The system behind our campaigns.
A platform that connects analysis, ad creatives, campaigns and evaluation into one loop. What runs today we use in every project. What is still missing we develop with the clients who need it.
01 Analysis
Company and labour market: who fits the role, where do these people live, what speaks to them?
As of today
Five parts are running. The sixth is the reason for the research project.
Recruiting Hub
All roles, applicants and channels in one place. Six-stage pipeline, automations that handle confirmations and follow-ups.
In useAnalysis
A company profile and an assessment of the labour market for the role, with recommendations for the approach. AI-assisted, checked by us.
In useCreatives in variants
Several ad creatives per role, generated from the recommendations, approved by a person. AI images are labelled.
In useMeta integration
Campaigns, creatives and results flow directly with Facebook and Instagram. Detects when a creative burns out and reports it.
In useClient view
You see applications, appointments and status per role without clicking through ad accounts.
In useOptimisation
The system should learn from results which creative properties work for which audience. The research core, in development.
Research, in developmentResearch
Every campaign should learn from the last. How is not yet solved.
Today the knowledge of why an ad worked sits in experience and spreadsheets. The project aims to turn that into a method that predicts from little campaign data which creatives work for which audience. The German research allowance certification body recognised it as a research project because these questions cannot be answered with existing methods.
The honest status
The learning method is not built yet. First, creatives must be describable and real campaigns must deliver data. Both are happening now, with pilot clients. We publish results once they are measured, not before.
Research, in developmentThe path
Developed on real roles, not on the drawing board.
Make creatives controllable by properties
Tone, image mood, text density, structure: every creative gets describable properties. Without them a learning method has nothing to build on.
Collect real campaign data
Results per creative from live campaigns, with pilot clients. That is the data the system should learn from.
Develop and test the learning method
A method that predicts from little data which creatives work and proposes the next generation. The research core of the project.
Open up to other agencies
Once the system works for our own clients, other recruiting agencies can use it.
What the system does not do
Four limits, set in code.
01No assessment of applicants
The system captures, makes contact and books appointments. Who gets hired is decided by the employer. No AI rates anyone.
02Creatives never see people
Creative generation only receives anonymous parameters. An allowlist in the code prevents applicant data from getting there.
03A person sets the audience
Job ads run broadly, without age or gender targeting. Who sees an ad is decided by Meta under its rules, not by our system.
04Checked before it runs
Every creative goes through approval. AI-generated images are labelled. Every message an AI writes is recognisable as such.
Certified
Recognised as a research project.
In June 2026 the German federal research allowance certification body classified the project as research and development. That is an expert judgement that something new is being created here, not a marketing promise. We name results only once they exist.
Pilot clients
Whoever recruits with us now helps shape the system.
Pilot clients get full support at normal conditions. In return their campaigns feed into development: which creatives run, what burns out, what the client view should show. What emerges, they use first.