Kiwi’s Four AI Security Roles Turn Its Screen Into a Gate

What the Job Board Reveals

Kiwi lists eleven open positions across four teams (six in Engineering, two in Product, two in Core Operations, one in Strategy) clustered in three European hubs: Brno (five), Prague (three), Barcelona (three). The raw count obscures the signal. Engineering dominates, and its titles reveal a specific thrust: two Senior Software Engineers for Flight Experience, two AI Engineer – Security roles, two Senior AI Engineer – Security roles. That trio of pairs (flight systems, AI security, senior AI security) accounts for the entire engineering pipeline.

Product seeks two Retention Managers, split between Prague and Brno. Core Operations wants a GL and Reporting Lead in each Czech city. Strategy has a single Senior Business Analyst for CS Analytics in Barcelona. No Customer Service or Commercial roles appear. The absences are as telling as the presences: Kiwi is not hiring for sales growth or support scaling. It is hiring for intelligence (both the artificial kind and the analytical kind) and for the financial rigor to manage what that intelligence produces.

The AI Security cluster is the sharpest indicator. Four roles, two at each seniority, posted simultaneously in Brno and Barcelona, suggest a deliberate build-out, not replacement hiring. Pairing "AI Engineer – Security" with "Senior AI Engineer – Security" implies a team structure: leads and contributors, duplicated across two sites for redundancy or time-zone coverage. Flight Experience carries only senior titles (no junior or mid-level posts), signaling the flight-search product is in refinement, not build. Retention Managers sit in Product, not Marketing, placing user lifecycle ownership with the people shaping the product surface. The lone Strategy analyst, focused on customer-service analytics, hints the next optimization target is the support operation itself.

Geographically, the Brno–Prague–Barcelona triangle mirrors a classic European tech footprint: Czech engineering depth, Spanish product and AI talent, enough overlap for follow-the-sun reviews. But duplicating AI Security roles in Brno and Barcelona (not Prague) suggests the security team anchors where model-training infrastructure lives, while Flight Experience stays in Prague where the core booking stack has historically sat. GL and Reporting Leads in both Czech cities point to a finance organization preparing for multi-entity complexity, possibly ahead of regulatory or structural changes.

Taken together, the board reads like a company securing its model layer, hardening its revenue engine, and instrumenting its support feedback loop — all at once.

The Screen Candidates Hit

The clearest public signal on Kiwi's screening timeline comes from a Glassdoor review dated April–May 2023: the candidate applied online and received an interview request in under a week, followed by a 30-minute phone screen. That single data point (recent, specific, first-hand) is the only documented stage in Kiwi's funnel appearing in open sources. Everything beyond it is inference or silence.

Older Indeed entries tell a different story, and the dates matter. A May 2021 FAQ summary reads "Come in answer questions know the job description details hired": language suggesting an in-person step and heavy emphasis on JD fluency. A February 2018 post calls the interview "easy" and puts the process at "about 2 days to a week depending on training availability." Both predate the current hiring surge by years; they reflect a different company size, talent market, and possibly a different business line (the 2018 reviewer references "Kiwi Services," not the divisions now recruiting). Treat them as historical texture, not current procedure.

No public trace exists of technical assessments, take-home projects, pair-programming sessions, or structured rubric-based panels, staples at comparable employers. Kiwi's public footprint lacks equivalents. That absence could mean the process is lean, customized per division, or that candidates simply aren't writing about it. The Glassdoor reviewer's 30-minute phone screen remains the only confirmed gate.

For applicants, the actionable takeaway is narrow but concrete: expect a fast first response (under a week per the 2023 report), prepare for a 30-minute phone screen covering motivation and JD alignment, and treat every subsequent stage as division-dependent. Candidates who cleared the phone screen report being asked to "know the job description details", phrasing that appeared in the 2021 FAQ. Master the JD's stated requirements; map each bullet to a specific project or metric from your background. That preparation pays off whether the next round is a whiteboard session, a take-home, or a panel — and it's the only lever you control before the process goes dark.

The Technical Profile: ERP, BI, and Lean: Not Robotics

Kiwi's postings on ZipRecruiter and Glassdoor reveal a technical profile leaning toward enterprise systems, manufacturing operations, and business analytics, not the frontier robotics or AI stacks the company's label might suggest. Two distinct postings from mid-2026 illustrate the split.

One listing, dated August 2026, seeks AS/400 and JDE (JD Edwards) experience, requiring three to four years in those systems. It explicitly prefers manufacturing background, corrugated supervisory experience, and prior work in a Lean environment. This is an ERP-and-operations role rooted in legacy midrange computing and shop-floor process improvement. The other, from June 2026, emphasizes proficiency across the Microsoft ecosystem (Excel, Word, PowerPoint, PowerBI, Power Pivot, SharePoint, Teams) and flags prior experience in a small non-profit, startup, or "unstructured environment" as preferred. A third Glassdoor entry lists customer service and restoration-industry experience, with a bachelor's in medical technology or related field required. A fourth pathway, cited from an education advisory site, notes a master's degree or applied business programme focused on Supply Chain & Logistics as the most common entry route.

