Making AI useful means making it understand your business
“AI initiatives require more context than ever. Atlan's metadata lakehouse is configurable, intuitive, and able to scale to hundreds of millions of assets.”
Andrew Reiskind
Chief Data Officer
“With Atlan we cataloged over 18 million assets and 1,300+ glossary terms in our first year, so teams can trust and reuse context across the exchange.”
Kiran Panja
Managing Director, Cloud and Data Engineering
“Atlan captures Workday's shared language to be leveraged by AI via its MCP server. As part of Atlan's AI labs, we're co-building the semantic layer that AI needs.”
Joe DosSantos
VP, Enterprise Data and Analytics
Most enterprise AI fails not because of the model, but because of missing context
The wall isn't the model. The models are extraordinary. The wall is that no agent can reason effectively about a business it doesn't understand. It doesn't know what your data means. It doesn't know how your teams actually work. It doesn't know the difference between how your company defines "revenue" and how the rest of the world does.
We call this the AI context gap. We think closing it is one of the most important unsolved problems in making AI genuinely useful for organizations.
“We built a revenue analysis agent and it couldn’t answer one question. We started to realize we were missing this translation layer. We had no way to interpret human language against the structure of the data.”
Joe DosSantos
VP, Enterprise Data & Analytics
One question for AI. Multiple layers of context.
Through our work with enterprises, we've found that even a simple agent task requires multiple layers of context working together. Miss one layer and the answer breaks.
Who is asking — and what decision?
Editorial or marketing?
What does “new” mean here?
Released this month or new to the platform?
How do you define “Top 10”?
Views, watch time, or ratings?
Which tables hold watch time?
Raw logs vs. aggregated metrics
How do you measure viewership?
Total plays, unique viewers, or hours streamed?
A shared context layer, built through human collaboration
We believe the right answer is not to embed context into individual agents — that fragments knowledge and creates inconsistency. Instead, we're building a universal context layer: a shared, living source of truth that any AI agent can draw from.
FoundationEnterprise Data Graph
We bring together metadata from hundreds of sources — business systems, data systems, BI tools, pipelines, warehouses — and convert it into a unified Enterprise Data Graph. Lineage, query history, semantics, and quality all interconnected.
EnrichmentAI-generated context
Using the data graph as input, AI automatically generates descriptions, links terms to business concepts, identifies metrics and KPIs, extracts common query patterns, and bootstraps an ontology of how your organization's data relates to its business.
CollaborationHuman-in-the-loop refinement
AI gets you 80% of the way. The remaining 20% requires human judgment — resolving conflicts between competing definitions, certifying which metric is canonical, annotating edge cases. We've designed this to feel like a natural collaboration, not a governance burden.
KnowledgeActive ontology
Entities, attributes, and relationships that encode what your organization knows — bootstrapped by AI, refined through collaboration. A living model of your business that agents can query and reason over.
MemoryEnterprise-wide memory
Every interaction, every correction, every piece of feedback becomes part of a persistent institutional memory. The system gets better with use, compounding knowledge across every team and every use case.
RuntimeLive context at decision time
When an agent answers a question, it matters who's asking and why. Runtime context provides the situational awareness — user identity, relevant policies, current permissions — that turns a generic answer into the right answer.
From connection to collaboration to activation
We've designed the process to be incremental. You don't need to solve the entire context problem before seeing value. Each step builds on the last, and the system improves continuously through use.
1. Connect your data estate
Bring together metadata from across your organization — business systems, data systems, BI tools, warehouses, pipelines — and unify it into a single Enterprise Data Graph that captures how everything relates.
2. Let AI generate the first layer of context
The system automatically produces descriptions, links terms, identifies metrics, and bootstraps an ontology from the evidence already present in your data. A strong starting point without manual effort.
3. Collaborate to refine and certify
Your teams review, debate, and improve the AI-generated context. They resolve conflicts, certify canonical definitions, and annotate the nuances that only humans understand.
4. Activate context across every agent
Governed context flows to any AI agent via SQL, APIs, or SDK. Feedback from real-world usage feeds back, creating a continuous loop where every interaction makes the context layer more complete.
The same challenges, across every industry
The specifics vary, but the underlying pattern is strikingly consistent. These are real challenges from teams we've worked with.
Cold start
"Critical business logic already exists, but not in a form AI can use. Getting to a credible first version feels slow, manual, and overwhelming."
— Leading UK retail group
Testing
"Validation relies on spot checks and intuition, not repeatable processes. Without a clear definition of 'done,' shipping feels risky. One wrong response could break trust."
— Global lifestyle brand
Scale
"We had early success, but as we added more models and data sources, experience started to degrade. We need a shared foundation for managing context across use cases."
— Leading CRM SaaS company
Principles guiding our work
We hold a few strong convictions about how the context layer should be built. These shape every decision we make.
Context is a team sport
Frontline teams — not just engineers — need to be able to read, question, and improve the context that shapes AI behavior. We design for collaboration first, because the best context comes from people working together.
Built for what comes next
The same context layer powers MCP, A2A, and whatever protocol emerges tomorrow. We believe context should outlive any single technology cycle, so we build for portability and permanence, not for today's stack.
Open and portable by default
Your context belongs to you. It should move freely across agents, models, and clouds. We think the worst outcome would be organizations locked into a single vendor's representation of their own knowledge.
One source of truth, many agents
Every agent should learn from the same living context. When one team improves a definition or certifies a metric, every agent across the enterprise gets smarter. Context compounds — and that compounding is the real value.
