Last week, thousands of humans of data and AI came together for Atlan Activate 2025 — our biggest product launch event yet. Over the course of the day, we unveiled the industry’s first App Framework, launched new Studios for AI Governance and Data Quality, introduced the Metadata Lakehouse, and debuted powerful new capabilities like AI Search and the MCP Server.
Together, these launches reimagined the future of data governance, catalogs, and metadata — built on one bold idea: context is the missing foundation for trustworthy AI.
The undeniable context problem #
Most enterprises today face what we call the AI Value Chasm — the gap between AI pilots and production. Pilots succeed, but production fails. In fact, 95% of GenAI pilots are failing — not because of the models, but because of missing context. As one CIO told us: “The models turned out to be easy. The data and context are what’s breaking us.”
Bridging this chasm requires solving three critical gaps, what we call the Context Gap:
- Data context: Knowledge scattered across CRMs, warehouses, SharePoint, Slack, and beyond.
- Business meaning: Terms like “customer” or “TAM” may mean different things across teams — without a shared language, neither humans nor AI can align.
- Governance principles: Regulations and risks evolve weekly, and AI adoption depends on governance at scale.
At Activate 2025, we showed how Atlan is building the context layer for humans and AI — unifying context across systems, collaborating on meaning, and activating it in every workflow.
The App Framework: Unifying context across systems #
Enterprises today face a sprawl of context. Data context lives in warehouses and pipelines. Business meaning is scattered across Slack threads, SharePoint docs, and Confluence pages. Governance rules are locked away in policies and legal binders. Every new system, every new AI app, multiplies this fragmentation.
Enterprises need a way to unify context across data, business meaning, and governance — across an entire ecosystem.
That’s why we launched the Atlan App Framework: developer infrastructure to unify enterprise context at scale. Think of it as an App Store for Context.
With the App Framework, developers, partners, and customers can build and deploy context-native apps in days, not months, using:
- Pre-built APIs and SDKs for rapid development.
- Reusable components for authentication, parsing, storage, and UI.
- Enterprise-grade runtime to deploy apps securely in any environment.
- A marketplace where apps can be shared, discovered, and consumed.
We’re proud to launch with 27 partners already building apps across governance, observability, quality, and AI risk management: Alteryx, ALTR, Anomalo, Ataccama, Bigeye, BigID, Dagster, dbt Labs, Elementary Data, Great Expectations, Immuta, Lovelytics, Monte Carlo, Pantomath, Prophecy, Qualytics, Revefi, Sentra, Soda, Tavro, Telmai, Trustlogix.
One example is Anomalo Unstructured, an app built on the Atlan App Framework to solve the challenge of visibility into unstructured data quality.
For one of the world’s largest quick-service restaurant brands, customer reviews arrive as PDFs, transcripts, and documents — valuable but hard to monitor at scale. With Anomalo’s app, these unstructured sources can now be automatically scanned, enriched with metadata, PII redacted, and surfaced directly inside Atlan’s governance workflows.
Unification isn’t a vendor problem. It’s an ecosystem solution — and the App Framework makes Atlan the launchpad for context in the AI era.
Context Products & Studios: Collaborating to create shared context #
The most valuable context lives with humans. Engineers carry it in pipelines, analysts define it in KPIs, legal teams encode it in policies.
Too often, this knowledge is fragmented across Slack, Confluence, SharePoint, or simply in people’s heads. If we can’t capture and share it, even humans struggle to agree on what “customer” means. For AI, it’s impossible.
Context Products turn this human context into discoverable, reusable building blocks — bundling data context, business meaning, and governance principles together.
They’re co-created in Atlan’s Collaboration Studios — purpose-built collaboration workspaces (Catalog, Glossary, Policy).
And at Activate, we introduced two new Studios for the AI era:
The AI Governance Studio #
AI is spreading faster than enterprises can govern it, creating three major challenges:
- Shadow AI → hundreds of unknown apps and models.
- Regulatory complexity → rules shifting weekly across countries and industries.
- Governance dependency → AI trust is only as strong as the data beneath it.
The AI Governance Studio addresses these with:
- A single, unified view of every AI app and model in your enterprise.
- App registration enriched with compliance metadata and automated, agentic risk classification.
- Lineage tracing from data → models → downstream apps.
This means developers can register apps seamlessly as part of their build process, governance teams can cut review times from weeks to minutes, compliance leaders can stay ahead of evolving regulations, and executives finally gain visibility into the sprawl of AI across the enterprise.
The Data Quality Studio #
For AI to be trusted, data must be fit for purpose. If hidden errors undermine dashboards, ML models, or AI apps, adoption stalls before it starts. Yet teams face a cold start when setting up checks, and leaders lack visibility across fragmented quality signals.
The Data Quality Studio solves this by providing:
- Native checks inside Snowflake and Databricks (no-code or SQL).
