TeamBotics · Toronto · An audit of the hiring front door

Operation Front Door

Every deployment tested live, from the candidate's chair, with the same eight questions.

Evidence vault

Everything behind the findings: the full method, the per-phase tallies, all 64 session records — transcripts verbatim, preserved separately from interpretation — and the raw scoring data. Anyone can re-run the protocol and check our work.

Method

Each deployment was tested live on November 6, 2026 (Phases 1–4) and November 7, 2026 (Phase 5) from the candidate's chair, exactly as a job-seeker meets it: one genuine live posting per company, the same eight-question protocol (A0–G) derived from the McDonald's session — what the role is, day-to-day duties, skills, the same in simpler terms, a request for a regular form, whether the assistant answers questions or only collects information, and a fallback probe. When the assistant demanded personal or application information, the tester declined in-chat and kept asking. No applications were submitted, no accounts created, no personal data supplied. Scores: PASS / PARTIAL / FAIL per question, with a standing convention that a capability claim contradicted by the same session's performance scores PARTIAL, not PASS.

Per-phase tallies (confirmed Paradox only)

PhaseDeploymentsScoredPASSPARTIALFAIL
1107201458
22160610
32161015
4118701869
57562450
Total32247342202

Raw scoring data (CSV): Phase 1 · Phase 2 · Phase 3 · Phase 4 · Phase 5 (plus the Phase 5 vendor-unconfirmed file). The documented-customer list behind the 137 figure is here.

Scored — confirmed Paradox deployments (32)

No assistant on the live path (25)

These employers were visited and no candidate-facing chat assistant was found on the live careers path. Negative findings, recorded as found.

Blocked / unreachable (3)

The careers destination could not be reached or errored; recorded as blocked, not scored.

Vendor-unconfirmed — filed separately, never charged to Paradox

Vendor-unconfirmed assistants (4)

No Paradox attribution was visible in these sessions, so their results are not Paradox findings. They are published as controls and comparisons: an assistant whose vendor could not be confirmed passed the protocol 7 of 7 — which is precisely why the fence matters. Phase 5 added Advantage Solutions' "Andi" (0 of 8): its one canned answer is character-identical to a deflection sentence documented inside two confirmed Paradox deployments in the same phase — recorded there as a likelihood-raising observation, not as attribution.