TalentInitiator, the demo walkthrough: transcript

TalentInitiator demo walkthrough. Demo data: fictional people and companies. The recording has no sound; these captions describe what is on screen.
Length: 4 min 58 s

Step 1
[0:00] Dana, a recruiter at Northline Staffing (Demo), starts on Today: what waits on a person.
[0:06] The workspace in French, then back. The demo-data banner stays on every screen.

Step 2
[0:12] A new job for Maple Grove Care Homes (Demo): Dana pastes the client’s email.
[0:17] The fields are read from the email. Sites in Ottawa and Gatineau bring in both provinces’ rules.
[0:25] The French requirement needs its reason, in French and in English.

Step 3
[0:29] The compliance panel: a $45,000 to $105,000 range blocks publishing.
[0:35] Ontario allows a $50,000 spread at most. $44,000 to $56,000 passes.
[0:40] “Canadian experience preferred” is flagged and rewritten; the required statements pass.

Step 4
[0:48] The rubric guard keeps out “Recent graduate”, which stands in for age.
[0:53] Approved, with Dana’s name, the time, the model and the prompt hash.
[0:58] Back on the posting, “Recent graduate” is flagged as an age proxy. Dana takes the rewrite in both languages.

Step 5
[1:06] Version 1 is on Northline’s careers page.
[1:10] Each job board shows its real state: Indeed sends a simulated receipt for now.
[1:15] Job Bank has no posting API, so the product prepares an export tagged with the NOC code.
[1:21] LinkedIn and Jobillico are planned, with no partner application filed yet. Nothing is scraped.
[1:27] The French posting on a phone: the pay range, the required statements, the reason for French.

Step 6
[1:34] Maple Grove’s PSW job has 48 fictional applicants. Dana imports five sample résumés.
[1:39] Each résumé is read, redacted and scored as you watch.
[1:42] One matches an applicant on file. A person decides whether to merge them.

Step 7
[1:47] The ranking: fit against the approved rubric, with references instead of names.
[1:52] Each score quotes the résumé lines it rests on, word for word.
[1:56] “Hidden from AI” lists every field removed before scoring.
[2:01] The top five side by side, names hidden.
[2:04] Interview priority suggests who to invite first. Dana chooses three.

Step 8
[2:11] Outreach to 15 applicants. Those who unsubscribed or withdrew consent are blocked, with the reason.
[2:18] Each draft is in the applicant’s language, with its CASL basis.
[2:22] Approved, then delivered to the demo outbox with the consent record.

Step 9
[2:28] Marie-Ève, on her phone in French: the privacy notice names AI evaluation in the United States.
[2:34] The screening chat starts by saying how AI is used and that a person decides.
[2:41] An off-script pay question gets the posted rate, from approved facts only.
[2:46] “Parler à une personne” (talk to a person) is there at every turn.

Step 10
[2:51] Her interview: storage in Canada, AI evaluation in the US, no face or voice analysis.
[2:58] She asks for extra time, and it is granted at once.
[3:02] Four answers by voice, one written. Each question allows one re-record.
[3:10] She checks the transcript before she submits.
[3:13] Submitted. The 45-day outcome clock starts now.

Step 11
[3:17] Dana opens the interview: a timestamped transcript, scored criterion by criterion.
[3:22] Every score cites her words, and the evaluation states what it cannot tell.
[3:28] Dana records her own decision: advance, meets certification and shift availability.
[3:33] Calendar sync is planned for a later release, so Dana proposes three times by hand. Marie-Ève gets them by email, in French.
[3:41] From the link in the email, she picks one on her phone and gets a calendar file.

Step 12
[3:46] On the pipeline, Dana moves C-00032 to “Not selected”.
[3:51] A decline is a decision: the drawer asks for a reason code and her own words.
[3:56] Pasting the AI’s text as the reason is refused. The reason has to be hers.
[4:01] With her own reason, the decline is recorded and the 45-day notice is queued.
[4:07] The bilingual outcome notice goes out by email.
[4:10] In her portal, “Pourquoi cette décision” (why this decision) explains it; she asks for a human review.
[4:16] The request lands in Priya’s human-review queue.

Step 13
[4:20] Dana submits two candidates to Maple Grove, with her reason for sharing them.
[4:26] The client sees references and evidence, never names or contact details.
[4:31] Sophie, at Maple Grove, opens her private link, reads the evidence and answers: interested.
[4:36] The compliance centre builds the job’s export pack.
[4:41] Priya, Northline’s owner, runs the adverse-impact report on the synthetic PSW cohort.
[4:46] One impact ratio is under 0.80, and groups under five are not compared.
[4:51] The audit log exports as CSV. Billing shows the Agency per-seat price times 5 seats.