Stop reading ChatGPT tutorials. Stop watching LinkedIn theater. Custom AI automations for German service firms — built by a Freiburg-trained M.Sc. computer scientist, not by a marketing agency with a prompt-engineering certificate.
Every morning, another "game-changer." GPT-5, Claude, agents, MCP, n8n. You know you should act — but between clients, staff, and the day-to-day, you can't even find time to evaluate, let alone implement.
You've booked the workshops. Collected the frameworks. Seen the roadmaps. What you don't have: a single system actually running in your firm and saving real hours. Strategy without implementation is theater.
The new agency. The young Steuerbüro. The AI-native broker. They deliver the same result at half the hours. Your margins know it first. Your clients notice next. The window is closing — quietly, but it is closing.
You don't need another lecture on AI. You need someone who understands the models, writes the code, and ships the system into production at your firm.
We sit with your key people, observe the actual work, talk to the team members doing the repetitive tasks today. Then we decide where to start — and where not to.
Code, model selection, API connections to your existing systems. Weekly reviews — you see progress, not status reports. The first automation is typically live around day 14.
We train your team, document every system, and stay active for 30 days. At the end, everything runs without us — and we've built the roadmap for the next twelve months together.
A non-exhaustive sample. Most engagements combine two or three of these, tailored to your industry, your stack, and your actual bottlenecks.
Incoming invoices are auto-classified, pre-coded, and pushed into DATEV or Lexware — with sanity checks. Dramatically reduces manual processing on standard documents.
Monthly client reports generated automatically from GA4, Meta, Google Ads, LinkedIn — with narrative, not just numbers. Your team focuses on strategy instead of copy-pasting screenshots.
From property data and photos, high-quality listings produced in your house tone. Routine tenant communications (scheduling, FAQs, standard requests) handled semi-automatically.
Contracts, briefs, and case files summarized with key passages flagged. First drafts generated from your own template library — your lawyers edit instead of write from scratch.
LinkedIn profiles scored against your role requirements, personalized first messages drafted, response clusters summarized. Your team runs the conversations — AI prepares them.
Client conversations turned into structured needs analyses, needs analyses turned into comparative quotes — with transparent reasoning, in your house style.
Structured extraction from tender documents, auto-pre-filling of bills of quantities based on your project history, consistency checks against relevant standards.
AI pre-translation with terminology consistency against your translation memories, targeted post-editing by your linguists. Double the volume, hold the quality.
I'm Yaëlle Gaillard. M.Sc. Computer Science at the University of Freiburg, focused on Machine Learning and Robot Learning. Before Freiburg: B.Sc. Physics in Poitiers, and two years of Classes Préparatoires in Tours — France's intensive prep program for engineering school entrance.
My current research builds and benchmarks diffusion-policy models for robotic manipulation against the 2025 state-of-the-art. Translated: I don't just understand what AI models do — I build them.
What German service firms get from me: someone who genuinely understands the models and the code, but is pragmatic enough to ship the simplest system that actually works — not an overengineered platform nobody will maintain six months from now.
I deliberately work 1:1 with a small number of clients. No juniors, no subcontractors, no white-label agency chain. When you talk to me, you're talking to the person who builds your systems.
That simple. Before we start, we agree in writing on what "it works" means — typically hours saved per week in a specific workflow. If we miss that target after go-live plus the 30-day optimization phase, you get a full refund of the engagement fee.
You carry no model risk, no implementation risk, no "will this fit my firm" risk. I do. That's exactly why I only take three to four clients per quarter — I need to be convinced it will work before we start.
A free 30-minute call. We figure out together whether your firm is a fit, and you walk away with concrete first levers — even if we don't end up working together. No pitch deck, no closing tricks.