AI your team actually uses. On your terms. Kept honest.

Most companies bought the licences and got scattered chat use, no rules and no measurable result. I start with a fixed-price AI Check, then build the agents that fit how you work, with your data kept where it has to be. Prices published below. You keep the keys.

Works with: ClaudeOpenAIMicrosoft Copilotopen-weight models · Runs on: your cloudyour hardware

Three problems, one method

Which of these sounds like your company?

Same method, same price ladder, same person. Pick the one that hurts.

AI onboarding

The trigger
Everyone has a licence, nobody has rules, and nobody can say what AI has done for the business.
What I do
Rules people actually follow, the right tool for each team, hands-on training, and the first workflow that pays. Measured, not assumed.

Custom agents

The trigger
Off-the-shelf copilots don't know your process, your systems or your limits.
What I do
Agents built for one job (intake, drafting, review, triage), wired into the systems you already run, with permissions, guardrails and tests against your real cases.

Data control

The trigger
Client files, HR or regulated data can't go to whichever AI provider a tool happens to use.
What I do
A gateway that decides which data may reach which model, every request logged, and open-weight models on your own hardware where nothing may leave.

How it works

Four steps. Every one has a price before it starts.

  1. STEP 1

    AI Check

    CHF 4'500 fixed · 2 weeks

    Which AI tools your people actually use, what data flows where, where AI would pay first, what needs rules, and what I would leave as it is. You get a written plan with the first build scoped and priced. If the answer is "you're fine", that's the answer, and it's still worth the number.

  2. STEP 2

    Build

    CHF 9'000–15'000 fixed · 4–6 weeks

    One agent or workflow, end to end: CHF 9'000. With a gateway or private models for sensitive data: CHF 12'000–15'000. Runs on your real data. Your team trained. Documented so you are not dependent on me.

  3. STEP 3

    Care

    CHF 1'200 per month · cancel monthly

    Model and tool updates, fixes when a vendor changes something, a monthly usage and spend report, business-hours escalation. Your IT holds first line; I'm second and third.

  4. STEP 4

    Sign-off

    Included in Care · or CHF 4'000 / year

    One signed page a quarter: which agents ran, what they cost, which data went to which provider and under which contract. The document your AI policy and your auditor both promise and nobody produces.

Pricing

Published and fixed.

No hourly rates. Every number is agreed before I start and does not move. Quoted in Swiss francs; EUR and USD on request at the same level.

WhatPriceIncludesTerms
AI Check CHF 4'500fixed
  • Inventory of the AI tools in use, and who uses them
  • Data-flow map: what goes to which provider today
  • Where AI pays first, ranked by effort and return
  • First build scoped and priced, or "leave it"
Two weeks. About four hours of one person's time on your side.
Build CHF 9'000–15'000fixed, quoted after the Check
  • One agent or workflow, wired into your systems, with guardrails and tests: CHF 9'000
  • With a gateway or private open-weight models for sensitive data: CHF 12'000–15'000
  • Team training session
  • Documentation, repo in your name
Four to six weeks. Licences and hardware are yours to buy; I spec them, I take no margin on them.
Care CHF 1'200per month
  • Model and tool updates, vendor-change fixes
  • Monthly usage and spend report
  • Business-hours escalation, second and third line
  • Quarterly Sign-off
Cancel with one month's notice. No 24/7 pager. Instead, a runbook your people can follow at 02:00.
Quarterly Sign-off CHF 4'000per year, standalone
  • Quarterly signed report
  • Which agents ran, and what they cost
  • Which data went to which provider, under which contract
  • Patch and model level recorded
For AI systems you run yourself and want kept honest.

Prices exclude VAT. Invoiced by GlobalWerk GmbH, Switzerland.

Not sure where to start? Start with the Check. It tells you what is worth building at all.Request a 30-minute call

What stays yours

Everything.

  • The repo, the keys, the box. One docker compose, one git repository, in your name.
  • Your hardware, your accounts. Models run in your cloud account or on your own server; the API keys are yours. I never sit between you and a vendor.
  • Your people, trained. The session is for the person who runs it, not for management.
  • A written map of what may go where under the data protection law you answer to: GDPR, the Swiss nFADP or your sector's rules. It survives whoever builds the next thing.

What I don't do

Three things, on purpose.

  • 24/7 on-call. Business hours, honest response times, and a runbook for the middle of the night.
  • Resell licences or hardware. I recommend; you buy directly. I take no margin on either.
  • Bill by the hour. Every number on this page is fixed before I start.
Open source, and you own the code. Fire me and it keeps running. That is the point.

Also

Training for the team that has to run it

CHF 2'000 half-day workshop, up to 15 people

Running AI infrastructure with Docker and GPUs, Claude Code in a real workflow, or building agents with their own role and guardrails. Hands-on, on your stack, recorded.

See the workshops

5'338
engineers taught on Udemy
4.55
average rating from 923 reviews

Track record

Ten years of running the thing, not just recommending it.

10+
years in production infrastructure: Docker, cloud, on-prem
2020
sold the Swiss DevOps consultancy I co-founded (56k.cloud)
5'338
students in a top-rated infrastructure course
Zug
based; working remotely across Europe and the US, in English and German
Docker Captain, first cohort Traefik Ambassador Conference speaker: DockerCon, universities AI and infrastructure lead inside a regulated Swiss company
Brian Christner speaking at a conference
Brian Christner

Who I am

I run AI and infrastructure inside a regulated Swiss company: dozens of automated workflows and agents in production, under real compliance rules, with people who did not ask for them. Before that I co-founded and sold a Swiss DevOps consultancy, and spent a decade teaching containers publicly as one of the first Docker Captains.

I mention the regulated part deliberately. Buying licences is easy. Getting AI used so that the auditor, the messy data and the sceptical sysadmin all sign off is the part I've already done.

Based in Zug, Switzerland. I work with companies in Europe and the US who want AI their people actually use: theirs, compliant, and boring in the best sense.

Next step

A 30-minute call.

If your situation isn't a fit, I'll say so on the call rather than sell you a Check you don't need.

No newsletter, no sharing. You get a reply from me, not a funnel.