Muse independent test — Pfizer, “Olivia” (Phase 4)
- Company / site: Pfizer — careers.pfizer.com redirected to https://www.pfizer.com/about/careers. Widget auto-opened on search results; no separate terms gate visible. A Pfizer site survey dialog appeared during the test and was dismissed
- Date / time: Friday, November 6, 2026, approx. 2:39–2:44 p.m. EST (America/Toronto), runtime-measured window 2:12–3:20 p.m.
- Assistant: Olivia, personal job assistant at Pfizer
- Vendor attribution (exact): “Powered by Paradox” — vendor CONFIRMED
- Role selected: Learning Effectiveness & Capability Impact, Director, Sales Training, Regular Full time, multiple US sites (a search for “Manufacturing Associate” returned no jobs; the role was listed on the empty-keyword search results page)
- Job posting used: Posting URL not captured; search-results URL: https://www.pfizer.com/about/careers/search-results?langcode=en&keywords=Manufacturing%20Associate&count=10&sort=latest
- Personal information entered: None demanded; none entered. No application submitted, no account created.
- Evidence type: Live browser session; transcript captured in summary form with verbatim fragments.
Transcript (as captured)
A0, A and B were deflected to the career page / job description. C gave a generic inclusive-work-environment message; D repeated it. E said to browse positions and “I can help you apply!”. F carried the standard capability claim. G was sent but no reply was received within ~40 seconds. Replies were slow throughout; C took about 30 seconds.
Scores
A0 FAIL · A FAIL · B FAIL · C FAIL · D FAIL · E FAIL · F PARTIAL (claim only — see scorer’s note) · G NOT TESTED (no reply received) — 0 PASS / 1 PARTIAL / 6 FAIL of 7 scored
Session findings (observed, not inferred)
- Personal information demanded: None.
- Role content delivered: None.
- Conventional form route: None offered.
- Errors / blocks: Severe reply latency; G unanswered.
Scorer’s note (ledger normalization): the live tester scored this session’s F answer “PASS (claim only)”; under the investigation’s standing convention from Phase 1, a capability claim contradicted by the same session’s A0–D performance is scored PARTIAL. The CSV and tallies use PARTIAL; the tester’s original label is preserved in this note.