v2.51 to v2.63
The Assistant gets straight to the answer. Each answer opens with a line that sums up what the Assistant looked at, for example “2 interviews found · 2 analyses read”. One click expands it. The answer comes next, followed by what it’s based on.
- Interview passages: each quoted excerpt shows its speaker.
- Documents: interview notes, a brief, or an outreach draft opens in a card with its actions.
- Sources: under the answer, one chip per candidate, job, or search consulted. One click shows the details.
- Your feedback: say under each answer whether it helped.
- One row per profile: the name, the current job, and four status dots for skills, experience, interview, and logistics.
- A shorter file: the verdict and the four criteria on a single strip, the summary, then strengths and watch-outs.
- Every claim names its source: one click opens the interview excerpt at its timestamp, the passage from the CV, or the field from the candidate’s profile in your ATS.
- A question about the profiles: the field under the list opens the Assistant in a conversation linked to the search.
- Interface language: French or English, chosen by each user.
- Language of notes and analyses: your organization decides which language the interview notes, key takeaways, insights, and deal-breakers are written in. You can also ask for the language of the interview.
- Meeting badge: the recordings list tells meetings apart from interviews, with a filter by category.
- Analyze anyway: if a meeting was actually an interview, one click from its page runs the full analysis.
- Organization setting: detection can be turned off in the integration settings.
- The request: what was observed, where, and what had already been tried.
- Support reply: the diagnosis and what was fixed.
- Still outstanding: a check expected on your side or an announced fix.
- Without an interview: the interview criterion is left out of the calculation instead of counting as a middle value, which makes the ranking more accurate.
- View interview: from a profile that came from your ATS, the interview used to evaluate it opens in one click.
- Run search again: a search run again within 24 hours with the same requirements reuses the evaluations already done, and the result arrives much faster.
v2.44 to v2.50
The home page is rebuilt. The dashboard brings together the Assistant, the four ways to start an interview, your recent interviews, and today’s calendar. You can ask a question or call a candidate without leaving the page.The interview page takes a new shape. A single bar on the right groups the actions, the analysis, and the ATS panel. Interview notes are now always produced after each analyzed interview.
- Key takeaways: a new view, shorter than the interview notes, that sums up the interview in a few lines.
- Transformations: format interview notes, change their length, or translate them from the Actions panel.
- Speaker identification: a window suggests the names it heard and lets you correct who said what.
- Candidate CV: you upload it directly in the Analysis panel.
v2.30 to v2.43
Analytics arrives. You describe in plain language what you want to track, and Hirify builds the dashboard. Five starting points are offered, from pipeline health to interview activity. A Data reliability page shows which analyses are available, which ones are still locked, and the step that makes them available.The Assistant can build and find these dashboards, displayed right in the conversation.Nicoka joins the ATSs you can connect, alongside SmartRecruiters and Teamtailor.Your recruiting data gets richer. Syncing now brings in:
- Job families: a taxonomy that segments your jobs, which you can correct from the settings.
- Locations: jobs are mapped to French regions.
- Salary expectations: on the candidate side, and salary ranges on the job side, as annual salary or daily rate.
- Application status history: the data behind the velocity analyses.
v2.20 to v2.29
The Assistant gets more precise. It writes summaries, produces better candidate profiles, and accepts longer requests.Speaker detection gets stricter. The roles shown match what the interview says, and misattributed turns can be corrected by hand.
