Build
Every system starts from a platform we maintain, so a project begins about 40 percent complete. Delivery takes weeks, and the result is guaranteed for three years.
01Packaged builds
Each one is adapted to your data and workflow. Only the starting point is standard.
From 24,900 EUR6 to 10 weeks
From 17,900 EUR5 to 8 weeks
From 12,900 EUR4 to 6 weeks
02Built to your brief
4 to 8 weeks · 15,900 to 39,000 EUR
A production AI application on our platform, with a data layer, admin, model gateway, evaluations and deployment.
You get
The running system, source code, documentation, a handover session and the three-year guarantee.
Not included
Features added after acceptance, and infrastructure costs.
1 to 4 weeks · 4,900 to 18,000 EUR
Agent workflows connected to the tools you already use: CRM, ERP, email, telephony and spreadsheets.
You get
Working automations with monitoring, approval steps and an off switch.
Not included
Changes to your existing systems beyond the integration points.
2 to 5 weeks · 6,900 to 19,000 EUR
Internal software without AI: record systems that replace paper, dashboards and portals.
You get
The running tool, source code, documentation and the three-year guarantee.
Not included
Entry of historical records, quoted separately.
8 to 16 weeks · 39,000 to 95,000 EUR
Larger builds with several components: fine-tuned local models, deployment on your own GPUs and multiple integrations.
You get
Everything in AI Core, plus a hardware specification, model cards, runbooks and an external penetration test.
Not included
The hardware itself. We specify it and you buy it.
Not sure where to start? An AI Proof gives you a working prototype on your own data in two weeks, for 7,900 EUR. Half of it is credited against the build.
See the price list03The platform
Pattern is the platform every system is built on: a maintained set of packages, templates and deployment profiles. Each new system starts from it and passes the same tests and security checks.
That is why a project starts about 40 percent complete, why fixes reach every client quickly, and why we can guarantee our work for three years.
Models are configuration, so moving to a newer model is a reviewed change rather than a rebuild. You own the code we write for you, and Pattern is licensed to you permanently.
04On-premises
Local models keep your data on site. On-prem deployment costs 4,900 EUR and includes an offline installer.
Class A · Small
Runs 7B to 14B models at 4-bit, plus speech recognition, embeddings and rerankers.
3,000 to 5,000 EUR once · you buy it
Class B · Single pro GPU
Runs 30B to 70B models, several at once, and fine-tunes small models. The standard setup for Conversation Intelligence.
About 12,000 EUR to buy · or rented in the EU
Class C · Cloud accelerators
Large models, fine-tuning jobs and bursts, at EU providers.
Hourly · passed through at cost
05Security and governance
OWASP ASVS level 2 and the OWASP Top 10 for LLM applications, with a threat model for every system.
Every change is reviewed by a second model, and by a senior engineer when it touches sign-in, data export, payments or infrastructure.
Signed, scanned images and a software bill of materials with every release.
Personal data stays on local or EU-hosted models. It never goes to a US-hosted model.
An AI Act register entry for every system, and a transparency notice on every assistant.
Contact
A 30-minute call with our engineering team. Bring a sample of your data if you can, and a fixed quote follows within three business days.
30-minute call
Booking calendar (Cal.com) loads here
Or write to [Email address]