Global Data Management Community

While AI is a strategic priority for most organizations, few have successfully scaled it or demonstrated measurable returns. This presentation explores a client case study that illustrates true meaning of Scaling AI in the Enterprise, built on shared data, governance, workflows, and a single user interface.

Every enterprise can get the same models. Agents are surging into production. And in every company racing to keep up, one person decides whether that AI can be trusted: the person who owns its data, its meaning, and its permissions. That's you. This is the moment your work stops being back office and becomes the backbone of your company's entire AI program. Join Collibra for a keynote on what to do today, this quarter, and this year to put yourself at the center of enterprise AI, and get a first look at the breakthrough capabilities that make governed context and control the foundation everything else builds on.

How data domains, business ownership, metadata and logical models help turn enterprise data into trusted assets for humans and AI agents.

You have your agents on record, but the model deciding on their behalf sits in no catalogue. You don't know what data it was built on, who governs its behaviour, who can switch it off, or where it runs when you hand it access to production. I'll show why a model belongs under governance like any other data actor, why you need an alternative to a single vendor, why you should benchmark on your own data rather than public leaderboards, and what is genuinely available today — from frontier APIs to sovereign deployment in the Czech Republic. You'll leave with an architecture in which swapping a model is a configuration change, not a project.

How do you prepare enterprise data so that AI agents can work with it reliably during software development? We will demonstrate how to generate a relevant knowledge base, unify data across systems, and securely grant access to production data. Everything will be complemented by concrete real-world examples from both new and long-term legacy projects.

Every organisation claims its data is "AI-ready". Few can say what that means beyond "clean and governed". This talk defines it in three steps. First, the theory: what an agent needs from data that a human analyst never asked for, and why a good catalogue falls short. Second, an agent team at Raiffeisenbank, setting up authorised persons' access rights in corporate banking across 140 protocol templates. Third, the demonstration: how we turned a 46-page prompt into graphs. A glossary, a concept model, and selection rules agents can read and business can verify. And what it cost and where it failed.

Data Governance and AI Governance are often developed separately, reflecting different objects, regulations, roles and levels of maturity. Conversational AI and emerging agents increasingly challenge this separation. They depend on governed data and metadata, mediate access to organisational knowledge, generate governance artefacts and may act across systems and processes. Their influence is already visible in data stewardship: many activities are supported through drafting, summarisation, retrieval and semantic improvement, while tasks such as lineage documentation, schema governance, data classification and regulatory monitoring are beginning to change more substantially. These developments shift stewardship from producing artefacts towards validating outputs, managing uncertainty and coordinating accountability. Although specialist governance expertise remains necessary, conversational AI strengthens the case for operational integration through shared lineage, lifecycle management, roles and clearly connected decision rights.

As AI evolves from answering questions to taking actions, data must evolve as well. This session explores what makes data "agentic-ready" and why discoverability, context, governance and trust are becoming essential foundations for autonomous AI systems. Learn how regulated organizations can build trusted agents through trusted data while balancing innovation, risk and regulatory expectations.

Most organizations run data governance and AI governance as separate programs with separate sponsors, separate councils, and separate backlogs. Then they wonder why AI use cases stall on data quality and why governance work never gets funded. This talk is a case study from Eurowag, a FTSE 250 fintech serving the European commercial road transport industry, which made a different structural bet: the Data Governance Council reports into the AI Steering Committee. One executive-sponsored body decides AI use-case intake, data product funding, and Data Governance together. The result changes the conversation in both directions. AI ambition creates the business urgency that data governance always lacked, and governance maturity becomes a visible gate on AI delivery instead of a parallel hobby. We will cover the operating model, what it took to get executives to fund data products as standalone investments, where the structure creates tension, and why this integration is the prerequisite for governing agents: an agent is just an AI use case with write access, and it will be governed by whichever body you build today.

Thanks in part to conferences like DAMA, we’ve all managed to build great data foundations with beautifully defined semantics, right 😉? But what if you wanted to make life as difficult as possible for your human users and AI agents? Let them earn their salaries and tokens! In this session, we’ll walk you through several battle-tested patterns, complete with practical tips and tricks on how to destroy your data governance. P.S.: There is a slight risk that if someone takes our presentation as a cautionary tale, they might accidentally build a highly effective, user- and AI-friendly system. Please note that we accept no responsibility for such unintended consequences.😜

Nothing should reach production because of AI hype or fear of missing out. It should reach production because it meets the conditions every production system has to meet: it acts under a known identity, it touches only the data it is entitled to, its behavior is evaluated rather than assumed, it is monitored, and everything it did can be reconstructed afterwards. Agents are no exception - access management, behavior evaluation, monitoring and auditability are not features you add once the agent works, they are what makes running it permissible at all. Drawing on almost two years of building enterprise-grade agentic platforms in banking, this talk shows how we built those conditions into the platform layer, so that individual agents inherit them by default rather than each team re-solving them.

We run AI agents across half a dozen providers, and the first question anyone asks is: What is this actually costing us? The answer is usually, "Well, it depends.". I will talk about how we're building a unified view of AI spend across providers - what it caught, what it unlocked, and the pushback that came with making spend visible in the first place.

Governed data plus a governed agent still does not add up to coherent action, because an agent can only reason across the context the enterprise makes available to it. Where that context is fragmented, AI does not remove incoherence — it operationalises it at machine speed, and a locally valid decision becomes a globally wrong one. I'll show what connected enterprise context looks like: an evidence-backed model of relationships across purpose, process, information, authority and accountability. You'll leave with the structural condition that sits between trusted data and trusted action — coherence.

Introduced and moderated by Mirek Umlauf (FIS VŠE). Ivan Merta (MSD), David Jez (E.ON), and Petr Pleticha (Allwyn ČR) on what changes when agents get write access to production data — where their organizations stand today, and what they would (and wouldn't) let an agent do.