CollateData

AI agents don’t fail because the model is weak.' They fail because your data has structure… but not shared meaning. “Regional revenue” means one thing to Finance and another to Sales. An LLM will confidently pick one and move on. That’s how AI pilots stall. Today, we’re launching Collate’s Semantic Intelligence Graph. Built on OpenMetadata, it transforms metadata into a semantic graph that makes business meaning machine-readable and reusable across your stack. New capabilities: 🤖 AI Studio – A suite of AI agents for data quality, governance, documentation, and SQL. Tune them to your standards or build new agents grounded in your semantic graph. 🔌 AI SDK – Invoke semantic agents from your custom AI applications built on Python, Java, or TypeScript with no external dependencies, or integrate with external systems like GitHub, CI/CD pipelines, and n8n workflows. 🌐 Open Standards – OpenLineage, Open Data Contract Standard (ODCS) The impact: • Fewer hallucinated answers • Faster AI deployment • Governance before bad data ships • Clear auditability • Less engineering rework This is the shift from AI experimentation to production-ready, trusted AI systems. Full announcement: getcollate.io/blog/collate-l… March 5 webinar: datasciconnect.com/events/webinar… #SemanticIntelligence #EnterpriseAI #AIAgents #DataGovernance #AITrust

We polled data leaders on AI and governance integration. 100% were in the bottom two tiers. AI that can't read your governance context infers instead, and does so confidently. Here's what fixing that looks like: buff.ly/PYMWHob #DataGovernance #SemanticIntelligence

Collate 1.12 just dropped the most detailed data diff tool I have seen. Here is the problem: your DBT model runs, the data looks different, and you have no idea why. You export both tables to Excel, create VLOOKUP formulas, and spend an hour manually comparing rows. Or worse, you just hope the transformation worked correctly. The new data diff in Collate 1.12 ends that nightmare. What it shows you: • Column-level differences with color coding (renamed fields, missing columns) • Row-by-row comparison of matching records • Character-level differences within cells The demo compares a raw customer table against a transformed one. You can see that "ID" became "customer ID," spot exactly which rows are missing data, and identify character-level differences at a glance. No exporting. No spreadsheets. No guesswork. What surprised me: this level of granularity simply does not exist on other data observability platforms. They tell you tables differ. Collate shows you precisely how they differ. If you are debugging pipelines or validating transformations, this is a massive time saver. 🎥Watch and see: youtu.be/4xsM9qpEmvY #DataQuality #DataObservability #DataEngineering #DBT #DataManagement

In this new 🖱 Interactive Demo, you will learn about Data Products in Collate. They are a flexible concept that allows you to combine various data asset objects into logical groupings for different audiences. Try it here: collate.storylane.io/share/oavzebbl… #OpenMetadata #DataEngineering #DataProducts #DataMesh

Data quality testing has been significantly upgraded in Collate 1.12. If you have ever written a custom SQL query for a data quality check, copied it 50 times for different tables, and then tried to explain to business users how to run it... you know the pain. The new Test Library changes the game. Here is what it does: • Turns your custom SQL into reusable test templates with dynamic variables • Makes those tests available to business users through a simple UI (no code required) • Gives you control over which tests are enabled across your organization The demo shows a column-level test built with {{table_name}} and {{column_name}} variables that resolve at runtime, along with a custom parameter that users can set when applying the test. Apply it once, scale it everywhere. Data quality should not require a computer science degree to implement. This is how you make it accessible. 🎥Watch and see: youtu.be/021XrWkrFvM #DataQuality #DataGovernance #DataEngineering #OpenSource #DataManagement

Most orgs invest in AI first, then discover their data is too inconsistent to trust the outputs. RATP did the opposite — built the metadata foundation first. 14 data products. AI chatbots in stations. Predictive maintenance. AI that actually works. "Metadata is a strategic asset." Full story → buff.ly/rDGS6Ms #DataGovernance #EnterpriseAI #MetadataManagement @SemanticIntelligence

🎥The new episode of Data 30 with Jo Perez is a great explanation of Knowledge Graphs, what they are, and why you should care. Dive into some real-world use cases on how Knowledge Graphs can impact your organization. 👉 Watch here: buff.ly/UMT2UbX

🚀 Collate version 1.12 has enhanced the auto-classification workflow, helping teams automatically identify PII and sensitive data. With the newly surfaced auto-classification engine, your team can now see exactly which recognizer flagged a column, what pattern it matched, how many times it matched, and how confident the system is in that tag. It supports human-in-the-loop so you can review tags to make sure they’re as expected. Key takeaways ✅Get full visibility into built-in NLP recognizers (credit card, email, and more) with the ability to enable, disable, or customize them 🎯Support for your own custom recognizers using regex patterns, keyword terms, or column name matching 💡An auditable tag feedback loop so governance teams can continuously improve classification accuracy This is what responsible data governance actually looks like -- not just automation, but automation you can understand, customize, and trust. If your team is managing sensitive data at scale, this video is worth a look. Watch the full demo youtu.be/CUsY_Zi3kJA #DataGovernance #DataCatalog #SensitiveData #PIIDetection #DataCompliance #Collate #MetadataManagement

