2026 Agenda
More speakers and sessions will be announced soon. Check back for updates.
The gap between a successful AI pilot and a production system isn’t a model problem. It’s a context problem.
Practitioners from WPP, Chevron, Anthropic, Accenture, and OpenHands discuss what breaks when agentic AI meets real enterprise data. Learn how semantic consistency, deterministic logic, and shared business context are creating the foundation for durable data architectures.
The terms overlap, but the architectures don’t. Juan Sequeda, Sanjeev Mohan, and Jessica Talisman join Dave Mariani to untangle knowledge graphs, ontologies, and semantic layers. Learn why getting the distinction right is foundational to AI that actually reasons correctly over enterprise data.
AI agents don’t hallucinate because the models are bad. They hallucinate because the data has no shared meaning. When agents query raw tables directly, “revenue” becomes whatever the context implies. The semantic layer fixes that: enforcing consistent metric definitions, applying security policy before any query executes, and making every agent response auditable after the fact. Daniel Gray, VP of Solution Engineering at AtScale, shows what that infrastructure looks like in production across Snowflake, Databricks, and BigQuery.
Carrefour France bypassed the traditional pilot phase, migrating its full enterprise scope of 3,000 KPIs and 1,000 users to a universal semantic layer on BigQuery. By focusing on query-cost optimization and co-developing a custom Google Sheets connector, they matched the efficiency of legacy systems while gaining modern flexibility. This session explores the technical rigour behind this “Day 1” scale and how it now powers Carrefour’s production-ready conversational AI agents for Finance and Merchandising.
18 months ago, the Global BI team at Blue Yonder made a critical decision: they would no longer be a BI team. They consolidated their analytics organization under a single data infrastructure charter, are rebuilding 800+ tables into a governed dimensional semantic layer powered by AtScale, and are wiring that semantic layer into AI tools through AtScale’s MCP server. In this session, Blue Yonder shares the team transformation, the architectural rebuild, and a real example where a multi-day analyst workflow collapsed into a single AI query. All grounded in semantic-first infrastructure.
Thousands of franchise locations. One legacy cube. Countless conflicting reports.
Papa Johns retired a fragile OLAP environment and unified analytics across Excel and Tableau with a single semantic layer. Franchisees, finance, and corporate now see the same KPIs everywhere: comps, delivery times, customer satisfaction. This session covers how governed semantics drive trust, adoption, and faster decisions across thousands of locations, without sacrificing operator-level data privacy.
What does a semantic layer look like in production across three different enterprises?
Slickdeals scaled analytics across finance, product, operations, and sales, serving consistent metrics into both Tableau and Excel without sacrificing domain relevance. TELUS built a centralized semantic layer using SML as the authoritative repository for KPIs, powering everything from executive reporting to engineering root cause analysis. Vodafone Portugal used AtScale to retire a legacy on-premises OLAP environment and migrate to a governed, cloud-native analytics architecture, improving performance, reducing cost, and enabling self-service access across the business.
One consistent finding: a semantic layer is the foundation that makes analytics trustworthy, portable, and scalable across the enterprise.
Semantic fragmentation happens when business logic is embedded differently across platforms, BI tools, and AI systems. It’s one of the most underestimated blockers to reliable enterprise AI.
The Open Semantic Interchange (OSI) addresses this directly: an open source, vendor-agnostic specification for how semantic metadata is defined and shared across the stack. With the specification finalized and ecosystem participation expanding, this panel shifts the conversation from “why open semantics” to “what does it actually take to make this work in production.”
