The challenge
An alliance of several IT companies passes client requests between its members and staffs them from a combined talent pool. Those requests arrive by email and from meeting notes, each in a different shape. Someone has to read them, break them down into roles, search CVs across every member company, assemble a team and prepare anonymised CVs for the client — and then work out who owes whom when the request belongs to one company and the candidate to another. All of it by hand, in spreadsheets and email threads: slow, and with a real risk that the best candidate at a partner firm never gets found.
What we’re building
JARVIS is a multi-tenant platform for the whole alliance. AI reads incoming emails (Microsoft 365, Gmail) and meeting notes, splits them by topic and turns each request into a structured case — client, roles, headcount, required skills, rates. Nothing is created blindly: a human confirms each case in the inbox with one click. From a few lines of input, AI generates a job description and sorts requirements into must-haves and nice-to-haves. Matching then searches across every partner’s pool — vector search, hard filters and AI re-ranking that annotates each candidate with strengths, concerns and a recommendation. Results become a shortlist, the shortlist becomes anonymised CVs for approval and submission to the client. Placements, margin splits between companies and the alliance’s running balance are tracked automatically, and a traffic-light dashboard shows what needs attention now.
Results
The platform’s foundation — a multi-tenant data model and an alliance layer with data and permission boundaries between companies — is in place, with further modules shipping on top. The goal is clear: a request in an inbox reaches a shortlist drawn from the entire alliance in minutes rather than days, no suitable person in a partner’s pool goes unnoticed, and settling up between companies stops being a spreadsheet exercise.