HR-Focused AI "MENDAN" Updates "Matching Agent" Service, Continuously Improving Accuracy via Accumulated Placement Data
Zen office Co., Ltd. launched the "Matching Agent" for its HR-specialized voice AI "MENDAN," which enhances matching accuracy by combining voice analysis with CRM placement data. The flywheel model improves performance over time, achieving results such as a 4% increase in interview-to-placement rates.
📋 Article Processing Timeline
- 📰 Published: May 26, 2026 at 19:10
- 🔍 Collected: May 26, 2026 at 10:31
- 🤖 AI Analyzed: May 28, 2026 at 12:46 (50h 14m after Collected)
Zen office Co., Ltd. (Headquarters: Osaka; CEO: Yasuhiro Iwase) has launched the "Matching Agent," a new service update for its HR-specialized voice x AI agent, "MENDAN." This feature combines data extracted from interview audio with job requirements to propose openings with a high probability of receiving offers.
This function is characterized by a flywheel model where matching accuracy improves as it is used, by continuously learning from structured candidate information obtained from interview audio and historical placement data stored in CRMs. It shifts the focus from keyword-based searches to optimizing proposal sequences based on offer rates, enabling AI to replicate the once-personalized "winning combinations."
■ Improving Accuracy through Accumulated Placement Data
In the recruitment industry, data volume in CRMs is increasing annually. However, front-line proposal work often remains focused on basic condition searches like "salary, location, and job category." The Matching Agent addresses these issues: inconsistency in application rates, lack of non-public information like client preferences, and the reliance on individual veteran expertise.
■ Features of "Matching Agent"
1. Automated Information Structuring from Audio: AI extracts data such as career history, motivations, preferences, and competencies from interview audio. It also structures Must/Want requirements and past hiring trends for jobs, making the matching logic transparent.
2. Continuous Improvement via CRM Data: Automatically integrates with CRMs like PORTERS, Salesforce, and kintone to accumulate data from every stage—proposal, application, interview, and offer. Accuracy improves significantly after 3 to 6 months of use.
3. Optimizing Priority with Case Ranking: Ranks job openings based on unit price, offer rates, and urgency, allowing junior staff to make decisions at a veteran level.
4. Automatic CRM Synchronization: Supports customizable field mapping for major CRMs, ensuring seamless updates to resumes and activity logs.
■ Implementation Results
- Interview-to-Placement Rate: +4% improvement.
- Monthly Job Acquisitions: +47 additional jobs.
- Reduction in Missed Proposals: Every job in the inventory is reliably considered for proposals.
- Faster Onboarding for New Staff: Standardizing decision logic across the organization.
■ Implementation Flow
The process involves hearing/analysis, field design, data linkage, and operational launch followed by verification of improvement in offer rates.
This function is characterized by a flywheel model where matching accuracy improves as it is used, by continuously learning from structured candidate information obtained from interview audio and historical placement data stored in CRMs. It shifts the focus from keyword-based searches to optimizing proposal sequences based on offer rates, enabling AI to replicate the once-personalized "winning combinations."
■ Improving Accuracy through Accumulated Placement Data
In the recruitment industry, data volume in CRMs is increasing annually. However, front-line proposal work often remains focused on basic condition searches like "salary, location, and job category." The Matching Agent addresses these issues: inconsistency in application rates, lack of non-public information like client preferences, and the reliance on individual veteran expertise.
■ Features of "Matching Agent"
1. Automated Information Structuring from Audio: AI extracts data such as career history, motivations, preferences, and competencies from interview audio. It also structures Must/Want requirements and past hiring trends for jobs, making the matching logic transparent.
2. Continuous Improvement via CRM Data: Automatically integrates with CRMs like PORTERS, Salesforce, and kintone to accumulate data from every stage—proposal, application, interview, and offer. Accuracy improves significantly after 3 to 6 months of use.
3. Optimizing Priority with Case Ranking: Ranks job openings based on unit price, offer rates, and urgency, allowing junior staff to make decisions at a veteran level.
4. Automatic CRM Synchronization: Supports customizable field mapping for major CRMs, ensuring seamless updates to resumes and activity logs.
■ Implementation Results
- Interview-to-Placement Rate: +4% improvement.
- Monthly Job Acquisitions: +47 additional jobs.
- Reduction in Missed Proposals: Every job in the inventory is reliably considered for proposals.
- Faster Onboarding for New Staff: Standardizing decision logic across the organization.
■ Implementation Flow
The process involves hearing/analysis, field design, data linkage, and operational launch followed by verification of improvement in offer rates.
FAQ
「マッチングエージェント」の主な機能は何ですか？
面談音声から求職者情報を構造化し、CRMの決定データと連携して内定可能性の高い求人を自動提案する機能です。
どのようなCRMと連携が可能ですか？
PORTERS、Salesforce、kintoneなどの主要なCRMに対応しており、各社の運用に合わせた項目マッピングも可能です。
マッチング精度はどのように向上しますか？
提案結果、応募、面接通過、内定の各段階のデータを継続的に学習する「フライホイール構造」により、使うほど精度が向上します。
導入によって期待できる具体的な効果は何ですか？
導入企業の実績では、面接から決定までの通過率が4%向上し、月間の求人獲得数が47件増加しています。
新人アドバイザーの育成にどう役立ちますか？
案件がランク付け表示されるため、新人でもベテランと同水準の優先順位判断が可能になり、立ち上がり期間を短縮できます。