Recruitment Research with Evidence Attached

The starting point

Put accumulated recruiting knowledge to work

A new search often begins with knowledge collected during earlier work: a promising conversation, an older resume, a previous finalist, or a useful detail in a recruiter note. Those records have value, but answering a new question requires finding them, understanding their context, and deciding whether the evidence still applies.

We built a platform that connects conversational research with a structured longlist workflow. Recruiters can ask questions across existing records, continue with follow-up questions, and bring documents into the conversation. For a defined search, they can move from a role brief to candidate assessments with the supporting excerpts attached.

Research design

Make the search plan something a recruiter can change

The longlist workflow turns a role scope and recruiter brief into explicit criteria. It distinguishes hard requirements, preferences, and contextual information, then identifies comparable earlier searches. Controls set the requested list size and whether research can include people one level below the target role.

Optional plan review pauses the workflow before deeper screening and reading. A recruiter can inspect and edit the criteria and comparable searches, then approve the next stage. This creates a practical intervention point: an assumption about the role can be corrected before it shapes the candidate research. Plan review is optional.

Discovery and assessment

Find relevant records, then read what they actually support

Discovery combines structured relationships and indexed keyword retrieval across earlier searches, work history, recruiter notes, and resumes. AI screening helps organize the pool for deeper reading. The resulting assessment connects each candidate to the criteria and records why that person surfaced.

Retrieval is the beginning of the assessment. A note may discuss several people. An employer named in a resume may be a client rather than a past employer. An outreach message may describe the job being offered. The workflow therefore checks excerpts and their context before using them to support candidate findings.

Evidence and uncertainty

Keep the reason for a recommendation within reach

Candidate assessments carry source excerpts, references, and criterion-level verdicts. Those verdicts distinguish supported requirements, partial evidence, and requirements that were not evidenced in the records. A missing detail remains a question for further research rather than an automatic statement that the person lacks the qualification.

The longlist includes coverage information, such as the channels used, evidence read, and unread material. When fewer supported candidates are returned than requested, a shortfall explanation accompanies the result. This helps a recruiter judge the scope of the research and decide whether to broaden the brief, inspect more evidence, or follow up directly.

Human review

Keep the recommendation and the decision distinct

A generated longlist is a research report for review. Recruiters can record an accepted, rejected, or deferred decision with a reason, separately from the generated assessment. That separation preserves the difference between what the system proposed and what the person responsible for the search chose.