- AI-suggested rules based on schema scans.
- Slack-based collaboration to resolve thresholds.
- A unified view that brings in partner signals (Anomalo, Monte Carlo, etc.).
For engineers, this means faster coverage without complex pipelines. For analysts, it’s the ability to weigh in on thresholds without writing code. For governance and observability teams, a single pane to monitor quality across tools. And for executives, the confidence to track trust KPIs across domains and scale AI with reduced risk.
Metadata Lakehouse, AI Search & MCP Server: Activating context in workflows #
Unifying and co-creating context isn’t enough — it has to be activated where humans and AI actually work. In the BI era, Atlan pushed metadata into Tableau, Looker, and Power BI so users had context where they worked. The AI era is no different. Unless context flows into copilots, agents, and AI platforms, trust breaks down.
But does context actually improve AI outcomes? We tested this with 145+ natural language questions across different complexity levels. In the baseline setup, models only saw basic schema information. In the enhanced setup, we enriched prompts with Atlan’s metadata: business glossary terms, descriptions, lineage, data quality signals.
The results were decisive. Adding metadata significantly improved SQL accuracy across every difficulty level, with complex queries seeing the most dramatic improvements. The statistical significance (p < 2e-10) proved that metadata acts as a “semantic scaffold,” helping models interpret user intent and generate reliable results. Without context, AI produces confident but wrong answers. With context, it becomes trustworthy and production-ready.
At Activate 2025, we introduced three major innovations to make context two-way, scalable, and actionable:
The Metadata Lakehouse #
Metadata is no longer static documentation — in the AI era, it’s both two-way traffic and big data:
- AI doesn’t just consume context; it generates new signals like prompts, lineage, and usage at massive scale.
- Millions of metadata records need to be stored, versioned, and queried in real time.
The Metadata Lakehouse is the warehouse for metadata:
- Iceberg-native → open, versioned, interoperable.
- Two-way by design → captures context generated by both humans and AI.
- Scalable → handles millions of records like big data.
- Queryable anywhere → natively in Snowflake and Databricks, or exposed to any BI/AI system.
- Analytics-ready → build dashboards on metadata just like data (e.g., a CDO dashboard showing AI readiness, adoption, and compliance gaps).
AI Search capabilities #
Finding the right context is one of the most painful parts of data work. Without grounding in enterprise metadata, AI search tools often produce “confident garbage out.”
Atlan’s AI Search changes that by grounding every answer in enterprise metadata. Examples include:
- “What metrics measure CAC?” → glossary definition.
- “Who owns CAC?” → the metric owner.
- “Do we have dashboards for CAC?” → live dashboards with lineage and previews.
Because results are tied to governance signals like certifications and quality, answers aren’t hallucinated or shallow — they’re precise, contextual, and trusted.
Atlan MCP Server #
Copilots like ChatGPT and Claude are powerful, but blind. They can generate, but they can’t reliably act on enterprise context. Without the right bridge, they remain just chatbots.
The MCP Server is that bridge — a universal adapter that safely connects copilots and IDEs to Atlan:
- Handles authentication, reliability, and governance guardrails.
- Embeds context directly into agentic workflows.
- Turns copilots into true teammates, capable of:
- Creating glossary terms from Confluence docs.
- Auto-documenting new databases.
- Generating compliance reports.
- Powering Slack or Teams copilots that answer domain-specific questions.
Imagine a copilot that doesn’t just chat, but reads your documentation, creates glossary terms, and keeps them updated automatically. That’s what MCP unlocks.
Building the future together: Atlan AI Labs #
The launches at Activate showed how enterprises can Unify, Collaborate, and Activate context. But some of the hardest challenges of AI-native governance remain unsolved:
- Preventing hallucinations at scale.
- Versioning context as it evolves daily.
- Keeping pace with regulations that change weekly.
These aren’t problems that we can solve alone. They demand collaboration across customers, partners, and researchers. That’s why we introduced Atlan AI Labs — an open innovation hub for trustworthy enterprise AI.
AI Labs isn’t about polished slides. It’s about rolling up sleeves together:
- Working sessions instead of presentations.
- Prototypes instead of promises.
- Tangible progress by the end of every cycle.
Focus areas include:
- Reliable talk-to-data benchmarks.
- Domain-specific AI agents powered by trusted context.
- Continuous compliance agents for regulations like the EU AI Act.
- Converting unstructured data into governed knowledge.
- Research partnerships to push the boundaries of AI governance.
Closing notes #
Activate 2025 wasn’t just a launch event — it marked a turning point. Metadata is no longer static documentation. It is the living context layer that powers humans, AI, and enterprises alike.
👉 Explore what’s new → AI Governance Studio, Data Quality Studio, Metadata Lakehouse, AI Search, MCP Server, App Framework
Together, we can close the AI Value Chasm — and build a future where AI and humans share the same trusted foundation: context.