AI Job Fit
Console
The console watches postings published on career sites, matches them against your bench every night, and prepares the file to send to the client. You arrive in the morning to a list that is already sorted.
In final testing · Built on the engine already running the extensionAt night, career sites are re-read with nobody in front of the screen. The rest of the time, a brief that arrives by phone or email is typed in by hand and follows exactly the same path. The only step that needs a decision is the last one, and it waits for a deliberate click, never an automatic one.
Screenshots below show the console's actual interface, which is currently French only.
A pile that empties. It holds only postings picked up overnight where someone on your bench actually fits, best match first, because the morning question is who you can place. Three ways out: put a profile forward, keep it for later, drop it. Opening a posting to read it does not count as a decision, so nothing ever slips through unseen.
The short list is computed in batch overnight, so it appears instantly. Five dimensions, the gaps, the strengths, and the requirements that genuinely rule someone out. Putting a profile forward costs nothing and calls nothing: generating the file is the one expensive step, and it takes a deliberate click.
On request, the consultant's CV is rewritten for the posting. Reordered, tightened, never invented, and the difference from the original is shown line by line. You read it, adjust it, export it. The file is an attachment to an application, never the thing that decides its status.
Every consultant put forward on a posting is followed on its own: to send, sent, chased, interview, placed, turned down. Two people on the same posting are two separate stories, and a no on one says nothing about the other. Anything that has not moved for ten days says so, because a file sent three weeks ago looks exactly like one sent this morning.
Each profile is imported once and becomes a structured record. That record is what gets matched against the market every night, not a PDF someone would have to open.
Availability, skills, and the history of postings proposed and decisions taken. You see at a glance what has been tried for that profile, and what hasn't.
Paste a careers page address: the platform is recognised on its own in most cases. A source that goes quiet for several nights is flagged, because a site that changes its template must not vanish silently. Each one can be renamed, repointed at another URL, or pulled out of the watch afterwards: nothing is ever deleted, so what has already been sent to a client stays intact.
Every status and every decision is a timestamped event. You always know who decided what and when, and the history can't be rewritten.
A posting pulled, a consultant gone: nothing is deleted, everything is archived. You can always go back to what happened six months ago.
A semantic pre-filter discards obviously unrelated pairs before any AI call. On the real bench, that removes 95% of the work, and as much of the bill.
A file reorders and tightens what the consultant actually did. The gap with the original CV is shown line by line, and you read it over before sending.
The database, your consultants' CVs and the decision log live on a European server. Only the text a scoring actually needs goes out to the AI model, and none of it is reused to train it.
The console is a service sitting on your database, not a closed box. A bridge to your ERP or your CRM, Odoo for one, a field specific to your trade, an export in the format your client expects: those are developments we take on for you.
It already runs internally on a real bench. If you place consultants and the above speaks to you, tell us in two lines how you work today. That is what will decide what gets built next.