Jack & Jill: evaluate the two-agent hiring marketplace

Air Street’s September 2026 investment account describes a $40 million Series A following a $20 million seed round. ^[1]

These are two financing events, not revenue or a valuation. The later announcement updates this case’s earlier seed-stage account. Our analysis should ask what operating evidence the business produces after raising capital, rather than treating investor participation as proof of hiring quality.

The company describes Jack as the candidate-side agent and Jill as the hiring-side agent, coordinating introductions. ^[2]

Our interpretation is that this design combines two kinds of information: what a person wants next and what a role requires. That coordination is the hypothesis to test. It does not by itself prove that either side is represented accurately, that every opportunity is available, or that automated recommendations outperform a comparable process.

Creandum’s October 2025 account reported that 49,000 candidates in the UK had spoken with Jack; it did not describe 49,000 successful hires. ^[3]

Keep the units visible. People, conversations, introductions, interviews, offers and job starts are different measures. A person may have several conversations or interviews. Counts from different dates or reporting populations cannot produce a meaningful conversion rate simply by dividing one by another. Unknown follow-up remains unknown.

Nor does a success fee automatically resolve every incentive conflict. Our analysis is that a marketplace must also test whose preferences are represented and whether both parties can correct an inaccurate description. This is a research question, not a finding of misconduct.

Analysis & application

To assess an agent-mediated service, design a consent-based pilot with a defined observation window.

1. Define the eligible roles and participating candidates. Use explicit job requirements and user-stated preferences; do not infer sensitive personal traits.
2. Let both sides review and correct the information used for matching, and control whether an introduction is made.
3. Track unique people and roles through introduction, interview, offer and accepted start. Record repeat events separately.
4. Compare time, effort and agreed job-relevant outcomes with a suitable baseline. Separate language or access barriers from job qualifications; provide a human review route.
5. Follow up at a stated interval with permission. Report missing responses, sample limitations and costs per outcome. Keep final employment decisions with the candidate and employer.

Analysis & application

The sources reviewed are company and investor accounts, not an independent outcome audit. They do not establish a causal improvement in hiring quality, retention or equitable access. We have not run the proposed pilot on Jack & Jill.

A useful next result would link a clearly defined cohort to verified outcomes, explain corrections and declined introductions, and compare like-for-like roles over the same period. A larger network may increase possible matches, but it is not interchangeable with a better match. Keep dated funding and activity reports as context, and test the quality of the actual exchange.