Qureos

Fifteen systems are named individually in the homepage integration strip, across three classes. Applicant tracking covers Greenhouse, Lever, Workday, Taleo and iCIMS, with 50+ more. Job boards cover LinkedIn, Indeed and Glassdoor alongside the regional Bayt and Naukri, with 10+ more. CRM and HR systems cover Salesforce, HubSpot, SAP, BambooHR and Zoho, with 40+ more. Bayt and Naukri alongside the western boards is the regional coverage a GCC buyer would look for.

The integrations act in both directions. In Qureos's words: "Push candidates directly into your ATS. Iris syncs status updates, notes, and stages automatically." Listings follow "post once, distribute everywhere," with Qureos syndicating them "across all major job networks instantly," and "candidate data, communications, and hiring stages flow into your CRM automatically."

Those are authenticated writes into systems Qureos does not own, and the vendor frames the posture explicitly: "your existing stack stays the system of record and Qureos operates as the AI layer on top of the tools your team already uses." Outbound action also reaches candidates over email, LinkedIn and WhatsApp and books phone and video interviews.

The counts do not agree. The homepage FAQ states "100+ ATS, HRIS, and custom integrations," while the strip's own arithmetic sums to roughly 115 across the three classes. Either way, the named systems and the documented two way sync are what matter.

Sourcewww.qureos.com homepage integration strip and FAQ, /integrationsread 2026-09-12

The sequence is published as four numbered steps, each a distinct act by a distinct participant. In step 1 the recruiter defines the role, from job title and must have skills to experience level and responsibilities, and "Iris uses this to source the right candidates and tailor how it screens and interviews them," so one input sets every later stage.

In step 2, "Iris promotes your job with sponsored listings on 2,000+ job boards and social media ads, searches 2M+ candidate profiles, and requests referrals, reaching both active job seekers and passive talent": three parallel sourcing paths from one brief. In step 3, "Iris reviews every application, conducts phone or video interviews with shortlisted candidates, and delivers detailed screening reports." In step 4 the recruiter takes over.

The span is long, from writing the job description and distributing it to sourcing, reviewing applications and interviewing by phone or video on its own. Scoring, reports and outreach across three channels with follow ups come next, and the pipeline is managed in Qureos's own ATS with two way sync outward.

The recruiter briefs and decides, Iris sources, screens and interviews, and the candidate is contacted and interviewed, while the incumbent ATS or CRM stays the system of record.

No orchestration runtime is named, no handoff between sub agents is described, and there is no authoring surface: the flow is Qureos's hiring process set by the brief rather than one a customer designs.

Sourcewww.qureos.com homepage and /iris-featuresread 2026-09-12

Qureos maintains its own candidate corpus and retrieves from it against criteria the customer sets. It holds a pool of "2M+ verified candidates," searches it on every request, and reaches beyond it: "Qureos AI searches across LinkedIn, job boards, and its own talent pool to surface high-fit candidates." The pool stands between runs and is described as verified, which is maintenance rather than accumulation.

Retrieval is driven by criteria and the mechanism is stated: "You write the criteria that matter for the role, in order of priority. Iris reads every candidate profile against them and returns a percentage match score. The highest-scoring candidates appear first." Every profile read against customer set, priority ordered criteria is retrieval with a ranking function, not keyword filtering, and the recruiter describes the role in plain English rather than building a query.

The customer's own data is in scope too. Qureos ships an ATS holding the customer's pipeline, syncs two way with 100 plus ATS, HRIS and CRM systems so the existing stack stays the system of record, and can reengage past applicants through outreach campaigns, which means prior applications persist and are read again. Grounding shows at the response level: every candidate assessment carries clear evidence behind the fit rating.

One figure is unreconciled. The LinkedIn Outreach feature page describes a talent pool "spanning over 1 billion professionals across 200 countries" while the homepage states 2M+ verified candidates. Those are probably different objects, LinkedIn's population against Qureos's own index, but no page says so, and a 500 fold gap changes what a reader would expect.

