Strada Asked 1,500 Execs and AI Literacy Came Last. Hire Juniors.
Strada surveyed 1,500 execs and ranked AI literacy dead last for juniors. Here is how to source the entry-level pool the headlines say does not exist.
If you read only the Stanford HAI 2026 AI Index, you would freeze your junior pipeline this week. If you read only the Strada Institute survey that dropped May 20, you would double it. Both reports are correct, and the gap between them is where founders and engineering leaders are about to make a very expensive mistake.
The contradiction is not actually a contradiction
The Stanford HAI 2026 AI Index, released April 13, runs over 400 pages. The headline number everyone quoted: US software developers aged 22 to 25 are down roughly 20% since 2024, while older developer headcount grew. AI skill demand in the information sector jumped from 7.8% to 13.2%. The conclusion most readers drew: AI ate the junior engineer.
Then on May 20, the Strada Institute for the Future of Work published a survey of about 1,500 executives and senior talent leaders across every industry and firm size. In 2025, 46% said AI increased entry-level hiring against 13% who said it decreased it. That is roughly four to one in favor of more juniors. For 2026, the ratio narrows to 2.7 to 1, but it still skews positive. The same week Meta notified 8,000 employees, canceled 6,000 open requisitions, and Intuit cut 3,000.
These findings look like they are at war. They are not. Stanford is reading ADP payroll data: who is currently drawing a paycheck as a developer aged 22 to 25. That is a lagging indicator dominated by FAANG, Bay Area, and Seattle. Strada is asking employers across healthcare, manufacturing, finance ops, mid-market SaaS, and regulated industries what they plan to do next. Both are true. FAANG entry hiring collapsed. The rest of the economy is expanding AI-augmented junior roles.
If you are running talent at a Series A through Series C company, the Strada number is the one that matters to you. The Stanford number describes your hiring pool, not your competition.
AI literacy ranked dead last
Strada asked employers to rank the skills they actually want from entry-level hires. The order, from most to least valued:
- Critical thinking
- Communication
- Collaboration
- "Work experience in a similar role" (beat GPA and academic honors)
- ...several more...
- AI literacy, dead last
AI literacy was the only skill in the survey where graduate performance was rated higher than employer demand. Supply already exceeds appetite. Every bootcamp, every CS curriculum, every LinkedIn Learning path has spent two years optimizing for the one thing employers care about least.
This has a direct, painful consequence for sourcing junior developers. If you paste "AI literacy" or "prompt engineering" or "LLM" into a boolean string on LinkedIn Recruiter, you are filtering for the most commoditized signal and against what employers actually said they want. You are sourcing against the demand curve.
The skills employers ranked highest are also the ones boolean cannot find. There is no keyword for "judgment." There is no skill tag for "shipped a real project in a regulated industry and talked to actual users." Recruiters either reverse-engineer proxies (specific employers, specific repos, specific project descriptions) or they give up and post on Indeed.
This is where natural-language search beats keyword stacking. With Refolk, you describe the person the way a hiring manager would describe them ("junior backend engineer, two years of real production experience, worked on something with compliance constraints, not from a FAANG") and get a ranked shortlist. The hiring manager's sentence is the query. You stop translating intent into 14 nested ORs.
The companies going deepest on AI hire more juniors, not fewer
The single most counterintuitive finding in the Strada data: 59% of firms with strategic, company-wide AI integration reported growth in entry-level hiring. The most AI-mature companies hire more juniors, not fewer.
The mechanism is straightforward. When AI handles routine tasks, the marginal junior ramps into judgment work faster. Onboarding costs drop. The break-even point on a new grad moves from month nine to month four. 60% of tech employers who explored AI saw entry-level roles get more analytical and judgment-based. 54% saw routine task volume drop. The economics flipped: juniors are now a cheaper way to buy judgment hours, not a more expensive way to buy task hours.
If you are sourcing right now, target the AI-mature companies, do not avoid them. They are quietly the largest buyers of entry-level talent in 2026.
Recruiters who paste AI literacy into boolean are filtering for the one skill employers ranked last.
Why "Junior Software Engineer" returns almost nothing
Run a US-only search for the exact title "Junior Software Engineer" on any major sourcing platform. Refolk's index returns about 2,077 profiles. That is the entire boolean-visible US junior pool. It is comically small relative to the actual entry-level developer market.
The reason: real juniors do not get titled "Junior Software Engineer." They get titled "Software Engineer," "Software Engineer I," "Software Engineer II," "Associate Software Engineer," "Member of Technical Staff," or no level marker at all. The top current employers of self-identified "Junior Software Engineer" titles in our index skew heavily non-FAANG: Deloitte, Walmart Global Tech, Walmart eCommerce, Asana, and a long tail of consultancies. Those are not bad pools. They are just not the pool.
Meanwhile, Meta's reorganization is creating titles that did not exist six months ago. The four new AI-focused orgs absorbing roughly 7,000 internal transfers are Applied AI Engineering, Agent Transformation Accelerator XFN, Central Analytics, and adjacent pods. The new title categories include "AI Builder," "AI Pod Lead," and "AI Org Lead." None of those are in your boolean string. None of them were in your boolean string in November either. Any sourcing template built on conventional titles is now stale within a quarter.
