← All services Draft
AI Automation
Automating workflows with AI. Document handling, data extraction, content pipelines, vibe-coding setups for teams. Realistic expectations, measurable outcomes.
In 2026, AI is neither magical nor trivial. It’s a new toolbox that works well when targeted at one concrete workflow — and disappoints when “thrown in somewhere” without a clear goal.
We work with teams that have a specific bottleneck: too much manual document handling, content production that’s too slow, coding that lags without AI support. We build the automation that resolves exactly that bottleneck — and not more.
How we work
We always start with the concept phase. AI projects nearly always fail on unclear requirements, not on tech. 2–3 days are enough to sharpen the use case, pick the right stack, and define success metrics.
In the main project we build the prototype, test with real data, and ship a productive setup your team can operate. We’re not a black-box AI consultant — we deliver code that sits in your repo.
Typical applications
- Document extraction: invoices, contracts, forms structured out
- Content prep: summarising, restructuring, translating
- Workflow priming: email classification, ticket routing, first-draft replies
- Data quality: dedup, normalising inconsistent fields
- Vibe-coding for teams: setup, training, workflows with Claude Code / Cursor / Aider — we do this in-house and share what we’ve learned
What we care about
- Pragmatism over hype. Not every task needs an LLM agent. Sometimes a rule-based solution is cheaper and more reliable.
- Privacy and EU hosting where possible. We advise on the trade-offs (accuracy ↔ confidentiality ↔ latency ↔ cost).
- Measurability. Before we build, we define what has to feel better afterwards.
- No SaaS lock-in. Your workflow logic stays with you — we deliver code in your repo, not a dependency you have to keep paying for.
What’s included (typical)
- Concept phase (use case sharpening, stack choice, success metrics)
- Prototype tested with real data
- Production setup (API integration, logging, error handling, cost monitoring)
- Training for your team (use, maintenance, extension)
- Documentation + code in your repo
When this isn’t right
If you just need a chatbot, off-the-shelf SaaS is usually better. If you’re building an entire platform with AI features, head to Platform Development — we combine the two there.
How it continues
AI projects start with the concept phase. To sort out what’s realistic in your setting first: free intro call.
FAQ
- Do you build ChatGPT plugins?
- We build productive automations embedded in your workflows — not standalone chatbots. Plugins are often the wrong shape.
- How do you measure success?
- Concrete metrics before we start: time per case, error rate, number of manual interventions. We only build what measurably improves after.
- Which models?
- Pragmatic — usually Anthropic Claude for reasoning tasks, OpenAI/open-source for cheap classification. EU hosting where possible. We pick per task.
Other services
All services →- Fixed price
Business Website
A marketing site that converts. Clearly scoped, fixed price, fixed delivery time. From intro call to launch in 4–6 weeks.
- Concept phase first
Platform Development
We build platforms organisations can live with for years. SaaS, internal tools, marketplaces, admin systems. Concept phase defines scope, then a fixed price.
- Concept phase first
Corporate Design
Brand identity that holds up. Logo, colour system, typography, design language. Built with regard for everyday usability over years, not just for the launch photo shoot.
- Fixed price
Concept Phase
2–3 days of focused team time. We sharpen scope, write the concept, calculate the main project. You keep everything — even if you build with someone else afterwards.