Only 1 in 200 AI Applicants Gets Hired

The Numbers Behind the Shortage

Staffing firm KORE1 measures the demand-to-supply ratio for AI engineers at 3.2 to 1 (three employers for every qualified candidate). Senior AI engineer roles take 90 to 120 days to fill, against about 25 days for a generic software role, three to five times slower for the exact roles carrying the most business urgency. Nearly 20,000 companies cannot fill roles after 30-plus days of posting. The cost of that delay is not just the empty seat; it is the delayed roadmap, the interviewing load on existing senior engineers, and the risk the hire washes out anyway.

The squeeze shows up in recruiter workloads. Recruiting teams are 14 percent smaller than in 2021, yet they handle 93 percent more applications, Deloitte's data shows, and manage 40 percent more open roles. Hires per recruiter have dropped 43 percent. Technical roles now average 35 to 36 interviews and 26 interviewer hours per hire. Only one in 200 applicants is ultimately hired.

Meanwhile, 69 percent of employers report difficulty finding qualified candidates, and nearly half rank attracting them as their number-one challenge. Half cite lack of relevant experience as the primary obstacle — not pay. Nine in ten anticipate hiring obstacles in 2026, with skill gaps topping the list. The World Economic Forum projects a 40 percent skills gap by 2027. ManpowerGroup's 2026 Talent Shortage Survey finds 72 percent of employers reporting hiring difficulty, with AI skills surpassing all others as the most difficult to hire for.

The 3.2-to-1 ratio means every qualified candidate holds multiple offers simultaneously. Roughly half of employed tech professionals are already passively job-hunting, per Dice's 2025 Tech Salary Report. Counteroffers are constant. More than half of companies report lacking sufficient AI talent, and 54 percent cite lack of skills as their primary AI-adoption barrier. Experience compounds the premium: seniors earn two to three times junior rates across every region.

The Screen: Proof Over Potential

Gartner projects that by 2025, more than three-quarters of venture capital and early-stage investor executive reviews will be informed by AI and data analytics. The pressure to prove you can ship reliable, governable AI cascades down to hiring. VCs reportedly spend over 100 hours per startup evaluating investments; a regulated entity building AI competency from scratch applies comparable rigor because the cost of a mis-hire compounds across regulatory, reputational, and technical surfaces.

Three forces shape any serious AI screen in a regulated environment. First, bias in evaluation tools: PitchBook's VC Exit Predictor back-tested at 74 percent accuracy on historical exits but still showed a 1 percent gap in success predictions between male and female CEOs. Tech Brew found that four of six signals CB Insights uses are proxies for race, socioeconomic status, gender, and disability. A financial-market infrastructure company cannot afford a screen that replicates those patterns — its regulators would notice. Second, the "tool backwards" problem: enterprises buy platforms first, then retrofit process. A credible screen tests whether a candidate starts from the business problem (market surveillance, listing analytics, capital-formation workflows) and works backward to the model, not the reverse. Third, the trust-security-governance triad: "The most the key word is trust, the key word is security, the key word is governance," said Saurabh Atre of L&T Technology Services, "will define whether enterprises adopt AI successfully." A stock exchange lives or dies on trust; its AI screen must verify that a candidate builds systems auditable by examiners, not just accurate on a holdout set.

A multi-stage screen at similarly regulated firms typically runs like this: an initial take-home using anonymized exchange data (order-book snapshots, corporate-filing text, settlement timestamps) to assess whether the candidate can frame a problem, choose an architecture, and document failure modes. A live coding session focused not on algorithmic puzzles but on data-quality debugging: handling missing timestamps, reconciling conflicting feeds, explaining why a transformer might hallucinate a material fact in a prospectus summary. A system-design round where the candidate sketches an end-to-end pipeline (ingestion, feature store, model registry, monitoring, rollback) and defends each choice against latency, explainability, and audit requirements. A behavioral panel with compliance, legal, and product stakeholders to test translation skills: can this engineer explain drift detection to a general counsel who last coded in Fortran?

The talent math makes this selectivity rational. A LinkedIn post cited in the research calls the employment chart "a massive narrative explosion on employment in AI" and argues that coding and engineering jobs are the canary in the coal mine — "prepare for an avalanche." When the canary is a single role at a stock exchange, the screen isn't a gate; it's a stress test for the organization's own AI maturity. If the exchange can't define what "good" looks like in enough detail to filter for it, the hire will fail — and the exchange will have signaled that its AI ambition is performative.

Candidates who clear such screens share a pattern: they treat the interview as a design review, not an exam. They ask about data lineage, model governance, and incident-response playbooks before writing a line of code. They've shipped something regulated (fintech, medtech, defense) and carry the scars. The screen selects for people who have already internalized the constraints the exchange is only now learning to articulate.