Together, the requirements paint a company building out operational backbone roles: ERP administration, business intelligence, supply-chain coordination, regulated-industry compliance. The AS/400/JDE ask is notable; those platforms dominate mid-market manufacturing and distribution, and the corrugated-and-Lean specificity suggests Kiwi runs or services packaging and production lines. The Microsoft stack, particularly PowerBI and Power Pivot, signals a push to modernize reporting and self-service analytics atop that ERP core. The startup/non-profit preference hints that at least one opening sits in a newer, less structured division where process maturity is still being built.

What's absent is instructive. No Python, TensorFlow, ROS, CUDA, embedded C++. No cloud-native tooling (Kubernetes, Terraform, serverless). No computer-vision, sensor-fusion, or real-time control frameworks. The "AI-first" narrative appearing in broader New Zealand market commentary (where Employment Hero reported developers coding at roughly four times their prior rate and CX adoption jumping from 15% to over 70% after adopting the value in 2025) does not surface in Kiwi's own listed requirements. Manufacturing shows the largest wage premium for AI skills nationally (2.7% of postings, 9,600 roles in 2025), but Kiwi's ads don't yet reflect that premium in their stated tech stacks.

For candidates, the implication is clear: the screening will test ERP fluency, Lean execution, and BI storytelling before it touches model training or robotics integration. A resume leading with AS/400/JDE configuration, Kaizen events, or PowerBI dashboarding for shop-floor KPIs maps to the posted specs. One leading with LLM fine-tuning or SLAM algorithms must translate those skills into the language of process variance reduction and data-driven operational control, because that is the vocabulary Kiwi's job descriptions actually speak.

The AI Filter Reshaping Every Application

The New Zealand job market has shifted hard toward AI-mediated hiring. Employment Hero, whose local customer base has grown 60 percent in two years, ran more than 2,500 AI-led interviews in April 2026 alone. Its chief executive Neil Webster told BNZ Business Breakfast that employers are drowning in volume: "We're actually seeing for each role hundreds and hundreds of applicants applying for each role." That flood means the first gate is almost always algorithmic — and the playbook for clearing it looks different from the old advice.

First, stop feeding the bot the same prompt everyone else uses. A recruiter interviewed by Newsroom in February described a client who received 12 cover letters word-for-word identical because 12 different candidates had pasted the same job description into the same AI tool. The system flagged them instantly. "Authenticity is what sets candidates apart," the recruiter said. With 165,000 people unemployed (up 10,000 year-on-year) and the jobless rate at 5.4 percent, the shortcut that feels like a time-saver is the fastest route to the reject pile.

Second, treat the AI screen as structured data extraction, not conversation. Employment Hero's system ranks candidates against the job specification before a human ever sees a name. Webster was explicit: "Firstly they're using the AI to work out who should be on that shortlist... the AI does that and gives you a shortlist." Your résumé and any asynchronous responses must map cleanly to the keywords and competencies the posting lists. Mirror the language of the ad. Quantify outcomes. If the role asks for "Python automation of test pipelines," say "Built Python automation that cut test-cycle time 40 percent across 12 repos", not "experienced with test automation."

Third, know the human is still in the loop — but later. Employment Hero's firm rule: "We never want the AI to make a decision; we want the AI to make a recommendation and to save the employer time." The algorithm sorts; the hiring manager decides. So the strategy splits: optimize for the machine parse first, then prepare a distinct narrative for the human panel. The panel will probe the same claims the AI surfaced. Inconsistency between your AI-facing résumé and your live answers is a red flag.

Fourth, expect bias audits to be part of the process. Webster said Employment Hero trains its models against millions of candidates and reviews outputs for bias. That doesn't eliminate it — he acknowledged risk exists whether decisions are human or machine — but it means the system is tuned for consistency over gut feel. Candidates with non-standard backgrounds (career switches, gaps, overseas experience) should front-foot the translation: explicitly map each transferable skill to the role's requirements so the model doesn't have to infer.

Fifth, use the market's counter-intuitive signal. Despite rising unemployment, 76 percent of businesses surveyed for the latest Salary Guide plan to hire this year, up from 66 percent last year. The labour force participation rate jumped simultaneously (more people decided to look), which inflated the unemployment figure even as jobs grew. Westpac chief economist Kelly Eckhold called it "early evidence of economic strength" and believes the rate has peaked. For candidates, hiring intent is real; the bottleneck is the screening layer. The play is volume with precision: apply where your keyword match is genuine, customize each application, track which formulations clear the AI gate.

