TalentHubfor Deutsche Telekom
A recruiting platform for four parties with competing interests, working on the same data while each only sees what they are allowed to see.
Where it started
Direct sales recruits through a network of independent sales partners. Applications come from social media campaigns, are qualified by phone by an external agency and passed to the right partner. Until then this ran on spreadsheets, mailing lists and phone lists.
Why nothing off the shelf
Typical applicant tracking systems know one company and its jobs. Here several legally independent partners work in the same system, each with their own brand, calendars and employment models.
Four levels, four views
Operator, agency, sales partners and the corporate approval desk are separate permission levels. The separation lives in the data model, not as a filter in the interface. Try it yourself below.
Try it yourself
Who sees what?
Pick a role. The table shows what that role sees of the same data, and what stays hidden.
| Application | Source | Status | Partner | Contact |
|---|---|---|---|---|
| #1042 | New | Nord | a.k…@mail.de | |
| #1043 | Qualified | Süd | m.b…@web.de | |
| #1044 | TikTok | Appointment | Nord | j.s…@gmx.de |
| #1045 | Approved | West | l.m…@mail.de | |
| #1046 | Appointment | Süd | p.r…@web.de |
The operator sees everything in order to run the process.
Sample data, simplified.
An applicant’s journey
Intake via landing page and quiz flow with automatic field mapping. Then source tracking, duplicate check, documented approval, phone qualification, partner assignment by postcode and radius, and appointment booking with capacity control.
Speaking in the partner’s name
22 automated email flows accompany the process. Sender, branding and legal details are stored per partner. To the applicant it looks as if the partner wrote personally.
Privacy in the foundation
Whatever belongs to the application rests on Art. 6(1)(b) GDPR. Anything beyond needs individually given consent. Access under Art. 15 and anonymisation under Art. 17 are system functions, not manual tasks.
How we worked
The platform was built with AI-assisted development as a way of working, not an experiment. Concept, architecture and review stayed with people. A feature request from a morning workshop could be live the same day.
applications in the first month
active email flows
permission levels
Today the process is traceable in one place: where an applicant stands, who touched it last, which email went out and on what legal basis.
Next project
Service Hub