Set in Tampere during the Finnish summer, the three-day event brought together researchers, practitioners, industry leaders, and students to explore how generative AI is changing software engineering and organisational work. The summer school combined keynotes, technical talks, hands-on sessions, industry demonstrations, and networking opportunities, creating an active space for discussing both the possibilities and the responsibilities of AI.
One of the strongest aspects of GAISE 2026 was its practical orientation. The event was not only about what AI could become in the future, but also about how it is already being used, tested, governed, and questioned in real organisations.
Alongside the talks and keynotes, the programme included several workshops, hands-on labs, demonstrations, and show-and-tell sessions, which made the summer school feel less like a sequence of presentations and more like an active working space.
Participants could engage with topics such as:
- Agent-first IDEs
- AI-assisted ideation
- Jailbreaking LLMs
- Agentic workflows
- Agentic RAG
- Small language models
- Practical research services
This workshop format added real value, as it demonstrated how these ideas work in practice.
Day 1: Autonomous AI Systems and New Forms of Debt
The summer school opened with the theme of autonomous AI systems, setting the tone for many of the discussions that followed.
The keynote "When Software Stops Waiting: Life in Autonomous AI Systems" highlighted how autonomous agents are becoming more visible in software systems, while also reminding participants that human judgement remains essential.
One beautifully crafted observation was that tokens are the new currency. Architecture, prioritisation, responsibility, and decision-making cannot simply be handed over to automation.
Markus Borg’s keynote, "Agents Code, Teams Erode – From Technical to Cognitive Debt", added another important layer to this discussion.
His talk captured both the excitement and the danger of AI-assisted development. AI may help teams produce software faster, but it can also increase technical debt if teams stop paying attention to maintainability, ownership, and understanding.
A particularly useful point was the idea that teams now need to manage not only technical debt, but also:
- Technical debt
- Intent debt
- Cognitive debt
In other words, it is not enough for the code to work. Teams also need to preserve the reasoning, goals, constraints, and shared understanding behind it.
The message was practical rather than pessimistic:
- Use agentic guardrails
- Monitor churn and code bloat
- Maintain human oversight
- Treat understanding as something worth measuring
It was a timely reminder that AI can write code quickly, but whether future teams can live with that code is another matter entirely.
Day 2: The AI-Native World
The second day moved into the idea of the AI-native world and explored how organisations and software teams may operate when AI becomes more deeply embedded in everyday work.
The programme included demonstrations, workshops, hands-on labs, and show-and-tell sessions, reinforcing the practical nature of the event.
Architecting in the AI Era
Dr. Muhammad Waseem’s session, "Architecting in the AI Era", was particularly interesting, as it demonstrated software architecture through a product involving 18 agents.
This raised important questions about the future role of software architects and how coordinated agentic systems might support architectural reasoning, design decisions, and development workflows.
AI Won’t Save Your Research
Dr. Kai-Kristian Kemell’s session, "AI Won’t Save Your Research – Things Researchers Should Never Do", brought a different but equally important perspective.
The session highlighted that AI can support:
- Reading
- Writing
- Exploring ideas
However, it should not replace:
- Critical thinking
- Careful citation practices
- Human review
The session specifically warned against using AI to review papers and highlighted the risks of relying on LLMs to find academic sources, as they often fail to provide real and reliable citations.
Free-tier tools were also presented as especially limited for serious research work.
Executive Perspectives
The day also included an executive track for industry participants, helping organisations think more strategically about how to steer AI adoption rather than simply reacting to technological change.
The day concluded with the gala dinner at Harald Viking Restaurant, adding a more informal and social dimension to the summer school.
Tampere’s summer also made a dramatic appearance that evening—the rain arrived and reminded everyone that no Finnish summer programme is complete without a weather-related plot twist.
Day 3: Governance, Knowledge, and Responsibility
By the final day, many of the earlier conversations started coming together around responsibility, production readiness, governance, and organisational impact.
Building Production-Grade AI Systems
Sami Lahti from Koivu Solutions delivered an engaging session on building production-grade AI systems, with a strong focus on coding and the role of skills in generating production-ready AI code from day one.
Preserving Organisational Memory
Another session, "Preserving Organisational Memory in the Era of AI Agents", addressed the challenge of ageing workforces and the loss of tacit knowledge when experienced employees leave organisations.
The presentation explored how AI agents could support:
- Knowledge transfer
- Organisational memory preservation
- Faster onboarding
It also introduced possible KPIs such as:
- Successor Readiness Score
- Knowledge Concentration Map
- Dependency-on-Individuals Index
- Cross-Country Divergence
These metrics offered new ways to think about organisational intelligence beyond traditional documentation.
Engineering Trustworthy AI in Europe
Salla Westerstrand’s session, "Engineering Trustworthy AI in Europe: Compliance & Responsibility", provided a fitting conclusion to the event.
After three days of discussions on agents, software architecture, coding, organisational change, and AI adoption, this session brought the focus back to ethics and governance.
A particularly memorable framework was:
EU AI Act = Minimum
Governance = Medium
Ethics = Direction
This was a powerful way to close the loop.
Compliance is necessary. Governance helps organisations put principles into practice. But ethics gives AI development its broader purpose and direction.
Final Reflections
Overall, GAISE 2026 was an intense, engaging, and memorable summer school.
Across three days, it offered a rich view of how generative AI is reshaping:
- Software engineering
- Research
- Organisational knowledge
- Industry practice
Beyond the talks and demonstrations, one of the most valuable aspects of the event was the opportunity to meet people working with AI from different perspectives:
- Researchers
- Students
- Practitioners
- Industry leaders
- Organisers
With so many sessions running across the academic, industry, and executive tracks, the communications team did a brilliant job of keeping the energy up and gently nudging everyone through a programme that was rich, packed, and just the right side of overwhelming.
GAISE 2026 would not have been successful without the organising team, who brought together speakers from multiple countries and participants from across Europe while keeping such a packed three-track programme running smoothly and with remarkable energy.
What's Next?
I still wonder:
What’s next?
(XP2027 and GAISE27 together? 😉)
About the Author