The Kiwi market rewards candidates who understand the two-tier filter. Beat the algorithm with structure and specificity. Win the human with the authenticity the algorithm can't fake.

Where the Frontier-Tech Scramble Stands

Kiwi's hiring sprint lands inside an environment that has shifted more in the past eighteen months than in the prior decade. California now hosts 33 of the world's top 50 private AI companies, and the state's new AI‑Unemployment Tracker (updated monthly since June 2026) shows early displacement signals concentrated in the Bay Area, tech-heavy sectors, and college‑educated workers with high AI exposure. The tracker has not found large‑scale layoffs, but it has documented a sustained rise in unemployment‑insurance claims from those same high‑exposure occupations since ChatGPT‑3.5's release. That pattern mirrors what recruiters across frontier tech describe: demand for specialists who can move models from experiment to production, not just prototype them.

The numbers bear out the pressure. Gartner reports that only one in 50 AI initiatives delivers transformative value, and fewer than a third of generative‑AI experiments have reached production. Yet 70 percent of organizations actively explore or implement LLM use cases, and two‑thirds say they are increasing investment because they have already seen strong value. The result is a scramble for people who can bridge data‑lifecycle management (75 percent of surveyed organizations have increased spending there) and the practical engineering of smaller, customized open‑source models, which over 75 percent of firms now prefer over monolithic closed alternatives.

Remote work has rewired the talent pool. The share of paid days worked remotely jumped from 7.2 percent in 2019 to 27.7 percent in 2024, and the proportion of workers with zero commute doubled to roughly 15 percent. For a company like Kiwi, candidates no longer cluster near headquarters; they expect distributed‑first processes, asynchronous evaluation, and compensation bands reflecting a national — not local — market. Board data puts Stripe's median at $250,000 and shows the intensity:

At the same time, the screening arms race has turned adversarial. Gartner predicts a quarter of candidate profiles could be fake by 2028, driven by LLM‑generated résumés, deepfake video interviews, and synthetic code samples. Phishing messages crafted by large language models now achieve a 54 percent click‑through rate, on par with human‑expert social engineering. Frontier‑tech recruiters report spending more cycles verifying authenticity than evaluating skill. Kiwi's screening process functions as a fraud filter as much as a competence test.

The broader signal is clear: companies treating hiring as a one‑off transaction are losing. Gartner warns of skills gaps, morale erosion, and the cost of rehiring at premium rates when early hires burn out. The firms gaining ground — ASML, Stripe, and now Kiwi — are building repeatable pipelines: structured assessments, transparent rubrics, feedback loops that turn rejected candidates into future referrals. In a market where 62 percent of workers say they would switch for better pay and stability, the screening process itself has become a retention tool. Kiwi's current openings are not just four seats to fill; they are the next calibration of a hiring engine the rest of the sector is watching.

Why Competitors and Investors Should Watch

The hiring surge at Kiwi arrives against a backdrop already rewriting recruitment. Employment Hero's system now ranks candidates, builds shortlists, and conducts first-round interviews — functions that used to consume entire recruiting teams. Webster's "human in the loop" rule means the AI recommends but never decides, yet the time savings are real. Companies still relying on manual resume review will lose speed, and in frontier tech speed compounds. A competitor taking three weeks to schedule a technical screen while Kiwi's pipeline moves candidates through an AI filter in days will see its offer-acceptance rate erode. The ripple is not just operational; it is cultural. Candidates now expect an AI touchpoint early. Those encountering a purely human front end may interpret it as a lagging indicator of the company's own tech adoption.

Investors read a different layer. A hiring spree backed by AI screening suggests capital efficiency and scalability. If Kiwi can process thousands of applicants without linearly scaling recruiting headcount, the marginal cost of each additional hire drops. That matters for runway models. It also signals the company is building for volume (whether product demand, data ingestion, or model training throughput). First-party board data reinforces the pattern: ASML and Stripe hire aggressively across the same frontier‑tech stack. When multiple players expand simultaneously, the talent pool tightens further, pushing compensation up and making AI screening not a luxury but a necessity.

For investors, the takeaway is not about one company's headcount. It is about the infrastructure layer underneath. The platforms powering this screening (Employment Hero and its peers) are becoming the de facto gatekeepers of talent flow. Ownership or exclusive access to that layer may prove more valuable than any single hire.

The Kicker

Webster's team at Employment Hero ran 2,500 AI interviews in a single April. The next month, they'll run more. Kiwi's four AI Security roles will close, then four more will open. The screen doesn't blink.

Working in frontier tech? Zero G Talent tracks the openings: see every open ASML role, browse frontier tech jobs, openings at Stripe, and the people building the field.