Kurt Cotoaga

AI strategy that survives in operations

The situation:

Companies buy ChatGPT licenses, send teams to 60-minute workshops — and are still clueless when the EU AI Act comes into effect. Pilots get lost. Governance is missing. Knowledge remains with individuals.

What I do:

EU AI Act Readiness: Risk classification, documentation, literacy programs — pragmatic, not legally inflated.

AI implementation that works on Monday — not just in the demo slot

Expanding mental models: Empowering teams to independently evaluate and use AI tools.

Cleaning up after failed AI pilots — diagnosing what was really missing

Cognitive Sovereignty: Empowering organizations instead of making them dependent.

Result:

→ Diagnosis first, then therapy — not the other way around

→ Teams learn it themselves. I make myself redundant.

→ Fixed price. No FTE graveyards.

Tools: Claude, ChatGPT, Gemini, Perplexity, Grok, MCP, VectorDBs, LangChain, Prompt & Context Engineering, EU AI Act Framework

Almost failed SAFe ART with 250 employees (distributed, remote, multiple service providers). PI Planning confidence at 2.5 out of 5. ASPICE audit was imminent. Shared services between ARTs were being crushed. No functioning CI/CD pipeline.

- Connected the portfolio with the teams (both directions – not just top-down)

- Restored a real end-to-end CI/CD pipeline

- Reactivated SAFe best practices AND aligned them with the ASPICE framework

- Optimized traditional contract structures for Lean-Agile

-Innovation:Introduced first AI-powered workflows for transparency (ChatGPT for PI Planning preparation, automated ASPICE documentation)

→ ASPICE audit passed (on the first attempt)

→ PI Planning confidence increased from 2.5 to 4.0 (+60%)

→ Restored confidence and motivation in the ART

→ Established business value focus (instead of pure delivery theater)

→ Positioned AI workflows as a competitive advantage

Tools:ChatGPT, Claude, Jira, Confluence, Structure, Miro, Mentimeter, Teams, BPMN 2.0, UML 2.5

REWE acquires Lekkerland. IT-Ops suddenly has to serve multiple domains according to the proprietary REWE model. Team overwhelmed. Stakeholders frustrated. No transparency. Scrum rejection in the team. 50 people (distributed, remote, multiple service providers).

- Identified value streams (which domains are served how?)

- Made work transparent via IT-Ops Kanban (no Scrum mandate)

- Found balance: Planned Work vs. Work-on-Work

-Liberating Structures Workshops:Team self-organized (no top-down directives)

- Open Space Events: Made progress visible throughout the entire headquarters

→ Cycle time -50% (without prior KPIs, but perceived halving)

→ Backlog size -50%

→ Team accepted self-chosen way of working (instead of mandated Scrum)

→ Transparency established internally + externally

→ Scaling: Open Space Events throughout the entire headquarters

AI Leadership:Led Agility Competence Center into ChatGPT possibilities (2022-2023, before mainstream adoption)

Tools:ChatGPT, Jira, Confluence, ConceptBoard, Mentimeter, Teams