Keynote Handout
One Champion per Area
The underestimated lever for AI adoption
In 10 minutes: Are you ready for an AI champion?
This self-assessment shows you in 10 minutes where your organisation stands — not as criticism, but as orientation before the first concrete step. Knowing your starting point saves 4 to 8 weeks of ramp-up time in a pilot.
Award 0–3 points per question.
Do I have someone who understands both IT and the business side at the same time?
Are my employees already using AI tools?
Is our position on AI tools documented in writing?
Is there a concrete process with measurable time expenditure?
Is there financial room for a pilot (€3,000–€8,000)?
Does a manager have at least 2 hours per month for the champion plus the champion's own time?
Is there a list of GDPR-compliant AI tools?
Is the documentation obligation fulfilled?
Evaluation
Total points: ___ / 24
| Punkte | Stufe | Empfehlung |
|---|---|---|
| 0–8 | Build the foundation | Start with policy and tool whitelist. Without this foundation a pilot wastes time and credibility. |
| 9–16 | Select a champion | Champion selection is your next step. The prerequisites are there — the person is still missing. |
| 17–24 | Start the pilot | Your pilot is ready. Process and framework are in place. Just start. |
What this looks like in 3 minutes
Many champions ask: "Where do I start if I have no developer skills?" The answer is smaller than you think. Six individual steps, each under 30 seconds. Combined, a real workflow.
ChatGPT Vision
Photo → Minutes
Upload a photo of a flipchart — get back a structured set of minutes with a task list.
Gemini 3 Flash
PDF stack → Summary
30-page specification condensed into core risks and deadlines in 5 seconds.
Claude Project
Project container
Add documents, set a system prompt, get answers with sources from the uploaded files.
Claude Artifact
CSV → Chart
Drop in a table — first analysis as a bar chart including interpretation.
ChatGPT Pulse
Morning briefing
Daily 5-minute summary as input for the team stand-up.
Claude
Email → Meeting
Forward an incoming email — three meeting proposals with context in seconds.
The chain: client project in three minutes
Each step on its own is trivial. Together they form a repeatable process. A real-world example:
- Create a project container "Client Project Müller GmbH", set system prompt: "Only answer using sources from the uploaded documents."
- Upload three specifications as a PDF stack — a summary of the core requirements appears automatically.
- Upload the price table as CSV — first calculation as a bar chart.
- Every new document added to the container gets connected with context. The champion maintains nothing — they work, and the system builds itself.
What this means
- No software investment — access to existing consumer apps.
- No IT project — the champion builds the process for their team themselves.
- Measurable from week one — the container grows, efficiency becomes visible.
- Scalable — what works for client projects also works for HR onboarding, supplier evaluation, and tender review.
What managers need to know since 02/2025
AI regulation has not been a future topic since 2 February 2025. Four areas affect you directly.
EU AI Act Art. 4 — Training obligation (since 02.02.2025)
What counts as training:
- Internal workshops with documented content and participant list
- External certifications (e.g. EU AI Act foundation course)
- Verifiable e-learning modules with completion confirmation
What does not suffice:
A Slack message, a mention at a company meeting, informal demonstrations.
At least once a year, recommended every 12 months, documented in writing. Rule of thumb: 2 hours of basic training per person per year as a minimum entry point.
Immediate action: create a training record template, maintain a participant list, note the date.
Risk classes — where your organisation stands
| Kategorie | Beispiele | Konsequenz |
|---|---|---|
| Prohibited | Social scoring, emotion recognition in the workplace, biometric mass surveillance | Prohibited from 02/2025 — no use possible |
| High-Risk | HR scoring (hiring, dismissal), educational assessment, credit decisions | Prior conformity assessment required, logging obligation |
| Limited Risk | Chatbots, AI-generated texts | Transparency obligation: users must know that AI is involved |
| Minimal Risk | Spam filters, product recommendations | No special obligations |
Most mid-market companies operate in the Limited Risk and Minimal Risk areas. Exception: anyone using AI for HR decisions immediately falls into High-Risk.
GDPR Art. 22 — No machine as the sole decision-maker
Decisions with legal consequences or significant impact on a person may not be made exclusively by automated means.
Examples with legal consequences:
Dismissal, rejection of a job application, credit refusal, insurance exclusion.
Human-in-the-loop suffices as a protection mechanism. A person reviews the AI recommendation before it takes effect. This review must be documentable — a click confirmation without substantive engagement does not suffice.
For every AI-supported process involving people: does a human make the final decision? Is that verifiable?
Organisational liability and sanctions
Austria: Analogous application of GmbH due diligence obligations (§ 43 GmbHG AT) — managing directors are personally liable for organisational deficiencies if known risks were not addressed.
