Common questions, real answers.

With a discovery workshop (€5–15K, 1–3 weeks). We diagnose where software, data or AI actually fits in your situation and produce a written recommendation. Most build engagements grow out of a discovery, but you can also bring us a clear scope.

Discovery workshops can usually start within 2–3 weeks. Build engagements depend on team availability — typical lead time is 4–6 weeks from signed proposal to first sprint.

You work directly with us. Our own people are distributed across Europe and beyond. We focus on the quality of our people and the work, not their location.

Weekly demo + async written progress updates. We use whichever tool you already use (Slack, Teams, email). One named project lead is your single point of contact.

Discovery is fixed price. Build engagements get a fixed quote per scope; we don't move on you mid-project. Partnership rates are rolling and adjustable monthly.

Both, depending on what fits the work. New product builds often work better as T&M with caps; well-defined integrations or rebuilds are usually fixed scope.

In partnership engagements with a strong product thesis, sometimes. Always case by case, never as a replacement for being paid for delivered work.

You. Code, architecture, and intellectual property are fully transferred to you. We retain the right to reference our involvement publicly with your permission.

TypeScript/React/Next.js for web apps, Python for data and ML, n8n and custom code for orchestration, Postgres + lakehouse for data, OpenAI / Anthropic / open-source LLMs for AI agents. We pick stack based on your team's ability to maintain it post-handover.

That's a constraint we design around. We document, train, and choose tooling your team can actually live with. If you don't have an internal team yet, we offer a partnership engagement to bridge that gap.

We follow OWASP guidance for web apps, encrypt data at rest and in transit, implement least-privilege access, and design with the EU AI Act and GDPR in mind from day one. For regulated industries we work with your security/compliance team directly.

Strongly pro. Most of our stack is open source. Where commercial tools win on time-to-value (e.g. Vercel, Supabase, n8n cloud), we use them — but we avoid lock-in patterns that would prevent you from migrating later.

When the underlying process isn't mapped, when the data doesn't exist, when the cost of errors is high and there's no human in the loop, or when a deterministic system would do the job better. We say no to a meaningful share of AI requests.

We classify the system's risk level early and design accordingly. For high-risk systems we implement the documentation, monitoring and human-oversight requirements. For limited-risk systems we ensure transparency obligations are met.

Yes — within the right scope. Single-purpose agents (document classification, RAG over a defined knowledge base, scheduled workflow agents) are production-stable today. Open-ended general-purpose agents are still research-grade for most enterprises.

Per-system: golden test sets, structured evaluations on each model change, human-rated samples, output drift monitoring. We won't deploy an AI feature you can't measure.