 — Web and digital interface designers. U.S. Department of Labor, Employment and Training Administration. Retrieved June 2026, from https://www.onetonline.org/link/summary/15-1255.00

Rumage, J. (2026, March 12). Computer science degrees are losing popularity in the AI era. Built In. https://builtin.com/articles/computer-science-degree-decline-ai

Rachitsky, L. (2026, March 24). State of the product job market in early 2026. Lenny’s Newsletter. https://www.lennysnewsletter.com/p/state-of-the-product-job-market-in-ee9