AI-generated
Embed AI permanently, not just introduce it.
Six weeks of structured build: one person learns, documents and trains two colleagues. The workflow stays, even when the tool changes.
Fixed-price engagement · 6 weeks · qualitative range, details in the initial consultation
Practical programme with EU AI Act compliance connection
Six weeks. One person. One documented workflow.
W1 — Champion selection
Identify the right person and narrow down the target process. Clarify tool approval.
- Champion profile: brief note on 4 selection criteria
- Process scoping document (max. 1 page)
- Tool whitelist start: 1–2 approved tools
W2 — First steps
Champion applies the approved tool to the real target process. First workflows emerge.
- First prompt draft for the target process
- Short protocol: what works, what does not
- Open questions for the next support session
W3 — Refine the workflow
Expand the prompt library, write down the step-by-step workflow.
- Written workflow (version 1): step-by-step for colleagues
- Prompt library (3–5 reusable building blocks)
W4 — Stabilise the workflow
Test workflow on real cases, document error sources, define human-in-the-loop.
- Workflow version 2 with human-in-the-loop markers
- Short error and exception list: when does the human step in
W5 — Train colleagues
Champion trains two colleagues. We observe what needs explaining.
- Training protocol: what worked, what was unclear
- Workflow version 3 with improvements from the training
W6 — Evaluation
Measure results, prepare rollout decision or document an honest stop.
- Pilot evaluation: time saved before/after
- Rollout proposal or stop finding (for internal use)
Who is suitable as a champion, and who is not.
Four observable criteria. No job title required, no coding experience. Validation possible in three steps.
Suitable
- Already uses AI tools like ChatGPT or Claude privately, without being prompted
- Names concrete, recurring process pain points, not just "AI is interesting"
- Understands both IT and the business side, can switch between both languages
- Experiments independently, has already tried something without waiting for permission
- Has 2 to 4 hours per week during the pilot for structured support, plus their own learning time
Not suitable
- Was nominated by management but shows no personal interest
- Cannot name concrete process pain points, only abstract "AI interest"
- Expects external consultants to build everything without becoming active themselves
- Has no time capacity, insufficient management support for release
3-step validation: 30-minute conversation › have them build a prototype themselves › have them train a colleague.
1st place BitGN Hackathon Vienna · April 2026 (AI Factory Austria)
79 out of 104 points, solo, 11 April 2026. 2nd place: 65 points. A multi-agent system with Soft-SGR and five-layer injection defence. A repeatable process, built daily since late 2024, not invented on the day. Exactly what I pass on in workshops and the champion pilot.
5 questions before our first call
Answer these 5 questions in advance - we will start more efficiently.
Who is the decision-maker? (Business unit, IT, CEO, Compliance)
Why we ask this
Depending on the stakeholder, we need different language, different output formats, and different escalation paths. A CTO asks different questions than a compliance lead.
What budget range is realistic? (Sparring from 450 EUR / Pilot 3,000-5,000 EUR / Custom Build five figures+)
Why we ask this
The budget range determines the format, not the ambition. With this information we can recommend the right format directly, rather than developing options first.
Do I have a concrete use case, or am I still exploring?
Why we ask this
Concrete use case: Sparring or Pilot. Open exploration: Audit as a starting point. This distinction saves both of us half an hour in the first call.
EU AI Act / GDPR: must the solution run on-prem, or is EU-hosted cloud sufficient?
Why we ask this
The hosting requirement fundamentally affects tool selection and pilot scope. Strict on-prem requirements limit available models - better to discuss this upfront.
Who on my team can continue building after the pilot? (1 developer, 1 PM, nobody)
Why we ask this
The Champion Pilot only works if someone can take the baton. If nobody is available today, we plan the programme accordingly - or recommend clarifying team readiness first.
No right or wrong answers. These questions help us suggest the right format without wasting time in the call.
Since 02.02.2025: What managing directors need to know.
Most SME pilots fall into the Minimal or Limited category. What matters: five documents that protect you in 80% of situations.
Prohibited
Social scoring, emotion recognition in the workplace, biometric mass surveillance
Prohibited since 02 Feb 2025, no use possible
High
HR scoring in hiring or dismissal, educational assessment, credit decisions
Prior conformity assessment required, logging obligation
Limited
Customer-facing chatbots, AI-generated texts
Transparency obligation: users must know that AI is in use
Minimal
Text assistance, internal research, summarising, spam filters
No special obligations. This is where most SME pilots land.
Compliance checklist (5 points, 80% covered)
AI policy in writing
Usage rules for AI tools: who may do what, with which data
Tool whitelist documented
Which tools have been reviewed and approved for data protection
Training records maintained
Date, content, participant list per Art. 4 EU AI Act
HR processes reviewed
No AI tool in hiring or dismissal without human-in-the-loop
Transparency notice in place
Users are informed when AI content is in use
In the Champion Pilot: The Champion Pilot directly addresses point 2 (tool whitelist) and point 3 (training records). Point 1 (AI policy) emerges as a side artefact.
Client confidentiality & on-prem: For law firms, tax advisors, and other professionals bound by secrecy, § 9 RAO and comparable confidentiality duties remain intact. On request, everything runs self-hosted on your own infrastructure (optional, surcharge), so client data never leaves your house. The full data-residency matrix is on the Implementation page.
Not a substitute for legal advice. The checklist provides a practice-oriented overview. Legal binding force is assessed by a lawyer.
DSGVO, hosting and transparent pricing
Where the tools run, which legal framework applies and what each format costs · expand for the full picture.
Price corridor
Audit from EUR 1,500 · Sparring from EUR 450 · Champion-Pilot EUR 3,000–5,000 · Custom build five figures+
Where does your data run? Transparency before compliance.
I maintain the full provider and hosting matrix (OpenAI Enterprise, Azure EU, Anthropic, on-prem) centrally on the implementation page.
View the hosting matrixShow detailed service scope and pricing
Audit, Sparring, Champion Pilot - clearly defined
Three formats. Three clearly defined scopes. No grey areas.
| Format | Price | Included | Not included |
|---|---|---|---|
| Audit | 1 day — from 1,500 EUR | AI Readiness Report (PDF, 15-20 pages), 30-minute results review, Concrete recommendations, Prioritisation matrix | Implementation or code, Coaching or follow-up sessions, Production code |
| Sparring | 90 min — from 450 EUR · Book a slot (Cal.eu) | 1 question worked through in depth, Architecture sketch or decision framework, Clear recommendation with reasoning, Short summary by email | Code or technical implementation, Rollout planning or project management, Follow-up coaching included |
| Champion Pilot | 6 weeks — 3,000-5,000 EUR | 1 trained AI champion within your team, Guided W1-W6 programme (2-4 hrs/week), ROI calculation with concrete metrics, Handover documentation and playbook | Custom software development (see /build), External rollouts to other sites, Ongoing operations support after pilot |
All prices net, excl. VAT. Exact quote after scope conversation.
Ready for the first step?
30-minute initial consultation, free of charge. We go through your situation and clarify whether a Champion Pilot fits.