Composable Analytics & the Future of the Semantic Layer: Building for the Age of AI Agents

Business intelligence has always been good at telling you what happened. It's never been great at telling you what to do about it. In this episode, AtScale's CEO and co-founder Dave Mariani sits down with Donald Farmer, Principal of Treehive Strategy, to talk about why AI is finally closing that gap.You'll get Donald's take on:How AI will start to take over the exploratory, discovery side of analyticsWhy "revenue" means something different to marketing, sales, and finance ,  and why semantic layers need to model relationships between data, not just definitionsHow treating semantic models like code (Git, CI/CD, AI-generated YAML) lets business users build them without needing to be a "unicorn"What Apache Ossie tells us about the industry's shared need for semantic standardizationDonald's advice for enterprises and software vendors trying to figure out their AI strategy right nowKey topics in this episode:The future of BI: routine dashboards vs. AI-driven discoveryMarcia Bates' model of "browsing" (orientation, glimpsing, examination, acquisition) and where BI always fell shortSemantic layers as the foundation for AI trust and governanceAI as translator, not just code generator, for building semantic modelsVendor lock-in, Apache Ossie, and Apache Iceberg as pieces of an open, portfolio-based data stackDonald's "go carefully, but go" advice for AI adoption

What happens when enterprise AI meets inconsistent metrics, fragmented dashboards, and conflicting business logic?In this episode of the Data-Driven Podcast, AtScale CTO and co-founder Dave Mariani sits down with Brad Lindsey and Jeremy Arendt from Blue Yonder to discuss how Blue Yonder transformed its analytics strategy from disconnected dashboards into a governed semantic layer foundation for AI and enterprise analytics.The conversation explores why semantic layers have become critical infrastructure for AI, how governed metrics enable trusted self-service analytics, and why enterprises must standardize business definitions before deploying AI agents at scale.Key topics include:Why Blue Yonder shifted from dashboard development to data infrastructureBuilding a universal semantic layer for AI, BI, Excel, and LLMsHow semantic models eliminate inconsistent metrics across the businessWhy semantic governance matters for agentic AIThe role of Model Context Protocol (MCP) and semantic context in enterprise AICreating reusable governed business logic for analytics and AIHow Blue Yonder reduced analysis work from 30 hours to 90 seconds using semantic models and AIScaling trusted self-service analytics without losing governanceThe future of semantic layers as operational infrastructure for AIThe discussion also highlights a major shift happening across enterprise data architecture: semantic layers are no longer just BI tooling. They are becoming the governed operational foundation for AI-powered decision making.Learn how Blue Yonder is preparing for a future where AI agents, dashboards, copilots, and analytics workflows all operate from the same trusted semantic foundation.

What does enterprise AI actually need to succeed: more data access or better context?In this episode of the Data-Driven Podcast, AtScale CTO Dave Mariani sits down with Coginiti CTO Matthew Mullins to unpack one of the most important debates in modern data architecture: access vs. understanding. The conversation explores the rise of the semantic layer, the role of open semantics, and why initiatives like the Open Semantic Interchange (OSI) matter for the future of AI and analytics.Matthew shares his journey from cognitive science and formal semantics into enterprise data, and how those foundations are becoming critical again in the age of AI. Together, Dave and Matthew break down:Why inconsistent metrics (like “29 definitions of customer”) still plague enterprisesThe real purpose of OSI and whether it should be an interchange format or a modeling standardWhy semantic layers are becoming core infrastructure for agentic AI and LLMsHow the Model Context Protocol (MCP) fits into the stack, and what it doesn’t solveWhy AI cannot replace human-defined business semanticsThe shift from dashboards to AI-driven, ad hoc business intelligenceThe key takeaway: LLMs don’t understand your data; they understand language about your data. Without a semantic layer to provide governed definitions, relationships, and context, AI systems will produce inconsistent or unreliable results.If you're building AI-powered analytics, designing data platforms, or evaluating open standards like OSI, this conversation outlines where the industry is heading.Learn more about AtScale’s universal semantic layer: https://www.atscale.com

In this episode of the AtScale Data-Driven Podcast, Dave Mariani sits down with Juan Sequeda, Principal Researcher at ServiceNow and former Head of the AI Lab at data.world, to explore what’s next for analytics, AI, and trust in 2026.As AI becomes embedded across the enterprise, one issue is emerging as a hard blocker: lack of context. Large Language Models are powerful, but without semantic layers, governed metadata, and shared business definitions, they struggle to produce reliable, explainable answers.Juan and Dave discuss why metadata is no longer “documentation,” but operational infrastructure, and how semantic layers are becoming the foundation for trusted AI systems. They break down why dashboards alone can’t deliver prescriptive analytics, how AI agents are shifting analytics from insights to action, and why enterprises are consolidating around platforms that can govern context at scale.The conversation also covers:Why AI and LLMs fail without semantic contextThe rise of semantic layers as enterprise AI infrastructureHow metadata, knowledge graphs, and governance convergeWhat AI means for dashboards, agents, and analytics workflowsThe future of tech jobs, systems thinking, and human skillsIf you’re a data leader, analytics architect, or AI practitioner trying to understand how to make AI trustworthy in production, this episode explains why semantics, not models, are the real bottleneck.🔗 Learn more about AtScale’s semantic layer and Model Context Protocol (MCP):https://www.atscale.com🔔 Subscribe for more conversations on AI, analytics, and data infrastructure.

