Evidence-based talent discovery: finding people through demonstrated work
The short answer
Evidence-based talent discovery starts with the work a person has chosen to publish, not an inferred score or a scraped resume. A useful result explains why someone matches, points to the evidence behind the match, states what remains unknown and preserves the person’s control over any next step.
Why keyword search falls short
Titles and skill lists are convenient filters, but they flatten context. Two people can both list “AI product” while one shipped a customer-facing workflow and the other completed a course.
A better discovery workflow
- Describe the work. Start with the outcomes, environment and constraints rather than a title.
- Search approved public identities. Include only people who opted into discovery.
- Return evidence beside each match. Show the specific project, role or public record that supports the result.
- Keep gaps visible. If availability, location, rate or a capability is not stated, say so.
- Ask before connecting. Discovery is not permission for outreach or data export.
What a useful result contains
Permission before introduction
A public profile can be discoverable without becoming a lead list. The professional chooses whether the profile appears in search. A company sees only approved public fields. Contact or introduction happens through a separate consent step.
- Public profile and search discoverability are separate choices
- No private source or hidden field is returned
- Every query is attributable to an authorised company user
- Results are rate-limited and cannot be bulk-exported
- The professional can disable future discovery
How to run a controlled pilot
Begin with a small number of approved companies and a narrow set of read-only queries. Review false matches, missing provenance and attempted enumeration before expanding access.
| Pilot control | Minimum requirement |
|---|---|
| Access | OAuth with an explicit talent-discovery scope |
| People | Public, discoverable profiles only |
| Results | Small, explained result sets |
| Actions | Read-only; no automated outreach |
| Audit | Company, actor, query, scope and outcome recorded |
Questions
Is this candidate verification?
No. Evidence explains support for specific claims; it does not certify a person or collapse them into a score.
Can an AI agent contact people automatically?
Not in a controlled pilot. Discovery is read-only, and an introduction requires a separate permissioned workflow.
What if information is missing?
The result should say that it is not stated instead of guessing.
Can professionals opt out?
Yes. Search discoverability must be a separate, reversible choice.
See it on a real profile.
Every claim on the Kleoverse example profile shows where it comes from, including the gaps.