Sourcewww.qureos.com/iris-features and homepageread 2026-09-12

The hiring decision stays with the recruiter, but no hold on the agent's own actions is documented. The vendor states the division plainly: "Iris does the sourcing and vetting, you handle the final decision"; "you take over for final interviews and make the hiring decision"; and in the FAQ, "Qureos handles the repetitive, high-volume tasks like screening, scheduling, and follow-ups. Your recruiters keep judgment and decision-making." That is a division of labor, not a gate.

There is a real configuration surface that bounds the agent before it runs. The recruiter writes the criteria that matter in priority order, sets optimized search criteria, chooses public or private listings, and can review and adjust the filters Iris sets. Every candidate assessment carries clear evidence behind the fit rating, so the person deciding has the basis in front of them.

The agent's own consequential acts go unchecked, though. Iris contacts candidates automatically across email, LinkedIn and WhatsApp and conducts phone or video interviews with them, and nothing documents an approval queue, a checkpoint, a threshold or a sign off before a named individual is messaged or interviewed by the AI. Qureos publishes the human's ownership of the hiring decision, which comes after everything the agent does.

Sourcewww.qureos.com homepage, /for-recruiters and /iris-featuresread 2026-09-12

No security page, trust center, named certification or access control is published. The navigation runs from Product to Support and Login, the footer carries features, resources and directories along with Terms of use and a Privacy policy, and neither has a security, trust, compliance or certifications entry; no trust subdomain appears either.

The compliance claim is self asserted rather than attested. The recruiter FAQ states that "Iris is fully compliant with GDPR and local data protection laws" and that "candidate data is stored securely and processed according to the highest privacy standards." Regulatory alignment claimed by the vendor about itself is a posture; no SOC 2, ISO, HIPAA or CSA certificate is named as held, no auditor is named, and no report is offered on request.

Access controls are missing too, with no SSO or SCIM, no RBAC or permission model, and no admin console anywhere on the site, so a compliance claim has nothing to stand on.

The consequence matters given what the product does. Iris holds candidate data across the GCC and beyond, conducts recorded video interviews and filters on nationality, and a buyer taking it through a security review has no published certificate, control list or report to work from.

Sourcewww.qureos.com homepage, /for-recruiters FAQ, navigation and footerread 2026-09-12

Reporting covers the recruiting process rather than tracing the agent's runs. A real time dashboard tracks outreach progress and job board performance, with funnel insights and detailed screening reports for each candidate. The most useful item is that "every candidate assessment includes clear evidence behind the fit rating, not just a score," so a reviewer can see why Iris rated a candidate as it did. Talent Insights ships as its own product page for data driven hiring.

Nothing keeps a record of the agent's conduct that can be reconstructed. No run log, action history or audit trail is named, nor any export or retention period.

Nothing states that a record is kept of which candidates Iris screened out and on what criterion, which messages it sent over which channel and when, what it asked in a phone or video interview, or how a match percentage was computed.

On a product that interviews named individuals on its own and filters on nationality, an audit trail is what a buyer would need, and it is not published. Dashboards answer how the pipeline is doing; an audit trail answers what the agent did to whom.

Sourcewww.qureos.com homepage FAQ, /iris-features and /talent-insightsread 2026-09-12

What Qureos keeps is application data, not agent memory. Talent pools and pipelines are the customer's application data model, held in Qureos's own ATS and synced from the customer's incumbent ATS, HRIS and CRM. A recruiting database persists because it is a database; the pool is the corpus Iris reads every profile from. Reengaging past applicants through outreach campaigns is a query over that store, not agent state.

The learning Qureos names improves the system rather than giving the agent memory. Its Adaptive AI Engine is described as "AI that adapts and evolves, making your hiring process smarter and more efficient." Learning that improves a matching model is not a store a customer can point at.

