Timbr.ai

Databricks has made serious investments in agentic AI. Mosaic AI, Agent Bricks, Genie Code. The tools to build, deploy, and orchestrate agents are all there. But every one of those agents still queries raw tables. And raw tables don't know what "revenue" means. They don't know which joins are valid, which metric definition is authoritative, or how your business concepts actually relate to each other. The agent generates SQL. It runs. The number looks reasonable. But it reasoned from schema, not from meaning. Databricks tells agents where data is and who can access it. Timbr tells agents what the data means, how business concepts relate to each other, and how to reason across your entire data estate, including the 60% that isn't in the Lakehouse. We just published a breakdown of where the gap actually lives, why Databricks' native AI tools need help, and what changes when agents operate with the context provided by an ontology-based semantic layer. 👉 https://lnkd.in/dMwbwGhG #Databricks #SemanticLayer #AIAgents #EnterpriseAI #NL2SQL #KnowledgeGraph #Ontology

When SQL ontologies make a difference...

Interested in Context Graphs? A well designed SQL ontology is the ideal context graph to drive AI and contextual intelligence. #ontologies #knowledgegraph #contextgraph #agenticai #llm #aiarchitecture

Claude in Excel is genuinely impressive. ⚡ But there’s a quiet problem underneath the excitement. When AI works inside Microsoft Excel, it doesn’t see your business. It sees a spreadsheet that has already lost its relationships, definitions, and rules in transit from the data warehouse. That’s not an AI limitation. It’s a data foundation problem. AI agents like Claude can reason, explain, and model beautifully if the data carries meaning. Most enterprise spreadsheets don’t. That's why the Gold Layer still isn't enough for AI. And why the Diamond Layer matters if you want Excel to be reliable, not just impressive. The teams moving fastest with AI aren’t chasing smarter agents. They’re fixing the foundation first. 🔗 Read the full breakdown here: Why Claude in Excel Needs Ontologies, And How Timbr Delivers Them - https://lnkd.in/d5vaRsEU #EnterpriseAI #DataStrategy #SemanticLayer #MicrosoftExcel #Ontology #DataEngineering #Claude

Most enterprise ontologies fail after initial success. Not because they were modeled incorrectly. But because they were architecturally isolated from the systems they were meant to govern. Here's what happens: The data warehouse schema changes. A migration runs. DBT models update. Everything stays aligned. Except the ontology. It sits in a separate triplestore, unchanged. Nobody on the data engineering team knows it exists. The knowledge engineer who maintains it doesn't see the schema migration PRs. Over time, "Active Contract" means one thing in the warehouse, another in the ontology, and something different in the BI layer. Nothing breaks. Queries still run. Dashboards still load. But Finance and Sales now show different numbers for the same metric. This is semantic divergence. And it's not a modeling problem. It's an operational design problem. Traditional ontology architectures treat semantics as separate from data systems. That separation creates a sync gap that widens over time. SQL-based, co-located ontologies solve this structurally. When the ontology lives in the same operational world as the data stack, it evolves using the same migration workflows, version control, and CI/CD practices teams already rely on. The ontology doesn't need manual synchronization. It changes in the same commit as the schema. 🔗 Read the full article: https://lnkd.in/dcXerswE #SemanticLayer #DataArchitecture #EnterpriseData #KnowledgeGraphs #DataEngineering #OntologyManagement #DataGovernance

Build governed enterprise AI data agents with Timbr ontologies. 🤖 Timbr now delivers native data agent creation and management, built directly into the platform. Most AI agent frameworks focus on how agents act - chat, workflows, tool orchestration. Timbr focuses on what agents know about your data. Timbr Data Agents are built on a governed semantic ontology that understands business entities, relationships, and metrics by design. How data agents differ from generic agent frameworks 👇 • Agents reason over business concepts (Customer, Order, SLA) - not raw tables or schemas • Data access is governed through a semantic layer, not prompt rules • Metrics and business logic are defined once and reused consistently • Queries are generated deterministically using SQL ontologies • Agents operate safely across warehouses, lakes, and operational systems Instead of asking an agent to figure out your data model, Timbr gives agents a clear, structured understanding of enterprise data from day one. This is how teams move from experimental copilots to production-grade data agents. 👀 Want to see how data agents work in practice? See how Timbr turns your data into an agent-ready semantic layer: 👉 https://lnkd.in/dNk4xfN3 #DataAgents #SemanticLayer #KnowledgeGraph #EnterpriseAI #LLM #NL2SQL