🚀 AI SDK is part of the 1.12 release of Collate Avoid the common struggles of building AI applications that automate work for your data practitioners, especially for agents that need to understand data meaning and business context. AI SDK streamlines agent development in Python, Java, and TypeScript by providing a self-contained agent framework that leverages Collate's semantic intelligence to ensure your agents take the right action on the right data. In this demo, we show that the Collate AI SDK provides an easier way to build AI applications that leverage the Collate semantic metadata graph to understand data meaning and business context, achieving accurate results. Key takeaways: ✅ Building agents is easy with the Collate Semantic Intelligence Platform with no external dependencies 🎯 Develop any type of agent that leverages the intelligence in the Collate platform 💡 Simplify data management tasks that normally take hours and days down to minutes 🎥Watch here: youtu.be/BwRLKZ8QkT4 #DataGovernance #AI #DataManagement #OpenMetadata #Collate

Most platforms say "AI-powered" but what they really mean is a complex setup for a black box you can't see inside or adjust. Collate AI Studio is different. You get a visual interface for building and customizing AI agents. No code. No waiting for engineering. Just define, configure, and deploy. Here's what that actually looks like: 🎯 Customize out-of-the-box agents (documentation, quality, etc.) with your own prompts and styles 🎯 Build completely new agents for your specific use cases. The demo shows a GDPR compliance agent that validates lineage and handles data deletion requests 🎯 Control exactly what each agent can do with bot permissions that work like service accounts 🎯 Build applications with Collate AI SDK in Python, Java, or TypeScript to integrate agents into your existing workflows The agents understand your actual data because they use Collate's semantic metadata graph, not just generic LLM training data. So when you ask about GDPR compliance, the answer references your real tables, columns, and lineage relationships. Worth a look if you're tired of AI that sounds smart but doesn't actually know your business. 🎥Watch and see: youtu.be/bXvj25GBDT4 #DataGovernance #AI #DataManagement #Automation

In this latest Collate Solution, Aydin and Shawn demonstrate how to add the MLflow open-source project for managing machine learning lifecycles to Collate. This short tutorial is an excellent introduction to incorporating machine learning models into your data flow management with Collate. 🎥Watch here: youtu.be/gTI0xsEPAM8

AI hallucinations aren't a model problem. They're a semantic chaos problem. When your data lacks unified meaning, AI agents make things up. Every time. Our new white paper shows how leading enterprises are solving this by building semantic foundations that give AI the context it needs to be trustworthy. Learn how to: → Build an enterprise knowledge graph for consistent AI understanding → Automate policy enforcement across distributed data estates → Create AI-ready data products through metadata-driven workflows → Establish semantic systems of record for human and AI workflows Download now → getcollate.io/resources/whit… #AITrust #DataGovernance #SemanticIntelligence

Build custom AI agents. Extend them into your stack. Govern sensitive data with full explainability. What's new: • AI Studio -- no-code agent builder with personas, abilities, and permissions • AI SDK (Python, Java, TypeScript) -- invoke agents from your own workflows • Auto-classification -- surfaced NLP recognizers with false positive feedback • Data Diff -- column, row, and character-level visual table comparison • GitHub Metadata Sync -- event-based metadata versioning as YAML Watch the full v1.12 overview buff.ly/u6Odk5O #Collate #AIforData #MetadataManagement #DataGovernance #DataEngineering #DataOps

For our new Collate Solutions, @progrockrec and Collate Solutions Engineer Jason Haugland, walk through connecting to CockroachDB from Collate, integrating metadata, illustrating basic features, and covering Ask Collate AI. A basic walkthrough of other capabilities is also covered in this short video. 🎥Watch here: buff.ly/MwYyXnw #dataengineering #dataquality #datagovernance #collate #cockroachdb #datalineage

Starting online in 1 hour! Our next online Meetup, "Building Semantically Intelligent Agents with AI Studio!" Join us as we introduce Collate’s AI Studio, enabling enterprises to build, deploy, customize, and tune AI agents to their unique data environments. Join here: meetup.com/openmetadata-m…

In < 48 hours! Attend our next online Meetup, "Building Semantically Intelligent Agents with AI Studio!" Join us as our community members, Alejandro Aboy and Erfan Hesami, will be sharing a new OpenMetadata demo with MCP! Join here: meetup.com/openmetadata-m…

🚀 Build AI agents that actually understand your data Join us Feb 26 @ 9 AM PT for a live demo of Collate AI Studio. See how semantic intelligence makes AI agents more accurate and capable—no more fighting complexity or dealing with hallucinations. What you'll see: ✓ How semantics solve the accuracy problem ✓ Real automation possibilities with Collate AI agents ✓ Live walkthrough of AI Studio & other new capabilities See how leading teams are automating data management tasks by creating custom agents that actually work. Register now → getcollate.io/resources/even… #AI #DataManagement #AIAgents #MachineLearning #Semantics