Service · from DKK 50,000

One concrete AI solution.
Built end-to-end.

In 4-6 weeks we build one pilot that solves a concrete problem and delivers documented ROI. Twenty percent of the price is tied to documented savings in the 30-day measurement window. That is risk sharing, not just promises.

n8n plus Claude API plus your systems. Hosted locally or at your premises. Audit trail on everything. Code in your own Git repo.

Book Discovery Maturity assessment first?
§ 01
Three variants

Choose by ambition

Focus pilot is the right choice if you just want to prove value. Standard hits broader. Scale pilot builds the foundation for an organisation-wide AI rollout.

01 / 4 weeks

Focus pilot

DKK 50,000-80,000

One concrete AI solution built end-to-end. An n8n flow or a custom script, documentation, training of 2 power users and a 30-day measurement framework.

Best when

You have one clear use case and want proof of value before a larger investment.

02 / 5-6 weeks

Standard pilot

DKK 80,000-120,000

Pilot with 2-3 connected use cases plus integration with your systems (CRM, ERP, document archive). Includes change management and continued training.

Best when

You want to go broader and see how AI works across processes.

03 / 6+ weeks

Scale pilot

DKK 120,000-150,000

Larger implementation with multiple integrations, custom UI, audit trail and deeper training of 5+ power users. Ready to scale after the pilot.

Best when

You have a plan to roll out AI across the organisation and want a strong foundation.

§ 02
Examples

What clients typically start with

Concrete use cases we have seen work in European mid-market. Your own pilot may look like one of these, or something entirely different.

01 / Law firm

Client onboarding RAG

Internal Claude portal with contracts, standard clauses and precedent. Reduces contract drafting time by ~50 percent.

02 / Accounting

Receipt classification

Automatic classification of scanned receipts against the chart of accounts. 80-90 percent correct out of the box with human review.

03 / IT services

Knowledge base plus RAG

Internal search across customer documentation. First-line support handles 30-40 percent more tickets without escalation.

04 / General

Email draft replies

n8n flow: new email → Claude classifies plus drafts reply → human review → send. Saves 15-30 min per employee daily.

§ 03
Delivery

How we work in seven steps

  1. 01
    Discovery
    30-minute conversation plus use case selection plus scope agreement
  2. 02
    Brief
    Pilot Canvas: problem, solution, data, success metrics, risks
  3. 03
    Build
    Typically built in n8n plus Claude API plus your systems
  4. 04
    Test
    End-to-end test, security review, GDPR check
  5. 05
    Training
    2-5 power users get 4-8 hours of hands-on training
  6. 06
    Go-live
    Pilot goes live with measurement framework and audit trail
  7. 07
    30-day measurement
    ROI report based on actual data after 30 days
§ 04
Deliverables

What you get

  • → Working AI solution running at your premises (n8n flow or custom script)
  • → Integration with your systems (CRM, ERP, document archive where relevant)
  • → Complete documentation: architecture, prompts, data flow, runbook
  • → Training of 2-5 power users (4-8 hours hands-on)
  • → 30-day measurement framework with baseline measurement before go-live
  • → ROI report after 30 days with concrete numbers
  • → Audit trail PDF documenting AI models, prompt versions and data flow
  • → Code in a private Git repo you can clone to your own infrastructure
  • → 30 days of support after go-live for fixes and adjustments
§ 05
Questions

What we are most often asked

What do you build on?
n8n as the agent orchestrator, Claude API (EU residency) as the primary LLM, Qdrant for vector search, the client's own APIs for integrations. Hosted locally or in the client's infrastructure. Pilot data does not leave the company without an explicit agreement.
What if ROI does not materialise?
We offer risk sharing: 20 percent of the price is tied to documented savings in the 30-day measurement window. If the baseline does not improve measurably, we refund that portion. Rarely needed, but important to say up front.
How hard is it to maintain afterwards?
With Focus pilot, your own power users can typically maintain it themselves. For Standard and Scale pilots, we recommend an Implementation Retainer or an internal AI champion role.
Next steps

30 minutes to find the right pilot

We start with a Discovery call: what is the most important use case, what data do you have, and how big an investment matches your ambition. Confidential and without obligation.

Book Discovery See the full catalogue (in Danish)