Large Language Models (LLMs) are transforming analytics, but only when paired with the right foundation. In this episode, AtScale’s CTO and co-founder Dave Mariani, co-founder Dianne Wood, and product director Petar Staykov show how Model Context Protocol (MCP) and the semantic layer unlock trusted, governed, enterprise-ready AI.You’ll get a full end-to-end walkthrough of Claude using AtScale’s MCP server to:Discover available semantic models via list-models tools Understand metrics, attributes, and dimensional context using describe-model Run fully governed analytical queries, without writing SQL, through run-queryGenerate insights, summarize trends, and even build dashboards autonomously Learn why LLMs struggle without structured business logic and how semantic layers eliminate hallucinations and governance drift. You’ll also learn how AI will soon help build semantic models, not just query them—enabling faster metric definitions, autonomous metadata generation, and an “army of agents” that maintain semantic consistency as businesses evolve. Key topics in this episode:• Why MCP is the JDBC-moment for LLMs• How semantic layers make AI deterministic, trustworthy, and governed• Real demos of LLM insights far beyond simple Q&A• Automating semantic model creation using agents and SML• The future: AI copilots for BI, analytics, and semantic modeling

Dave Mariani and Jens Kröhnert explore the "AI Frontier:” how semantic layers enable AI agents to operate autonomously without hallucinations or governance drift. Both veterans of scaling Microsoft BI stacks on Hadoop, they discuss why machines need semantic context even more than humans do.Topics include the AI Frontier operational model from Microsoft Research, automating semantic model development with Oraylis's Datamate tool, and Model Context Protocol (MCP) as universal infrastructure for LLMs. Dave demonstrates pairing Claude with AtScale's semantic layer to generate insights, identify sales patterns, and design campaigns. All grounded in governed metrics. Independent benchmarks show semantic layers improve LLM accuracy by 4-5x.Essential for data leaders preparing for agentic AI and generative BI at enterprise scale.Learn more: atscale.com

In this episode of AtScale’s Data-Driven Podcast, host Dave Mariani, Co-Founder & CTO at AtScale, is joined by Petar Staykov, Product Director, and Daniel Gray, VP of Solutions Engineering, for a deep dive into composable analytics—the architectural shift that’s transforming how enterprises model, govern, and scale data in the era of AI agents.Key TakeawayComposable analytics isn’t just a design pattern—it’s the foundation for governed, scalable, and AI-ready data ecosystems. Learn how AtScale’s Universal Semantic Layer and open-source Semantic Modeling Language (SML) are redefining interoperability across BI, AI, and agent frameworks.

In this episode of the Data-Driven Podcast, AtScale CTO and Co-founder Dave Mariani is joined by Rich Williams, SVP of Partnerships and Strategy at Hexaware, and Prashant Dahalkar, VP of the Cloud Data Practice. They dive into the future of analytics through the lens of agentic AI, semantic layers, and enterprise AI governance.Fresh from attending Snowflake and Databricks Summits, the trio unpacks the momentum behind AI agents, the rise of semantic standards, and how organizations can accelerate business outcomes with governed data architectures. Rich offers insight into how modern business users can finally talk to their data. At the same time, Prashant highlights the role of modular agents, MDM, and self-healing data tools in Hexaware’s data strategy.Whether you’re modernizing legacy stacks or scaling data literacy across your organization, this conversation sheds light on how enterprises can move from experimentation to enterprise-grade AI adoption without sacrificing trust or governance.

In this episode of the Data Driven Podcast, Dave Mariani, CTO and Co-Founder of AtScale, sits down with Dale Williamson, EMEA CTO at Databricks, to explore the transformative role of open source standards and semantic layers in modern analytics. They discuss the importance of interoperability in data infrastructure, the impact of generative AI on democratizing analytics, and how the right semantic layer can bridge complex data to natural language for everyone. Together, AtScale and Databricks envision a future where data insights are accessible to all, empowering organizations to build with flexibility, trust, and freedom from vendor lock-in.

In this episode, AtScale CTO and co-founder Dave Mariani chats with Greg Deckler, Vice President at a global IT services firm and a seven-time Microsoft MVP. They discuss Microsoft’s strategy with Power BI and its integration with Microsoft Fabric, with Deckler critiquing the rushed release and its early limitations. Both express skepticism about business users adopting more complex tools, citing ongoing challenges like measuring totals.They also discuss the impact of generative AI on business intelligence tools like Power BI, urging caution about data security and governance. They also highlight the underutilization of natural language queries in Power BI despite frequent demos. Deckler speculates the future of Power BI’s support for various data platforms and Microsoft’s strategy to integrate these with Fabric.To learn more, read Deckler’s 3 part blog series, “Is Fabric and Direct Lake a Game Changer For Microsoft?”