The baseline — three things that need to be true.
The size where AI pays back fastest: one automated process is a visible share of total cost, so the return shows up in months rather than years. Not a hard floor — the smallest deployment we have seen work is a 7-person firm running three workflows.
A process that runs every day and grows with the business: documents, inbox or phone. So far the only way to keep up has been to put more people on it, and that is the cost that compounds.
Nobody inside has taken AI to production yet, or you need to know whether hiring for it is worth a salary before you commit to one. A pilot answers that in weeks, and trains one of your people to own what we ship.
Clears the bar. But should you move now?
The big players are already there. Most companies your size are not — which is exactly why moving now is worth something. That window has a shelf life.
Eurostat 2026
Start small, prove the ROI, then scale to production.
An assistant for one role or team: prompts, knowledge base, integration into your work channels, a team playbook, and 30 days of support.
An inventory of the AI systems already in use, risk classification, a gap analysis of your deployer duties, and a 90-day roadmap with a named owner for every gap.
Two-week fixed sprint: readiness scorecard, a matrix of 10–15 use cases with ROI in €/year, a clickable prototype of the top case, EU AI Act and GDPR classification, and a 90-day roadmap.
1–3 business processes: n8n/Make + LLM, integrations, documentation, internal champion training, EU hosting.
Fixed-scope proof of concept to a go/no-go on your real data: working prototype, business viability with ROI, risks, roadmap, and cost to scale.
Inbound calls handled end to end: intake, qualification, booking, escalation. ElevenLabs/Vapi + telephony + CRM. Never miss a call.
Pilot agent (one channel, 1–2 integrations, guardrails, dashboard), then production: 4–5 integrations, prompt A/B testing, regression evals.
Benchmarked against 45 European providers with published prices, July 2026.
| Cheap shop | allinc lab | Boutique / agency | |
|---|---|---|---|
| PoC | from €5,000 | €8,000–12,000 | statworx €12,900 · Crux Digits €20,000 |
| Agents | €1,000–5,000 | €8,000–12,000 | €25,000–150,000 |
| Voice | €2,500–4,500 | €3,000–8,000 | from €90,000/year |
| Time to live | — | 2–6 weeks | 10–20 weeks |
An €8,000 proof of concept is about a week of senior engineering at the €1,200–1,800/day European market rate. We bill from a lower-cost EU market, which is why the same seniority costs less without moving your data outside the EU.
Since 2 August 2026 the high-risk obligations are in force: Art. 9–17 for providers, Art. 26 for deployers. The Art. 4 AI-literacy duty has applied since February 2025. Penalties reach €35M or 7% of global turnover for prohibited practices, €15M or 3% for high-risk breaches.
An inventory of every AI system already in use, risk classification against Art. 5/6/50, and a gap analysis of your Art. 26 deployer duties.
A named owner for every gap, a fundamental-rights impact assessment and a data-protection impact assessment (FRIA, DPIA) where required, a 90-day roadmap, and a one-page summary your board can actually read.
EU hosting by default. Self-hosted models where the data is sensitive, which removes the API vendor from the picture entirely.
Two ways in: the AI Readiness Assessment covers classification and a 90-day roadmap as part of a wider look at where AI would pay off, while the EU AI Act Audit goes deeper on compliance alone when that is the only thing driving this. Running a build inside your own perimeter instead of a hosted model adds 20–40% to its cost.
We prove value fast, then scale what works.
We map your processes, identify automation candidates, and score them by ROI, feasibility, and strategic value.
Build the highest-ROI automation in production. Real data, real integration, real measurement.
Roll out to full volume, add integrations, train teams, and establish monitoring dashboards.
Continuous improvement: new automations, performance tuning, and expanding to adjacent processes.
We agree one measurable KPI before we start — containment rate, hours saved, extraction accuracy — and write it into the contract. Miss it and you get 50% of the fee back, or a free extension until we hit it — your choice.
None of the 45 European AI providers we benchmarked offers a guarantee tied to a metric.
Benchmarked against 45 European providers with published prices, July 2026.
That is the failure mode we were hired against. Product background at scale, one KPI written into the contract, and a pilot that ends in a go/no-go rather than an invoice for the next phase.
A typical automation runs at €20–80/month for the VPS and €30–200/month for LLM API calls. Larger builds add €200–2,000/month of infrastructure. The 7-person firm mentioned above runs three workflows at €20–30/month. Self-hosting removes the API bill entirely.
That tier is crowded with providers who have never shipped: one competitor states outright that it has no clients yet, another shows Google reviews where client metrics should be. We scope against a KPI and write it into the contract.
A permanent senior AI engineer is a €90,000+/year commitment plus months of hiring — before you know whether the use case works at all. A pilot answers that question in 2–6 weeks, and trains one of your people to own what we build.
EU hosting by default, self-hosted models where the data is sensitive. Nothing leaves the EU, and for sensitive workloads there is no third-party API in the path at all.
You keep the written report, the roadmap, the prototype and the source code. The sprint fee is credited in full against the build if you go ahead within 60 days. There is no penalty for deciding not to.
The free scan takes 30 minutes and usually happens within a week. A two-week sprint starts as soon as the scope is signed.
You walk us through one process. We come back with 3–5 tasks worth automating, ranked by value, with a €/year figure on each. No deck, no obligation.
The call is with the person who would build it. Not a salesperson, not an account manager. The sprint fee is credited in full against the build if you go ahead within 60 days.
If EU AI Act readiness is what is driving this, ask for the readiness call instead — the high-risk obligations took effect on 2 August 2026.