What the Exchange Is Really Testing

The screening process at a long-term-focused exchange isn't filtering for generic AI competence. It's filtering for a specific intersection: engineers who can build reliable, auditable systems inside a regulated market infrastructure while thinking on the decade-long horizons the exchange was created to serve. The research around LTSE's leadership, its board-level survey data, and the AI tools its own insights team tracks all converge on a profile that looks less like a typical ML hire and more like a systems architect who treats probabilistic models as safety-critical components.

Start with the domain. LTSE operates an SEC-registered national securities exchange with a "Very Simple Market" (fully displayed quotes, simplified order types, no fees or rebates) that participates in the National Market System. Any AI deployed there touches price formation, order routing, or surveillance. A hallucination in a chatbot is embarrassing; a hallucination in a matching engine is a regulatory event. The exchange's founding principles require listed companies to measure success over long time horizons and publish detailed compliance policies. That same discipline applies internally. Candidates who have only optimized for benchmark leaderboards on public datasets will not pass the architecture review.

The leadership signals reinforce this. Eric Ries, LTSE's founder and chair, built the Lean Startup methodology around validated learning loops — not model accuracy metrics. His recent writing and speaking, including at the AI Conference in September 2026, frame AI as a force reshaping employment and organizational structure, not a feature to bolt on. Maliz Beams, LTSE's CEO, warned in the April 2026 CEO/board survey of 109 U.S. CEOs and board members that "more established organizations tend to underinvest in workforce adaptation relative to technology, leaving them exposed as AI adoption accelerates." Michelle Greene, board member and former interim CEO, added that companies with six-year-plus planning horizons "integrate AI into strategy, governance and workforce planning simultaneously." The screening tests whether a candidate can operate at that integration layer.

Reid Hoffman, a key LTSE backer and podcast guest, provides the clearest window into the intellectual bar. His undergraduate major was symbolic systems (AI and cognitive science combined), and he left the field in the late 1980s because "we're nowhere close to understanding human intelligence or cognition." He returned only when DeepMind demonstrated self-play at scale: applying massive compute to learning systems rather than programming knowledge in. He describes AI as a "meta tool" that amplifies every other tool, and co-founded Inflection to build personal intelligence agents. In the early OpenAI days, he worked directly with Sam Altman, Greg Brockman, Ilya Sutskever, and Mira Murati on a premise that the technology "isn't just owned by one or two big tech companies" but developed as "humanist technology." LTSE's screen looks for that same framing: candidates who reason about capability curves, compute scaling laws, and governance structures — not just prompt engineering.

The board survey quantifies what that framing means in practice. Only 16 percent of respondents anticipate net headcount reduction in 12 months, but 64 percent expect AI to shift roles and skills. Sixty percent of AI spending goes to software, tools, and platforms; just 27 percent to reskilling. Yet companies with long horizons are far more likely to treat AI investment as additive (89 percent versus 62 percent) and less likely to anticipate workforce cuts (33 percent versus 55 percent). They also maintain regular board-level AI risk review. LTSE's solitary AI hire will sit inside that governance gap — building tools the board can oversee, not black boxes the compliance team fears.

The VC-focused AI tools LTSE's own insights team catalogs reveal the technical stacks that matter for financial-market applications. Oracle DataFox and TechScout automate screening and market research across thousands of signals. Visible Alpha and ForwardLane compress analysis and monitoring cycles, adding sentiment analysis and anomaly alerting. Kanarys and DataRobot address bias and DEI metrics in model outputs. A candidate who cannot speak to model monitoring, drift detection, explainability, and bias auditing in production (not in a notebook) will not clear the technical rounds.

Mission alignment is the final filter. Hoffman recounts Jeff Weiner's advice: "Don't come work for me, come work for the mission." LTSE's mission is restructuring capital markets to reward long-term value creation. The AI role exists to advance that mission, not to publish papers. The screen tests whether a candidate can translate probabilistic output into deterministic guarantees that a regulator, a listed company, and a long-term investor can all rely on. That is a different skill set than the market currently rewards — which is exactly why the opening remains singular, and the bar remains high.

How Candidates Are Responding

Candidates facing multi-stage screens like LTSE's are not waiting passively. They are adapting in two divergent directions: some are building their own AI tooling to survive the initial filter, while others are opting out of any process that feels automated beyond the first touch.

The first response is tactical. One applicant learned that her résumé lost points for being two pages long, for inconsistent use of her middle initial, and for spelling out "percent" instead of using the % sign. Each fix raised her score. Tools like Jobscan promise to reverse-engineer the applicant tracking system, but Daniel Chait, CEO of Greenhouse, said no two ATSs behave alike and the logic shifts weekly. "As long as applicants believe AI is involved, they'll still try to game it using their own AI," a Wired investigation found. Another candidate instructed an LLM to comb listings, score them by work type, seniority, and salary requirements (separate bands for in-person, hybrid, and remote) and surface only the top matches. He effectively built a personal ATS to counter the employer's.

