Analyst(s): Brad Shimmin
Publication Date: June 12, 2026
Collate has announced Collate 2.0, an AI-native data governance and catalog platform rebuilt on OpenMetadata, with a semantic context graph at its core. The new release adds AI Studio for agentic workflow orchestration, a Context Center for semantic enrichment, agent memory, and a conversational interface for data discovery, all designed for AI agents whose reasoning accuracy depends on structured enterprise context, and for the data professionals whose job increasingly involves governing those agents.
What Is Covered in This Article:
- The semantic context graph as Collate 2.0’s foundational architectural layer, and how it differs from conventional catalog approaches to AI data access
- The three new capability pillars: AI Studio for agentic workflow orchestration, the Context Center for semantic enrichment, agent memory, and a conversational natural language interface for discovery
- Collate’s open-source foundation on OpenMetadata and the strategic implications of building on open infrastructure rather than defaulting to hyperscaler-native catalog tooling
- The evolving data professional — the AI Shepherd — and how Collate 2.0 organizes itself around governance and validation work
- Candid risks and open questions, including what a context-first architecture demands of enterprise teams and where proof still needs to materialize
The News: Collate has announced Collate 2.0, an AI-native data governance and catalog platform built on OpenMetadata, an open-source project that serves as its technical backbone. The headline architectural element of this release, however, is a semantic context graph, a structured knowledge layer intended to provide AI agents with the business context required to reason accurately over enterprise data rather than simply locating it.
The release introduces three primary capabilities. AI Studio provides an environment for building and managing agentic AI workflows tied to data governance tasks. The Context Center serves as the hub for semantic enrichment, weaving metadata, lineage, glossaries, and business definitions into a unified, graph-structured layer. A conversational natural language interface rounds out the release, enabling data discovery and interrogation through plain language and lowering the technical barrier for a broader set of stakeholders. Collate has positioned the platform to serve two audiences simultaneously: AI agents as consumers of enterprise context, and data professionals as the governors and validators of agent-generated output and actions.
Collate Turns OpenMetadata Into a Persistent Semantic Memory Layer for Enterprise AI Agents
Analyst Take: The persistent gap between what large language models (LLMs) can do and what enterprise AI governance actually requires has rarely been a model problem. It has been a context problem. Collate 2.0 and its semantic context graph make a direct architectural wager on that distinction, arguing that an agent’s ability to reason correctly over enterprise data depends less on raw model horsepower and more on the structured business meaning surrounding that data.
The market is starting to agree. According to Futurum’s 1H 2026 Data Intelligence, Analytics, and Infrastructure Decision Maker Survey, 44.5% of respondents plan to increase spending on the semantic layer over the next 24 months. This is a clear signal that enterprises are treating semantic infrastructure as a near-term budget priority rather than a someday aspiration, as in the past. That readiness is the backdrop against which Collate 2.0 arrives, and that matters because a semantic context graph is structurally different from bolting metadata tags onto an existing catalog.
The Context Problem That Catalog Versioning Never Solved
Traditional data catalogs were built to help people find and securely access data. They were never designed to help an autonomous agent reason over it. When an AI agent queries enterprise data without embedded definitions, lineage, ownership, and glossary context, it operates on syntactic similarity rather than semantic meaning, and the output reflects exactly that limitation. The agent can find something that looks relevant without understanding whether it is correct, current, or governed.
Collate’s semantic context graph attacks this at the infrastructure layer rather than the application layer, and the distinction carries actual architectural weight. When context lives in the infrastructure, it becomes persistent and reusable across many agents, instead of being painstakingly prompt-engineered for each individual use case. This is the same logic driving the broader interest in graph-based reasoning and knowledge-graph-grounded retrieval, where the structure of relationships (as opposed to just the proximity of vectors) determines the quality of an answer. A context graph gives agents a map of how enterprise concepts relate, which is precisely what syntactic retrieval cannot provide.
Three Capabilities, One Unified Intent
It’s important to read the three pillars (AI Studio, Context Center, and natural language capabilities) as a single governance loop rather than three separate features. AI Studio is the orchestration and audit surface – the mechanism for defining what agents are permitted to do with enterprise data, monitoring their behavior, and reconstructing their reasoning after the fact. It reflects the move toward supervised autonomy, where agents are granted latitude to act but remain inside an observable, governed perimeter.
The Context Center is where the semantic graph gets populated. Business glossaries, lineage, ownership metadata, and domain definitions converge there, and the quality of this enrichment directly determines the quality of agent reasoning downstream. The conversational interface, meanwhile, is the front-end expression of how well that underlying graph is structured. Natural language access reveals the coherence of the context layer beneath it. This forms an accurate, fluent conversation with enterprise data that can only happen when the semantic graph is genuinely sound. This allows Collate to deliberately make the experience persona-adaptive, serving data engineers, domain stewards, and business stakeholders without forcing them through a single undifferentiated interface.