🚀 Timbr now supports ServiceNow as a native data source. Automatically generate a semantic ontology from your ServiceNow schema. Incidents, Problems, Changes, Assets, and CMDB tables become business concepts with explicit relationships and meaning - not just operational records. What this means for your IT operations: ✔️ See the full picture - ServiceNow stores relationships, but you query flat tables. The ontology exposes how incidents connect to services, assets, owners, and SLAs explicitly. ✔️ Define once, reuse everywhere - Build metrics and operational logic in one place. Reuse them consistently across analytics, BI, and AI agents without rebuilding per use case. ✔️ Query semantically - Ask "Which critical services have repeated incidents impacting SLAs?" without writing complex joins. ✔️ Enterprise integration - Connect ServiceNow to data warehouses, data lakes, and business systems through one unified semantic model. Your ServiceNow data becomes part of a complete enterprise knowledge graph. 👀 Want to see it in action? Join a live demo: https://lnkd.in/dKDPjHSe #ServiceNow #ITSM #Ontology #SemanticLayer #ITOperations #EnterpriseData #KnowledgeGraph

Knowledge graphs didn't fail in enterprises. They got isolated. SPARQL-based semantic platforms delivered governed meaning, explicit relationships, and strong modeling foundations. Technically, they worked. But analysts didn't query them. BI tools worked around them. And AI systems never learned to use them. Meanwhile, everything else in the enterprise data stack standardized on SQL. Not because SQL is perfect. But because it became the interface. That interface decision quietly determined adoption. In a new Medium article, we look at SQL vs. SPARQL in the age of AI, why semantics stayed isolated for years, why SQL quietly won, and why bringing semantic intelligence into standard SQL may be the shift that finally makes knowledge graphs operational at scale. 🔗 Read the article: https://lnkd.in/dPyXNpHg #SemanticLayer #KnowledgeGraphs #SQL #AI #EnterpriseData

You open Snowflake and see hundreds of views, dozens of dashboards, and at least three definitions of "active customer." Someone asks: "What's our Q4 revenue by region?" Five different teams give five different answers. Same data. Same timeframe. Different SQL. Now GenAI enters the picture. The LLM generates a query. It runs. The number looks reasonable. And no one can verify if it's right because the business logic lives in scattered JOIN clauses that were never modeled explicitly. This isn't a Snowflake problem. It's not even an AI problem. It's a meaning problem. We just published why ontology-based semantic layers are becoming critical for Snowflake environments where AI needs to be trustworthy, not just fast. 👉 https://lnkd.in/dqtvRtGH #Snowflake #SemanticLayer #GenAI #Analytics

Relationships are first class citizens in SQL ontologies. This is key for LLM accurate data retrieval by SQL queries for large scale implementations. Why? read on. Your database has 847 views. 23 are actively used. 114 haven't been queried in over a year. The rest? No one's entirely sure. Someone created vw_customer_analysis_final_v3 in 2021. It's still running. The person who built it left in 2022. You're afraid to delete it because something might break. This is the view graveyard problem. Every new use case gets a new view. Every change spawns another version. The namespace gets cluttered. Documentation falls behind. And eventually, your data team spends more time managing views than actually analyzing data. The issue isn't the views themselves. It's that they don't scale as an architecture. When every analysis requires a custom view, you're not building infrastructure. You're accumulating technical debt with a SQL extension. This is why we built Timbr around relationships instead of views. Define the connections once, let users traverse them dynamically. No view sprawl. No version creep. No archaeological digs through stored procedures. How many views does your production database have? And how many can you confidently explain? We wrote about why relationships scale where views don't (and what happens when you try to manage 1,000+ views manually). 👉 https://lnkd.in/dupgqY8b #DataEngineering #SemanticLayer #DataArchitecture #Analytics