The Class of 2026 is a 60-day sourcing window
Oracle revoked over 50 campus offers from new grads this quarter. Microsoft offered voluntary separation to 8,750 US employees. Meta canceled 6,000 open reqs on top of the 8,000 notifications. The "first job, first offer, first salary" door is closing at the top of the market at the same time senior reductions hit.
For a smaller AI-mature company, this is the cleanest cohort you will see all year. These are engineers who already cleared the highest filters: they got the Meta offer, the Oracle internship return, the Microsoft new-grad slot. Someone else paid for the screening. Now they are available, motivated, and willing to take a Series B title instead of a FAANG title because the FAANG title evaporated.
The window is short. Last cycle (2023) it closed in roughly 90 days. This cycle is moving faster because severance is shorter and the public narrative is louder. If you have not built a list by mid-July, you are sourcing from the leftovers.
The sourcing motion here is specific. You are not looking for "junior engineers" broadly. You are looking for a named cohort: people who held a specific role at a specific employer for less than 18 months, ending in May or June 2026. That is a one-sentence description, and it maps cleanly to a natural-language query. "Engineers who started at Meta, Oracle, or Microsoft in 2024 or 2025 and are now open to work" is exactly the kind of search Refolk's AI hiring junior talent workflow handles in a single prompt instead of a six-tab boolean.
What the Strada Institute report actually tells you to do
Read the Strada Institute report against your current sourcing playbook and three things should change this week.
Stop filtering for AI literacy
Drop "AI literacy," "prompt engineering," and "LLM experience" from your required-skills list for junior roles. Move them to "nice to have" or remove them entirely. Employers ranked these dead last. Candidates over-index on them. You are paying a premium for a commodity.
Filter for judgment proxies instead
The top four employer-ranked skills (critical thinking, communication, collaboration, similar work experience) do not exist as keywords. The proxies that work:
- Shipped production code at a company with real users (not just side projects)
- Worked in a regulated industry (healthcare, fintech, defense) where judgment is forced
- Has writing in public: blog posts, conference talks, RFCs, design docs
- Held a role for more than 12 months at a single employer (rules out churn)
None of these are searchable as strings. All of them are describable as a sentence, which is the entire premise of entry-level tech hiring 2026 looking different from 2022. The query language has to change because the signal moved.
Target AI-mature mid-market, not FAANG refugees only
The 59% strategic-AI-integration cohort is your competition for talent and your model for hiring. Look at which mid-market companies (10,000 to 50,000 employees) have publicly announced AI platform teams in the last 18 months. Those companies are hiring juniors aggressively, and their alumni are the best signal for your own pipeline.
The headline narrative is wrong for your company
The Stanford HAI 2026 AI Index is real data. The 20% drop in 22-to-25-year-old developers drawing paychecks is happening. But it describes a specific slice of the labor market (FAANG-adjacent, paycheck-stage, payroll-recorded) that is not where most companies hire from anyway.
The Strada Institute report describes the rest of the economy: 1,500 employers across industries and firm sizes, all saying the same thing. AI is making juniors more valuable, not less. The skills they need are the ones boolean cannot find. The titles they hold are not the ones your template has.
If you let the dominant narrative drive your 2026 hiring plan, you will under-hire juniors at exactly the moment they got cheaper to ramp and more available to recruit. The contrarian move is also the obvious one once you read the primary sources. Hire juniors. Just stop sourcing them by typing "AI literacy" into a search box.
FAQ
Does the Strada Institute report contradict Stanford's 20% drop in young developers?
No, they measure different things. Stanford HAI's 2026 AI Index uses ADP payroll data to track who is currently drawing a paycheck as a US developer aged 22 to 25, which collapsed about 20% since 2024. Strada surveyed roughly 1,500 executives about their forward hiring plans across every industry, and found 46% increased entry-level hiring in 2025 against 13% who decreased it. FAANG entry-level hiring fell hard. The rest of the economy is expanding AI-augmented junior roles. Both findings are true at the same time.
Why is "AI literacy" a bad keyword for sourcing junior developers?
Strada's data shows AI literacy is the one skill where graduate supply already exceeds employer demand, and it ranked dead last among skills employers want from entry-level hires. Filtering for it means you are selecting candidates who optimized for the most commoditized signal and against what employers actually said they value: critical thinking, communication, collaboration, and prior similar work experience. None of those higher-ranked skills are searchable as keywords, which is why natural-language sourcing tools outperform boolean for this cohort.
Which companies are actually hiring entry-level engineers in 2026?
The companies going deepest on AI. 59% of firms with strategic, company-wide AI integration reported growth in entry-level hiring. Mid-market and non-FAANG employers dominate the visible junior pool. Refolk's index shows Deloitte, Walmart Global Tech, Walmart eCommerce, and Asana as top employers of self-identified "Junior Software Engineer" titles. The mechanism is that AI handling routine tasks shortens junior ramp time, which improves the economics of hiring entry-level talent rather than worsening them.
How long is the window to hire the Class of 2026?
Roughly 60 to 90 days from late May 2026. Oracle revoked over 50 campus offers, Microsoft offered voluntary separation to 8,750 US employees, and Meta canceled 6,000 open requisitions on top of 8,000 notifications. That released a cohort of pre-screened, FAANG-filtered engineers into the market simultaneously. The 2023 cycle closed in about 90 days. This cycle is moving faster because severance packages are shorter and the public narrative is louder, so a mid-July deadline is realistic for any company that wants the cleanest picks.