Nothing else appears: no named memory layer or retained preference set the agent reapplies across roles, no decision traces carried between searches that can be inspected, no stated scope or lifetime, and no customer control to inspect, export or purge agent state separately from the candidate records. Criteria are written per role by the recruiter, which is input for each run rather than remembered context. The features page, the homepage FAQ and the knowledge base all describe what Iris does per job without describing anything it keeps.

Sourcewww.qureos.com/iris-features, homepage FAQ and /iris-knowledge-baseread 2026-09-12

No deployment topology is published. The homepage, features and recruiter pages and the knowledge base carry no region, residency, data location or hosting statement, and there is no security or trust page where one would sit. Terms of use and a privacy policy are the only governance documents published.

The geography that is published is market coverage, not residency: jobs directories for the six GCC countries (the UAE, Saudi Arabia, Qatar, Oman, Kuwait and Bahrain), labor law guides for 29 plus countries, and interfaces in Arabic and English among 20 plus languages, plus an FAQ answering "We hire in multiple countries. Does this work outside the GCC? Yes." Where a platform recruits is not where its data sits, and the GDPR compliance claim is a posture, not a location.

The buyer consequence is real: a product whose core market is the GCC, which asserts GDPR compliance, and which holds recorded video interviews and nationality data, publishes nothing about where any of it is processed. The privacy policy at www.qureos.com/privacy-policy is the document most likely to name a processing location or a subprocessor.

Sourcewww.qureos.com homepage, /iris-features and footerread 2026-09-12

Ready to use units ship, but only one agent. The AI Job Description Generator and Interview Question Generator are named tools in the application at app.qureos.com, usable without configuring anything. Personalized outreach sequences, screening flows and optimized search criteria are reusable configurations a recruiter sets once per role shape. The site also publishes browsable template libraries for job descriptions, HR emails, interview questions, company policies and resumes.

There is one agent. Iris is a single named recruiter that sources, screens, interviews and reaches out; the nine items on the features page are its capabilities, not a set of complete agents a buyer selects among. They are optimized search criteria, job description, the adaptive AI engine, the ATS, public versus private listings, job boards, AI video interview, personalized outreach and LinkedIn outreach. Turning off AI Video Interview removes a step, not an agent.

The template libraries are also weaker than they look. They sit on the marketing site alongside glossaries, labor law guides and salary calculators, which reads as content marketing rather than assets inside the product, and no page states that they load into a workflow.

Sourcewww.qureos.com/iris-features and site navigationread 2026-09-12

Events, cadences and several channels are documented in depth. Outbound channels are named and plural: Iris engages candidates automatically across email, LinkedIn and WhatsApp with personalized messaging and follow ups, and LinkedIn Outreach and Personalized Outreach each ship as their own named feature. WhatsApp is the one that matters commercially in this market.

Synchronous channels are covered too: Iris conducts phone or video interviews with shortlisted candidates, so voice and video are conversations the agent holds rather than notifications, with AI Video Interview as its own feature page.

Distribution is a channel set in itself: sponsored listings on "2,000+ job boards and social media ads," syndicated instantly, with LinkedIn, Indeed and Glassdoor among the boards named. Reach extends across 20 plus languages with English and Arabic interfaces.

Events drive the flow at named points. Posting a job triggers promotion, sourcing and referral requests together; an application arriving triggers review of every application; a shortlist decision triggers an interview invitation; and past applicants can be contacted again through outreach campaigns, a trigger off the customer's own historic data rather than a new role. Follow ups run on their own cadence after first contact.

The job board count reads 100+ on the features page and 2,000+ on the homepage and in the FAQ; both are the vendor's own.

Sourcewww.qureos.com/iris-features and homepageread 2026-09-12

Nothing on the site names a model provider. None of Qureos's product pages, knowledge base or FAQ names a model, a provider or a family; the vendor speaks of its AI engine, AI agents and an Adaptive AI Engine without identifying what runs underneath.

