Roundtable III brings together everything we've learned from our previous sessions and grounds it in something concrete: working systems, real results, and lessons learned from the field. Alongside our roundtable discussion, our experts will walk you through a set of live use cases, so the conversation starts from a place of reality, not theory.

Together, we'll dig into the questions that matter most right now:

• AI Governance: How do we stay in control as our AI landscape scales?
• Agentic AI at Enterprise Level: What does it take to build and govern an agentic platform that works across your organisation?
• The Roles of Tomorrow: What skills and capabilities do you need to thrive in an agentic world?
• GenBI & SQL Agents: Is self-service analytics entering its next chapter?
• Buy vs. Build: How do you make the right call with (Gen)AI investments?
• Industry Use Cases: From AI-driven claims handling to transaction monitoring: what's working, and why?

A use case showcase
Before the discussion opens, Xomnia experts present real Agentic AI implementations. What was built, why it worked, and what didn't.

A focused roundtable
Facilitated by Fried Schölvinck, the discussion digs into the decisions that matter most for your organisation in 2026 and beyond.

A room worth being in
Alongside thought leaders and peers from your industry, we dig into the practical steps and strategic decisions that will define the AI-ready organisation in 2027. No slides, no sales pitches, just sharp minds working through real problems together.

Enterprise AI at scale
How we deployed a secure, company-wide GenAI platform to over 4,000 active users, saving at least 3,000 hours annually in customer service alone, while navigating the real complexity of legacy systems, stakeholder alignment, and measuring impact in uncharted territory

AI-powered agent assist
How we built real-time AI support for frontline call center agents, reducing handling times, automating call summaries, and scaling GenAI responsibly in a high-volume customer service environment.

Each showcase covers: the business challenge, the solution built, the lessons learned, and the measurable impact, giving you a concrete foundation for everything that follows in the roundtable.

🛡️ From smart to trustworthy: building AI systems you can rely on

The focus in AI development is shifting from intelligence to reliability. A smart model is useless if it's unpredictable.

Statement: The engineering challenge is no longer about writing the best prompt, it's about building the guardrails, permissions, and automated tests that surround the model.

• Developers now spend roughly 80% of their time working with AI and only 20% writing code themselves. Does that make your team faster. or does it create a dangerous gap in codebase comprehension?
• A coding agent can write a function brilliantly today and silently introduce fragile, poorly structured code tomorrow. What does a responsible deployment pipeline actually look like?
• How do you define and measure trustworthiness in an AI system, and who in your organisation owns that standard?

⚙️ From pilot to production: the operational reality

The smartest first move isn't a flashy customer-facing agent, it's automating a 'boring' internal process where it can fail safely.

Statement: The technology has moved from helpful to operational. The question is no longer whether agents can do the work, it's whether your organisation is ready to let them.

• What workflows in your organization are genuinely ready for autonomous AI, and which ones only feel ready?
• Traditional ROI frameworks are struggling to keep up with how fast the technology moves. Are you measuring time-to-value rather than financial return — and is that sustainable?
• What is the single most immediate, practical step your organization could take this quarter?

📊 Data quality: from prerequisite to active solution

In 2026, AI can help solve the very problem it suffers from, but only if you're honest about the state of your data.

Statement: Your agentic system is only as reliable as the data it operates on, and enterprise data is almost never as clean as you think.

• Enterprise knowledge bases accumulate contradictions over time,  outdated documentation, divergent updates, conflicting sources. When you deploy an agent on top of this, it doesn't just fail: it confidently presents incorrect information. How are you addressing this?
• Are you building internal evaluation datasets, and if not, how are you validating model performance on your specific workflows and terminology?
• How do you embed data quality management as an ongoing operational process rather than a one-time cleanup?

🏗️ Architecture: your company's AI brain

Everyone is focused on the LLM. But the model is a commodity, your real competitive advantage is a superior data and context platform.

Statement: The organizations winning with agentic AI aren't those with the smartest models. They're the ones with the cleanest context.

• How do you build your company's AI "brain", and who owns it?
• Agents now operate with modular "skills" that allow them to interact with specific tools, internal systems, and proprietary formats at runtime. How does this change your integration and architecture strategy?
• When do you implement MCP, and who defines the business rules that govern agent behaviour

🔒 Governance: The Red Button, Runaway Costs & the Multi-Agent Problem

A manual 'red button' is a comforting illusion. By the time you press it, an autonomous agent has already done the damage.

Statement: The governance challenge in 2026 isn't one agent making one bad decision. It's hundreds of agents making decisions at machine speed, collectively representing your organization.

• When agents operate across customer service, finance, operations, and software development simultaneously, how do you create a governance layer that maintains consistency without killing agility?
• Who in your organisation has the authority to approve, deploy, and terminate agents — and is that decision-making structure fast enough to keep up?
• How do you model and cap the costs of systems that can scale their own activity autonomously?

🔀 Platform consolidation vs. vendor proliferation: The new build vs. buy

Every major software vendor has now launched agentic AI capabilities. That makes the platform decision both more urgent and more complex.

Statement: For any serious European enterprise, defaulting to US platform ecosystems for core AI is an unacceptable long-term risk, but building everything yourself may be equally unrealistic.

• Microsoft, Google, Salesforce, SAP all are embedding agents into their existing products. Do you adopt each vendor's native solution, or build a unified orchestration layer across tools?
• What are the specific triggers cost, privacy, sovereignty, control that would push you toward self-hosting or custom orchestration?
• How do European compliance requirements, including GDPR and the EU AI Act, factor into your platform decisions?