The second response is refusal. A Greenhouse survey of 1,200 U.S. job seekers in April found that nearly four in 10 have withdrawn from a hiring process because an AI interview was required. One applicant keeps a list of companies that use AI interviewers and simply does not apply. Another calls it a "hard no." A candidate whose nystagmus causes involuntary eye movement was repeatedly interrupted by software demanding she look straight ahead. "That was probably the lowest point in my search," she said. "It was just demoralizing that I couldn't be treated like a human being." Only 8 percent of candidates believe AI makes hiring fairer, yet 80 percent of applicants in a Chicago Booth field experiment chose a voice-AI interview over a human when given the option, citing perceived objectivity, standardization, and convenience.

The arms race has produced a third, corrosive behavior: fabrication. One hiring manager screening 20 candidates for a technical AI role reported that at least half were "in some meaningful way, fake": experience that could not have happened as described, buzzword-perfect answers that collapsed on a single specific follow-up. A widely cited LinkedIn post put the share of candidates using AI to cheat in interviews at around 38 percent.

Chait calls the dynamic a doom loop: each side deploys AI to solve its own volume problem, worsening the other side's signal-to-noise ratio, which triggers more AI deployment. His advice to job seekers is blunt: "It's not you; it's the system, and it stinks. The answer is not just more spray-and-pray." For a role like LTSE's (one opening, a deep technical screen, no high-volume funnel), the candidates who advance will be those who treat the process as a genuine technical conversation, not a keyword-matching exercise. The screen is designed to detect the difference.

Why This Hire Makes the Exchange Credible

LTSE's solitary AI hire makes sense only when you understand what the exchange actually is. Founded by Eric Ries in 2016, approved by the SEC in 2019, and live since September 2020, the Long-Term Stock Exchange was built to "reverse the epidemic of short-term thinking" that Ries told CNBC makes it "difficult for companies to do the right thing." Its five SEC-approved principles (stakeholders, strategy, compensation, governance, investors) read like a governance framework waiting for a technology that can actually enforce it at scale. AI is that technology.

The leadership roster explains the urgency. Ries has spent years arguing that the definition of profit used for the last 25 years is wrong — that real profit maximizes human flourishing. Reid Hoffman, a longtime Ries collaborator, LinkedIn founder, OpenAI and Microsoft board member, and Inflection AI co-founder, has called the current moment a "cognitive industrial revolution" bigger than the internet, mobile, and cloud combined because "it combines them and crescendos them." On LTSE's own podcast, Hoffman echoed that view, describing AI as a meta tool enhancing every other tool; Ries said we're living in "the last years of the pre-AI revolution."

The data bears out the governance gap. LTSE's April 2026 survey with Chief Executive Group (109 CEOs and board members) found a sharp divide along planning horizons. Companies with strategies extending six years or more were far more likely to describe AI investment as additive to existing priorities (89 percent versus 62 percent for shorter-horizon peers) and far less likely to anticipate net workforce reductions over five years (33 percent versus 55 percent). Those same long-horizon companies shared three traits: they integrated AI into enterprise strategy, invested in both technology and workforce development, and maintained regular board-level review of AI risks and opportunities. As Greene summarized, "That alignment is what will allow them to capture value while managing risk more effectively over time."

LTSE needs to be in that cohort — not just philosophically, but operationally. An exchange that lists companies on long-term principles needs to evaluate those principles continuously. Market surveillance, listing compliance, stakeholder metric tracking, and governance scoring are data problems at heart, and they're exactly the class of problem where AI tools (the same ones VCs now use to screen thousands of startups via Oracle DataFox, Visible Alpha, ForwardLane, TechScout) deliver leverage. The survey showed 60 percent of such spending flows to software, tools, and platforms; only 19 percent goes to new technical talent. LTSE's single opening suggests it understands the sequence: you hire the talent first, then you build the tools, then you avoid the trap Beams identified (buying technology without the workforce adaptation to use it).

There's a signaling function too. Private companies in the survey were more than twice as likely to report active AI talent investment (33 percent versus 15 percent) and three times more likely to call AI core to competitive strategy (28 percent versus 9 percent). LTSE courts exactly those private companies (Asana listed in 2021, ThredUp dual-listed in 2023) and its pitch rests on being a partner that shares their time horizon. A stock exchange that can't demonstrate AI competence in its own house loses credibility with the very issuers it wants to attract.

The governance gap is stark. Only about one in five surveyed leaders described themselves as "very" or "extremely" confident in their organization's ability to manage AI risks. Thirteen percent of boards have no formal AI review at all; among sub-$100 million revenue companies, that rises to one-third. LTSE's principles demand better. Principle 2 (Strategy) and Principle 4 (Governance) cannot be implemented well without technical fluency at the exchange level. The hire isn't a luxury. It's the minimum viable investment to make the exchange's own thesis credible — a single seat, a multi-stage screen, and a three-to-one market that makes the filter the only way to win.

Working in AI? Zero G Talent tracks the openings: see every open Databricks role, browse AI jobs, openings at Anthropic, and the people building the field.