AI Builders

Chairpersons welcome and opening remarks

Discuss How Search and Context Layers work in BigTech Ecosystems. How investments are made in each layer, and pros and cons of each layer of investments. Finally, discuss how Data is Monetized in Big-Data Systems.

Every centralisation programme looks beautiful on paper. One platform. One governed store. One diagram that fits on a slide.

It was a reasonable bet, and consolidation solved real problems. But the diagram describes a company that doesn’t exist. Acquisitions sit half-integrated, subsidiaries run their own stacks, and sovereignty rules and cloud costs are pushing workloads back on premise. So the migration runs for years and is overtaken by the next platform decision before it lands.
Now AI has raised the stakes. To reason across your business it needs all your data and all your context with it. So a migration that never finished just got bigger.

But what if data and context could be queried and governed where they already live, across cloud and on premise, without migrating first? This session explores how you can leverage your existing architecture to start driving AI value without waiting for the migration to finish, and where we are taking it next for context.

Most agent governance focuses on access and approvals. The harder question is authority: what counts as a decision, and which should an agent own? In this session, Karen Lawson thinks about how to split consequential decisions between deterministic systems, AI agents, and people, and considers when the “human in the loop” becomes a rubber stamp. She introduces the 5 A’s (Authority, Access, Approval, Accountability, Adaptation), how to build authority into agent architecture, and what GDPR, the EU AI Act, and US state laws expect. Attendees leave with a test for when more autonomy is the wrong design.

Every executive in the room has been told some version of the same story: AI is moving fast and the organizations that hesitate will be left behind. It is a compelling story, but it is also incomplete. In this keynote, Marinela Profi — global strategy lead for agentic AI, a TED speaker and Human & AI Trust Expert — challenges the assumption that speed is the defining competitive variable in the agentic era, and makes the case for what actually separates the organizations that will win from those that will spend the next three years cleaning up after themselves. This is a talk about the four traps most AI strategies don’t see coming — because each one looks exactly like the right move. The ones that determine whether agentic AI becomes a durable advantage or an expensive lesson.

With an ever-growing ecosystem of data and AI technologies, selecting the right platform strategy is essential to avoid unnecessary complexity and technical debt. This panel explores how organisations can balance flexibility, scalability and cost when deciding whether to build, buy or combine solutions. Learn how to prevent vendor sprawl, avoid over-engineering and align platform investments with business maturity, ensuring teams remain productive while creating a foundation that can evolve with future data and AI ambitions.

Many AI agents now act on documents: contracts, invoices, onboarding packs. To make sure those agents decide safely, enterprise technical teams and their IT partners add checks around them, such as tracing, evaluations, permissions and human approval. Often, these checks run after the document is read. If a document is misread, the agent still gets clean, believable data, and the checks have nothing to flag. For an enterprise, that could mean decisions based on misread documents that no one checked.

anyformat’s CTO, Diego Pérez, shows how to add a reading layer before those checks. It reads multi-page tables and poor scans, linking each value to its source page with a confidence signal. Validation rules compare documents, and when they don’t match, a person reviews the mismatch before the agent acts.

Keep building AI agents, but give them a reading layer to keep them safe.

Boomi has been connecting applications and data for two decades, and increasingly that means connecting AI into the mix too. Agentstudio lets teams build their own agents. Boomi Companion lets them build integrations just by describing what they want. AI is already part of how the work gets done, and not only through the tools we ship. Developers are building their own agents and MCP servers, and often nobody is watching which applications and data they can reach. This session looks at where that gap actually shows up, and what it takes to close it without slowing AI builders down.