Maki describes itself as "the intelligence layer" that sits above existing ATS infrastructure. The positioning is deliberate: it is not a replacement system, it is an augmentation layer. Enterprises keep Workday, iCIMS, Greenhouse, or SAP SuccessFactors. Maki plugs into them, handles the candidate-facing intelligence, and pushes structured output back. The mechanism is an agent stack – three public-facing agents operating at different funnel stages, two operational agents handling scheduling and internal mobility, and a platform layer that orchestrates all of them.

Vision and product philosophy

Legardez-Coquin's core observation is simple and structural: in human resources, the majority of meaningful signals are produced in conversation – phone screens, interviews, manager reviews – and then immediately lost. "Everything disappears after my application process to that company, or sometimes after my five years as a sales executive enterprise in that company," he told Josh Bersin. The opportunity, as he frames it, is to capture that signal, structure it, and compound it into better hiring decisions over time.

That compounding is the thesis. An organization that conducted one million interviews last year, running on Maki, should make better decisions in the next million – because the system has learned which signals predicted performance and which didn't. The assessment engine does not just score candidates. It trains on outcomes.

The longer-term ambition is explicit. Talent acquisition is Phase 1. Phase 2 is dynamic talent management: capturing signals from within the workforce – the 40 hours a week every employee produces on the company's systems – and making those signals actionable. Legardez-Coquin's stated vision: an HR leader who can query their workforce intelligence layer and ask "who should I promote for that new department?" and receive a ranked, evidence-based answer. Not a report. A recommendation.

The architectural framing he uses is "system of work" rather than system of record. The ATS stores static data. Maki generates dynamic intelligence. In five to ten years, he argues, the structured signal Maki produces will flow directly to enterprise data warehouses – making the ATS as an intermediate layer redundant. Josh Bersin's summary from the same conversation: "You're building business accelerators here, not just HR employee improvement systems."

On the question of what distinguishes an agent from workflow automation, Legardez-Coquin is precise: "It's how we can build a system where an intelligence can take an action decision in a system with constraints and with ability to reasoning, but also to improve. The big difference between workflow automation and agents is the ability to take action and to train and improve over time." That definition is the product strategy in one sentence.

On recruiter displacement, the company's position is deliberate. "We don't believe recruiters will be replaced. We believe their life will get better and their task will change." The mission statement Maki uses internally: give human resources more than human power. Recruiters who previously spent Monday mornings under pressure from hundreds of unprocessed CVs now arrive to a ranked shortlist. The work that remains is judgment, not administration.

Science

The scientific engine predates the AI agent architecture. Maki employed psychometricians from Cambridge University at founding, building validated models across cognitive, behavioral, technical, and language domains before deploying conversational AI on top. The company measures 300+ skills. Every score is traceable to the evidence behind it. Independent external auditing runs continuously. The explicit claim is that assessments predict job performance – not that they correlate to interview preference.

Bias mitigation is structural, not cosmetic. Fairness controls and structured evaluation methods are embedded at the model level. This matters to enterprise procurement: EU AI Act readiness, explainable scoring, and auditable decisions are table-stakes requirements for large financial, consulting, and retail clients operating across jurisdictions.

Integrations

Maki integrates natively with Greenhouse, iCIMS, SAP SuccessFactors, Eightfold, Cornerstone, and Workday. The integration surface creates switching friction: once a client's ATS is flowing candidate data through Maki's agents, extracting and replacing the intelligence layer requires deliberate effort. That is the structural advantage of the "integration-first, replace-nothing" motion.

Trust and security

EU data residency by default. Advanced encryption at rest and in transit. Role-based admin controls with detailed audit logs. A formal Data Processing Agreement. Bias-audited AI with explainable scoring. The compliance posture targets regulated industries first – banking, insurance, audit – where procurement requires it.

Agents

Shiro – skill screening

Shiro operates at the top of the hiring funnel. It evaluates candidates across 400+ skills – cognitive, behavioral, technical, language – through short, scientifically validated assessments. Custom knock-out logic lets clients eliminate unqualified applicants at volume without recruiter involvement. The pitch is simple: screen every applicant, regardless of CV, in minutes.

Mochi – conversational screening

Mochi conducts adaptive voice and web-based screening interviews, 24 hours a day, in 40+ languages. It is not a chatbot. It clones recruiter voices, absorbs employer branding, and conducts conversations that candidates describe as indistinguishable from human screening calls. The March 2025 launch of AI-Powered Phone Conversations, subsequently deepened via a Deepgram partnership for real-time voice AI, pushed Mochi into high-volume frontline hiring at global scale. At H&M, across 60 markets, Mochi automated 80–90% of the entire screening process.

Ken – in-depth assessment

Ken operates late-funnel. It delivers structured, science-grade evaluations for high-stakes roles – coding challenges, problem-solving simulations, leadership assessments – with traceable scoring. Every decision is explainable and defensible. Ken targets corporate, technical, and white-collar hiring where mis-hire cost is highest and where replacing subjective interviews with objective structured data has the most leverage.