Building an AI Hiring Companion For Work.Life

What they built

Recruit.life - a read-only AI hiring companion that mirrors a company's ATS (Team Tailor) and acts as an intelligent analysis layer throughout the recruitment process.

Key capabilities:

• Auto-syncs candidates from Team Tailor via API (new applicants, stage moves, deletions/rejections all reflected in real time)
• Anonymises candidates and scores them 1–5 against the live job description, with evidence-led rationale
• Surfaces primary strengths, risk factors, competitive edge, and gaps per candidate
• Suggests tailored screening questions based on identified CV gaps
• Accepts uploaded interview transcripts or pasted notes (from tools like Granola) to update and deepen candidate assessments at each stage
• Generates detailed, personalised rejection emails based on accumulated evidence - no blank templates, no unfilled brackets
• Includes a DPO statement explaining data handling and candidate rights
• Does not store candidate data - Team Tailor remains the system of record

Why they built it

• Knowledge transfer problem: Ten years of hiring intuition lived in one person's head. If she went on holiday, the process quality dropped. The tool encodes her assessment logic so any team member can run recruitment compliantly and consistently.
• Data compliance concern: Using ChatGPT to analyse CVs felt risky - even without training on the data, it felt inappropriate to populate external LLMs with personal candidate information. A closed, internal tool removes that risk.
• Efficiency: Restarting ChatGPT conversations, re-uploading job descriptions, and re-uploading CVs for each candidate was slow. This tool automates all of that setup.

Tools used

• Lovable - used to build the front-end application
• Team Tailor - ATS; connected via its API as the data source and system of record
• Granola - used to capture transcripts from in-person interviews, which are then uploaded into the tool
• Team Tailor's native app - used to record transcripts for virtual/phone screens

How you could build this yourself

1. Audit your current hiring process - document the logic you use to assess candidates at each stage (what makes a 1 vs. a 5, what screening questions you always ask, what a good rejection email looks like). This becomes your AI's instruction set.
2. Check your ATS for API access - confirm whether your ATS (e.g. Team Tailor, Greenhouse, Lever) has API documentation you can access. This is what enables the live sync.
3. Build in Lovable - open Lovable and describe what you want: a read-only candidate dashboard that pulls from your ATS via API, scores candidates against a job description, and accepts uploaded interview notes. Paste in your API documentation to help it connect correctly.
4. Define your scoring rubric inside the tool - explicitly tell the AI what each score level means, what risk factors to flag, and what a strong rejection email should include. The more specific your instructions, the more consistent the output.
5. Set up your transcript workflow - decide how you'll capture interview notes (e.g. Granola for in-person, ATS app for virtual) and establish a habit of uploading transcripts after each stage to keep assessments current.
6. Add a DPO/data statement - include a visible explanation of how candidate data is handled, who stores it, and what candidates' rights are. Get sign-off from your DPO or legal contact before going live.
7. Test with a live role - run one active vacancy through the tool end-to-end before rolling out to the wider team. Check that stage moves, deletions, and rejections sync correctly from your ATS.
8. Hand over with confidence - once the logic is embedded, any team member can pick up a live process without a briefing from you.

Source

Built by Olya Yakzhina, People Lead at Work.Life, and presented as part of Open Org's "Shipped by HR" session series - where People leaders demo AI tools they've built for their businesses.