EU AI Act sanctions:
- Up to €35 million or 7% of global annual turnover for violations of prohibitions
- Up to €15 million or 3% of turnover for other violations
- Micro-enterprises: the lower of the two amounts applies
GDPR Art. 22 sanctions:
Up to €20 million or 4% of global annual turnover for violations of the prohibition on automated individual decisions.
Compliance checklist (5 points)
- AI policy in writing — usage rules for AI tools, who may do what, with which data
- Tool whitelist documented — which tools have been GDPR-reviewed and approved
- Training records maintained — date, content, participants (Art. 4 EU AI Act)
- HR processes reviewed — no AI tool in hiring/dismissal without human-in-the-loop
- Transparency notice in place — users are informed when AI content is in use
3 pilot variants, 4–12 weeks
No pilot is the same without context. These three variants cover 90% of starting situations.
| Variant | Budget | Duration |
|---|---|---|
| Small | €3,000–€5,000 | 4 weeks |
| Medium | €5,000–€8,000 | 6 weeks |
| Large | €20,000–€50,000 | 3 months |
All budget figures: personal estimate from conversations and public discussions.
Variant: Small
€3,000–€5,000Suitable when: a clear process pain exists, champion candidate identified, but no AI experience in the company yet.
| Area | Share | Amount (mid) |
|---|---|---|
| Training + champion onboarding | 40–50% | €1,600–€2,000 |
| Tool licences + setup | 20–30% | €800–€1,200 |
| External support / workshop | 20–30% | €800–€1,200 |
Variant: Medium
€5,000–€8,000Suitable when: self-assessment 9–16 points, ready for structured build, management support committed.
| Area | Share | Amount (mid) |
|---|---|---|
| Training + champion development | 35–45% | €2,100–€3,000 |
| Tool licences + integration | 25–30% | €1,500–€2,000 |
| Support + playbook creation | 25–30% | €1,500–€2,000 |
Variant: Large
€20,000–€50,000Suitable when: self-assessment 17–24 points, readiness for structural change, internal change management in place.
| Area | Share | Amount (mid) |
|---|---|---|
| Champion training (3–5 people) | 30–40% | €9,000–€17,000 |
| Tool rollout + whitelist build | 20–25% | €7,000–€10,000 |
| Governance, policy, compliance | 20–25% | €7,000–€10,000 |
| Ongoing external support | 15–20% | €5,000–€8,000 |
Success measurement
Two metrics are sufficient for pilot proof:
1. Time saved per champion per week
Before pilot: measure hours for target process per week. After pilot: repeat the same measurement. Target: at least 2–4 hours saved per week per champion.
2. Error rate before/after
Choose a process with a measurable error type. Record the baseline rate before the pilot. Compare after 4–6 weeks.
These two numbers are sufficient for an internal decision proposal for scaling.
The bottleneck is rarely the money. It is champion time and context quality.
Your first step in 72 hours
You don't need a 6-month roadmap. Three options — depending on where you stand today.
30-minute consultation (free)
Was:
We go through your self-assessment together. You get a clear assessment: where do you stand? What is your realistic next step?
Für wen:
If you are still unsure after the talk whether a pilot suits you.
No pitch. No proposal if it doesn't make sense.
Pilot kickoff workshop (from €1,500)
Was:
2 hours on site. Joint analysis of your strongest process pain point. Written recommendation with a concrete next step, champion profile and budget estimate.
Für wen:
If your self-assessment shows 9–16 points and you want to leave the meeting with a clear plan.
Ergebnis: 1–2 pages written recommendation you can use internally as a decision basis.
Guided champion build (pilot small–medium)
Was:
4–6 weeks structured build. Identify champion, train, implement first process, document quick win.
Für wen:
If your self-assessment shows 17–24 points and you want to start immediately.
Investition: €3,000–€8,000 (personal estimate from conversations)
Contact
- Website
- gotzendorfer.at
- linkedin.com/in/bg05
On the website you will find the BitGN PAC Hackathon case (1st place, Vienna, April 2026) and the BuchhaltGenie case (+71% OCR accuracy).
About the author
Bernhard Götzendorfer, 10 years IT experience (since apprenticeship 2015), 18 months intensive AI practice, 100+ own AI prototypes, 4,000+ development sessions. IT project manager branch network at Austrian Post (01/2024–02/2026), previously IT & e-commerce project manager at Europapier Austria (11 countries, 16 websites), managing director RocketWM GmbH (400+ websites, team of 4). IPMA-certified (2025). 1st place BitGN PAC Hackathon Vienna, 11 April 2026 — 79 out of 104 points, solo, 3 hours on site. EU AI Act certified.