Their systems were never built to answer. Volumes multiplied, AI polished every CV, and tools built for a slower market can no longer tell candidates apart, while hiring has never carried more strategic weight.
The job market needs new infrastructure, built on signals that can be trusted and verified: skills and capabilities.
See the productHow it worksThe intent is real. The evidence is missing. Nothing in the market produces the record that would make a skills-based decision provable, so hiring falls back on the CV, the school, the name.
Two people describe the same ability in ten ways; a job advert in ten more. 13,900 standardised skills give them one set of terms, and the ontology maps onto whichever reference the client already uses: ESCO in Europe, a national framework in the Gulf, an employer's own grid. Without that mapping, nothing above can be compared.
Every level is earned, never declared. Requirements are read twice, by independent passes, before they count. Levels come only from supervised work: a live machine-led interview, a timed test, code that ran. An unsupervised test proves nothing, so every assessment runs with anti-fraud monitoring built in.
With a shared definition of what makes a great fit, and scores based on real, anti-cheat assessments, recruiters make the right data-backed hiring decisions. And they have the record to prove it.
Ninety days after each hire, we learn whether the decision held, and the system adjusts what it weighs. The more it is used, the better it gets.
AI does the reading, the interviewing and the grading. The ranking is not AI, it follows rules a person can read, which is why every decision can be opened and explained.
Requirements are read across eight sources and mapped to the ontology, with a second pass checking the first. Candidates are matched against the verified pool and the gaps are named.
A conversation in natural language, not a form: a real-time voice interview, timed tests, code, anti-fraud. Levels come out earned.
Explainable weights, a publication threshold, a full log. The recruiter receives a pipeline they can work: prioritised, data-backed, manageable.
Every applicant is assessed the same way, at any number, and the shortlist arrives ready to work.
To the ministry, to the board, to the candidate who was turned down: criteria, levels and evidence attached to the decision.
What happened to the last hires corrects what the engine weighs on the next role.
Baselines are set with each client on the first cohort, and we report against them every cohort after that. Market evidence points the same way: skills-based search is linked to +12% quality of hire, and skills-based hiring to +10 points of retention.
Headcount doubling, roles opening faster than anyone can screen. Volume arrives before process does, and quality is what gives way first.
Client-facing, regulated, technical, safety-critical. One wrong hire is not a line in a spreadsheet: it is a lost account, an incident, a team that stalls.
National employment quotas turn a hiring shortfall into a penalty, and every decision into something that may have to be explained.
On the region's largest job site, 17.4 million applications chased 65,900 open roles last year. Agency fees already take 20 to 33% of a first-year salary. That is the budget we come to capture. Abu Dhabi is where those employers are.
We ran our own hiring on one of these platforms and stopped using it. That failure is the specification we built against.
A Gulf agency charges the employer 20 to 33% of a first annual salary, per hire. Arkemys is flat, per team, whatever the salary.
Between us we have stood on both sides of this transaction: the employer who cannot verify a claim, and the candidate no one will believe. Arkemys is the thing we each needed and could not buy.