Jack & Jill

Jack & Jill is worth evaluating for a direct-hiring team that wants an AI-assisted way to turn a role brief into an evolving candidate search and consent-based introductions. Its public documentation describes Jill, the employer-side agent, turning a description into a hiring brief, asking calibration questions, searching for candidates, and adjusting from feedback. It describes Jack, the candidate-side agent, as the agent Jill contacts when a team requests an introduction. [S1][S2]

This is not an outsourced hiring decision. The employer still needs to own job criteria, candidate review, outreach approval, interviews, selection, accommodations, lawful processing, and the records in its ATS. Jack & Jill's privacy policy says it uses automated decision-making for matchmaking and uses AI to assess alignment, generate fit scores, rankings, and profile summaries; it also says human platform users decide whether to engage candidates in its sourcing feature. Treat every score and recommendation as an input that needs a human check. [S3]

Best for: direct-hiring teams with an explicit brief, a recruiter or hiring manager who will review recommendations and introductions, and a defined process for candidate data, interviews, selection, and ATS records.

Not for: third-party recruiting agencies, a buyer seeking a fully automated employment decision, a confidential search without tested visibility and promotion settings, or a team without a clear owner for data rights, hiring fairness, and system-of-record changes.

Jack & Jill's documented employer workflow begins with a work-email signup and an employer role. A team can paste a job description or import roles from a careers page. Jill then researches context, drafts a structured hiring brief, asks calibration questions, and begins searching after the brief is set. The documentation says the initial setup to first candidates usually takes under 10 minutes; that is a vendor statement, not a time-to-hire promise. [S1]

1. Set the role boundary: the hiring manager supplies responsibilities, required skills, seniority, compensation range, reference profiles, and the candidate-facing pitch. Review the internal brief and separately choose team visibility, promotion, anonymisation, and who can change them. A private team role can still reach candidates when promotion is enabled. [S1]
2. Review candidate suggestions: Jill searches Jack's network and, according to the product docs, can also surface public profiles. A recruiter reviews the candidate, the stated reason for the match, gaps, availability context, and any existing pipeline record. Do not assume a fit score represents job performance or an employment decision. [S3]
3. Request an introduction: when the employer selects a person, Jill tells the candidate's Jack agent about the role and rationale. The candidate decides whether to proceed; only then does Jill email-introduce both sides. The vendor says no contact information is shared unless both sides opt in. The employer then runs its own interview process. [S2]
4. Sync deliberate pipeline events: for Ashby, Jack & Jill documents a one-way push that can create or update candidates and applications, move stages, deduplicate by email and LinkedIn profile, and map selected lifecycle events. The team chooses which events to sync. Test those choices before using an integration as an employment-record workflow. [S5]

Publicly documented facts: Jack & Jill positions Jill for companies hiring directly, rather than agencies, and says its network is strongest in London, San Francisco, and New York while introductions and hires can occur elsewhere. Its current pricing page says Jack is free for candidates and Jill charges companies 10% of first-year salary when it makes a hire; it also advertises a full refund if a hire leaves or is dismissed in the first three months. Verify the signed agreement, scope, exclusions, contractor terms, and refund conditions before relying on those public terms. [S4]

Candidate data and AI use: the privacy policy describes handling contact information, professional details, voice calls and transcripts, uploaded files, device and usage data, and public professional-profile information. It says candidate profiles may be shared with prospective employers under selected visibility settings and contact details only after the candidate and employer agree to an introduction. It also describes third-party AI and hosting providers and says retained data may be deleted on request except for limited legal or commercial retention, which it describes as up to two years. [S3]

Decision and fairness boundary: the policy says the service uses automated decision-making for matchmaking and says it audits decision outcomes for accuracy and bias. It does not, in the reviewed material, establish the buyer's exact audit method, sample sizes, role-specific error rates, adverse-impact analysis, configuration, appeal process, or compliance outcome. Request actual evidence for the jurisdictions and job families in scope before using a score or ranking in a consequential employment workflow. [S3]

Material unknowns: the reviewed sources do not establish a public enterprise security package, contractual availability commitment, model-provider and subprocessor list, buyer-specific data-processing agreement, integration permissions beyond the documented examples, exact candidate-network composition, expected response or hire rate, suitability for every country, or the contract terms that govern each hire. The Ashby guide confirms a one-directional push and configurable events, but each buyer still needs to test its own stage mapping, deduplication, access, audit, correction, and deletion requirements. [S5]

Start with one ordinary, non-confidential role where the team already has a written selection process. Name an accountable recruiter, hiring manager, privacy owner, and ATS administrator. Document the job criteria, what may be placed in the brief and pitch, who can access the role, which candidate-facing promotion settings are allowed, how recommendations are reviewed, and when an introduction can be requested.

Use a buyer-owned test set: a precise role; an ambiguous brief; a public-profile candidate; a candidate already in the ATS; a candidate who declines an introduction; a candidate who wants a correction or deletion; a confidential role; an accommodation request; a role change; a duplicate; a disconnection; and a manual continuation if the platform or integration is unavailable. For every case, retain the human rationale rather than assuming the AI's output is sufficient.

Measure setup effort, candidate-review time, number of introduction requests, candidate opt-in rate, duplicate rate, ATS reconciliation, time to a human first interview, candidate feedback, recruiter overrides, correction and deletion turnaround, and any discrepancy between the platform and system-of-record. Stop or narrow the pilot if a recommendation is used as an automatic employment decision, contact details move without the expected consent, a confidential role is promoted or exposed incorrectly, an integration writes an unreviewed record, a candidate-rights request cannot be fulfilled, or the team cannot explain and reconstruct a decision.