The Open Foundation Argument
Building on the OpenMetadata project is a philosophical position as much as a technical one, and it concerns who owns the metadata schema and context graph inside an enterprise. Hyperscaler-native catalogs exert a quiet gravitational pull: once metadata structures are bound to a single cloud provider’s schema, the cost of leaving climbs steeply. For this reason, Futurum views the metadata layer as the new battleground for data gravity and autonomy.
The trouble, according to Futurum research, is that the majority of enterprises don’t prioritize their metadata decision. The 1H 2026 Decision Maker Survey found that 41.3% of organizations land on cloud-native data catalogs by default rather than through deliberate architectural selection. That is the inertia Collate 2.0 is built to interrupt. An open, intentional alternative carries the most weight precisely because it can enable organizations to extend, connect, and migrate their context without rebuilding the semantic layer from the ground up. Doing so aligns with the broader steps the market is taking toward a composable, open-data ecosystem that will reshape this category over the coming months.
A Platform for the AI Shepherd
The data professional’s job has been quietly and comprehensively rewritten not by choice, but by the velocity of agentic AI capabilities across enterprise environments. Building pipelines and authoring queries now share the calendar with auditing AI output, validating agent reasoning, and communicating insight quality to the business. Collate 2.0 organizes itself around this evolved practitioner, whom Futurum now refers to as the AI Shepherd. AI Studio supplies the audit surface, the Context Center delivers the enrichment tools, and the conversational interface serves as the communication layer. Interestingly, as more software reorganizes around both human and machine consumers, these sorts of design choices read less like a feature roadmap and more like a job description for the person now responsible for keeping agents honest.
Honest Friction: Where the Architecture Must Prove Itself
It’s important to remember that a semantic context graph is only as good as the enrichment that fills it, and the Context Center forms a capability, not a content factory. Organizations with sparse metadata, inconsistent taxonomy, or fragmented ownership will discover that Collate 2.0 raises the bar for their underlying governance discipline. That is a prerequisite worth naming plainly rather than a flaw. In short, companies should not view these new tools as a shortcut to value. Companies must invest in understanding, documenting, and codifying institutional knowledge across their data estate. As inscribed at the Temple of Apollo at Delphi, “know thyself.”
Agent orchestration at scale also introduces fresh failure modes. When an agent reasons incorrectly despite having context, the debugging surface grows more complex, not simpler, and AI Studio will need robust explainability tooling to sit alongside its workflow management. Finally, because OpenMetadata is open source, Collate’s commercial differentiation lives almost entirely in the semantic context graph and the AI-native experience layer. That ground will need continual defending as the OpenMetadata ecosystem matures and attracts competing commercial wrappers.
What to Watch:
- The broader contributor ecosystem is already embracing Collate’s context graph approach, which will only elevate the stakes as Collate seeks to build its commercial moat while avoiding competition from within its own foundation.
- Snowflake, Databricks, AWS, Google Cloud, and Microsoft each ship native catalog and governance tooling. Watch how they fold graph-structured context into upcoming releases and whether Collate’s open architecture proves more durable than native integrations.
- The value of the context layer scales with the number of agentic frameworks that can consume it natively. Integration announcements that establish Collate as a default context provider for enterprise agent pipelines will be telling.
- Individual success stories will establish credibility for Collate, but the real validation arrives when organizations can quantify improvements in agent accuracy and audit quality after deploying the context graph.
See the complete announcement on the Collate 2.0 launch on the Collate website.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Other Insights From Futurum:
Semantic Layer Set to Become the Next Piece of Critical Infrastructure
Can a Database Truly Be a Genius? IBM’s Shift Toward Agentic Autonomy
Author Information
Brad Shimmin is Vice President and Practice Lead, Data Intelligence, Analytics, & Infrastructure at Futurum. He provides strategic direction and market analysis to help organizations maximize their investments in data and analytics. Currently, Brad is focused on helping companies establish an AI-first data strategy.
With over 30 years of experience in enterprise IT and emerging technologies, Brad is a distinguished thought leader specializing in data, analytics, artificial intelligence, and enterprise software development. Consulting with Fortune 100 vendors, Brad specializes in industry thought leadership, worldwide market analysis, client development, and strategic advisory services.
Brad earned his Bachelor of Arts from Utah State University, where he graduated Magna Cum Laude. Brad lives in Longmeadow, MA, with his beautiful wife and far too many LEGO sets.