No selection surface exists either: there is no model picker, admin setting or assignment per agent, and no way to bring your own key, and the product is configured through recruiting criteria such as skills, nationality and salary range rather than anything touching the model layer.

Two things look like model choice and are not. The homepage invites a reader to "Ask ChatGPT," "Ask Claude" or "Ask Perplexity" about Qureos, with three buttons pointing at those assistants, and a page at qureos.com/ai-info is headed "Hey AI, learn about us." Those are marketing prompts aimed at third party chatbots and content written for assistants to read; neither lets a customer choose the model that screens their candidates. The Adaptive AI Engine, described as "AI that adapts and evolves," is a model improving on its own, not customer selection.

The customer is not told what is running, let alone offered an alternative.

Sourcewww.qureos.com/iris-features, homepage and /iris-knowledge-baseread 2026-09-12

No route for an outside system to call Qureos is published. An integrations page at qureos.com/integrations is promoted in the footer as "Integrate with Qureos," and the homepage call to action "Integrate your ATS" points at a signup: a customer connecting their stack, not a developer building against Qureos. The FAQ describes "100+ ATS, HRIS, and custom integrations," where custom integrations reads as bespoke work rather than a published contract, and connector breadth is Qureos reaching outward rather than others reaching in.

Neither the navigation nor the footer has a developer, docs or API entry, and no docs or dev subdomain appears; app.qureos.com and sourcing.qureos.com are the application and the job posting flow. No API reference, endpoint list or SDK exists anywhere on the site, and no webhook or MCP server either.

One surface looks aimed at AI and is not: a page at qureos.com/ai-info titled "Hey AI, learn about us," plus buttons inviting a reader to ask ChatGPT, Claude or Perplexity about Qureos. Those are content written for assistants to read and prompts pointed at third party chatbots, and nothing an outside agent can call to make Qureos act.

Sourcewww.qureos.com homepage, /iris-features, navigation and footerread 2026-09-12

Qureos publishes a readable, comparable result against a human ground truth. The homepage FAQ states: "The AI CV screening runs at 95%+ accuracy, validated against real recruiter decisions in production." That is a figure, a named validation method and a benchmark that is the very thing the agent is trying to replicate: a score with a stated basis rather than a marketing claim.

Evidence for each assessment accompanies it: "every candidate assessment includes clear evidence behind the fit rating, not just a score," so the output can be inspected item by item and not only in aggregate. An Adaptive AI Engine described as "AI that adapts and evolves" sits around both.

The limitations are real. No test environment, sandbox, holdout, versioning or regression check is documented; the accuracy figure is vendor reported with no methodology beyond the one clause; and there is no evaluation surface for customers, so a buyer reads Qureos's number rather than running their own.

No bias or fairness audit of the screening or matching is published. That is worth stating plainly because Qureos's own hiring guide names NYC Local Law 144 bias audits for automated employment decision tools as a criterion it evaluates vendors against.

Sourcewww.qureos.com homepage FAQ and /iris-featuresread 2026-09-12

LinkedIn is an interface Qureos does not own, and the product acts inside it without a person driving each step.

LinkedIn Outreach is a named product feature with its own page, listed under Outreach and Follow Up alongside Personalized Outreach, and Iris engages candidates automatically across email, LinkedIn and WhatsApp with personalized messaging and follow ups. Qureos also states that its AI searches across LinkedIn as part of sourcing.

Email and WhatsApp are messaging channels; on the LinkedIn leg the product both reads and writes inside someone else's interface, unattended, on a cadence the recruiter sets.

It covers one interface, and the mechanism is undocumented. No page states how the LinkedIn leg executes, whether through a sanctioned partner path, an authenticated session or an extension. No hosted or local browser, desktop session, computer control or RPA is claimed anywhere on the site.

That the LinkedIn actions happen is documented on a dedicated feature page, www.qureos.com/iris-features/linkedin-outreach; how they happen is not.

Sourcewww.qureos.com/iris-features and homepage FAQread 2026-09-12