Frequently asked questions

A typical pilot completes in 4–6 weeks and is in production within 2–3 months. We give a concrete timeline after the discovery call, before any contract. Time depends on the use case, the state of your data, and internal approval processes — we surface blockers upfront, not mid-project.

We architect so your data stays under your control. We default to EU-region cloud (Azure EU, AWS Frankfurt) and operate GDPR-aligned. Self-hosted models are supported when regulation or policy demands it. We disclose every data flow transparently before contracting.

No. We build on top of your existing SaaS stack. We integrate CRM, ERP, project tools, comms and finance so data flows. If a system is genuinely a bottleneck, we say so — but replacement isn't our default answer.

An adoption programme is part of every engagement. Role-based workshops, internal champions and change communication are defaults, not add-ons. Without adoption AI delivers nothing — and that's already priced in. We measure usage and adjust the programme based on feedback.

We agree metrics before any code. Typical ones: hours saved per use case, processing speed, quality (errors, correctness), team active usage, and escalation volume. We don't promise "ROI %" without your data — figures always derive from your baseline.

We design with GDPR from the start: legal basis, data minimisation, retention limits, ability to respond to data-subject requests (delete, transfer, rectify). We default to EU-region services. We offer DPA contracts and DPIA support when needed.

The discovery call is always free. A single use-case pilot starts at €4,500. Ongoing partnership starts at €3,200/month, includes architect, optimisation and training. Enterprise is custom. All include source code handover, documentation, and metrics — no lock-in.

Yes. Source code, model configurations, and documentation are handed over to the client. We use open-source standards where possible (n8n, LangChain, pgvector). You're not locked in — if you choose to stop working with us, the solution remains yours.

In the discovery call we map the current state: where time burns, where decisions stall, where errors come from. We prioritise the highest-value, lowest-risk combination. Usually we start with a use case that has a clear metric and existing data — an early win builds confidence for the next ones.

We design guardrails in: source citations (RAG), human review at critical points, scoping answers to known facts, "I don't know" responses under uncertainty. We measure accuracy in production and update models regularly. For important decisions, a human is always the final decision-maker.

Yes. Data mapping and cleanup is often part of delivery. We start from a small use case where enough data exists, and only expand once the foundation is solid. If data isn't yet sufficient for any AI usage, we say so honestly in the discovery call.

Three options: (1) You take full ownership and we stay on call if needed. (2) An ongoing partnership where we handle optimisation and expansion. (3) We train your internal team to take over maintenance. The choice is yours — we don't build systems that require us.

Industry-agnostic, but focused on processes where humans do repetitive evaluative work — typically logistics, legal, manufacturing, e-commerce, insurance, healthcare, public sector, energy and consulting. Our team's background covers these industries from prior roles. If your industry isn't on the list we say so honestly.

Book a free 30-minute discovery call from the calendar. We walk through your situation, identify 2–3 opportunities, and email you a timeline and cost estimate within 1–2 business days. The decision is yours — discovery commits to nothing.

Not at all. At pilot end we transfer full ownership to you: code, configuration, documentation, runbook. You can continue yourself, take another partner, or shut the whole thing down. We design the architecture so you have no obligation to stay — which, surprisingly, is the main reason most teams do stay.

By default: EU region (Finland, Germany, Sweden or Ireland, by client choice). Model calls go to EU-based providers that do not train on your requests. Access is role-based, logged and auditable. In special cases we deliver fully self-hosted solutions where data never leaves your own datacenter.

This is common and healthy. We propose that IT is part of the first discovery meeting onwards. We walk them through architecture, encryption, auditability and escalation paths before a single user test happens. Most often, IT moves from sceptical to the project's best advocate once they see how it is built.

Three structural answers. (1) High-impact decisions are always human-approved — AI drafts, a human sends. (2) Each deployment has a named owner in your organisation, accountable for outcomes. (3) We log every decision so you can later check why the system made which call. Accountability cannot be "the AI" — it has to land on a human.

A small team is a deliberate choice, not a limit. We use carefully selected contractors when scaling requires it, but strategy, architecture decisions and delivery core stay in our team. That means you get the same person from discovery to production — which is surprisingly rare at larger consultancies.

We are model-agnostic: OpenAI, Anthropic (Claude), Mistral and open-source models like Llama 3 and Mixtral, depending on the solution. The common technical stack includes PostgreSQL + pgvector or Qdrant for retrieval, LangGraph for orchestration, Python and TypeScript for services. The choice depends on cost, latency, EU availability and accuracy needs — not our favourite logo.

Yes, but we always start with a diagnostic. In most cases the problem is not the model but the data structure, process fit or adoption. We run a 1–2 week "health-check" diagnosing why the current implementation isn't working. Sometimes we recommend fixing what exists, sometimes a rebuild — but only after the diagnostic, not based on sales.

Yes, and we are open about it. Sometimes a pilot reveals data isn't there, the process is too multi-faceted, or the business case was thin. In such cases we close the pilot cleanly and give you a documented learnings memo. You retain ownership of everything built. Beyond discovery, we always offer a satisfaction clause: pilot fees refundable if agreed metrics do not move within 90 days.