AI agents · Berlin
I design, build, and ship agents for marketing, sales, and operations: research, enrichment, outreach drafts, monitoring. They run in your own tools, and your team approves what goes out.
Use cases
The agent reads company sites, job posts, and reports, then writes the findings into your CRM before anyone picks up the phone.
First messages in your voice, based on real research. Nothing is sent until a person on your team says yes.
Rankings, backlinks, schema, and whether ChatGPT and Google AI answers mention you. A short weekly report, not a dashboard nobody opens.
Pricing pages, launches, and messaging changes, summarised when something actually changes.
Briefs built from search data and customer questions, drafts that follow your style guide, ready for an editor.
Point an agent at your docs, wikis, and campaign history, and your team gets answers grounded in your own data.
Packages
from €490
One AI agent for one marketing workflow, working in your stack
from €8,000
One AI agent designed, built, and shipped for a real workflow
Ready-made
My own library of production-ready agents for lead generation, ABM, SEO, and AI visibility. Take one agent, such as the Angle Test on one offer, or the whole engine for your motion.
Visit Demand Engines ↗Process
We choose the repetitive job that costs the most time. One workflow, one clear definition of done.
I map the steps, data, and approvals, then build the agent in your tools: n8n, OpenAI or Anthropic, your CRM.
The agent runs on your real inputs with a person approving every output until the results are right.
You get the source, a setup video or handover session, and docs your team can maintain.
FAQ
Start with Marketing Agents if you have one clear, repetitive marketing job and want to see an agent working within a week. Choose the AI Agent Sprint when the workflow touches several systems or teams and needs scoping first.
Not unless you decide it should. Every step that sends a message or changes data waits for a person to approve it. You can loosen that later, once you trust the output.
n8n or Make for the workflow, the OpenAI or Anthropic APIs for the model, Custom GPTs or Claude Projects for simple assistants, and Supabase with a small React app when an agent needs its own interface.
Agents run in your own accounts where possible, and I use API access, where OpenAI and Anthropic do not train models on your data by default. We sign a data processing agreement, and I prefer EU hosting for n8n. Have your data protection officer review the final setup.
Model and hosting costs depend on volume. I estimate them in the scoping call and design the agent so the monthly cost stays predictable.
You get the source code and documentation, so your team can adjust the agent. I can also maintain it for you; we agree the details at handover.