Jack and Jill AI

Jack and Jill are two AI recruiters: Jill screens candidates for companies, Jack helps candidates find work. It’s one of the fastest-growing AI startups in Europe, almost entirely engineers, and I joined as the only marketer, working straight with the two founders. The brief was simple to say and hard to do: turn a reactive LinkedIn presence into a proper content and launch engine, and build the whole thing on AI.

Building an AI content engine

I started by getting the two founders in a room and pinning down what we actually stood for: one audience, one problem, three things we did better than anyone else. That became a one-page brand strategy the whole company worked from, which quietly settled a lot of the founders’ conflicting instincts.

Then I built the machine. A multi-format content agent took raw inputs from our weekly founder sessions and turned them into LinkedIn posts in each founder’s voice, X threads, customer emails and video scripts. The whole team could publish daily without a content team behind them. The company LinkedIn went from 5,000 to 20,000 followers in three weeks, and a new X account from zero to 5,000 in the first.

Launching, fast

The interesting part of an AI product company is how often you get to launch. I ran five product launches in four weeks, scaling content across seven platforms, with launch videos I scripted and filmed in an in-office studio we set up. The best of them cleared 100,000 impressions on LinkedIn and brought in a thousand sign-ups each.

That same engine drove a US market entry off the back of a research campaign on bias in AI hiring: an interactive game and a white paper that turned a buyer objection into our strongest credibility asset. It brought in 10,000 sign-ups in a month with no extra headcount.

Proving what to build next

With no product-marketing resource, the team needed a faster way to decide what to build. I surveyed users, then vibe-coded a working CV-scoring tool in three days to test one idea with real behaviour rather than opinions. It scored higher on NPS than our main product, so we built and launched it, and I reused the same validate-then-launch playbook from then on.