Ataccama open-sources data quality layer for AI agents via Ossie

Ataccama, the enterprise data trust platform, is open-sourcing a converter that pipes governed business definitions and live data quality signals into Apache Ossie, the open semantic specification incubating inside the Apache Software Foundation. The move gives AI agents and analytics tools a way to read not only what enterprise data means, but whether it is currently reliable enough to act on, all without requiring that data to be consolidated onto a single platform first.

The distinction matters because the agentic AI era is raising the stakes of data errors in a qualitatively new way. During the business-intelligence era, a bad figure in a dashboard tended to get caught: it conflicted with a rival report, a memory of last quarter's board deck, or a trusted baseline. Autonomous AI agents have none of those human cross-checks. As Ataccama's Chief Product Officer Jessica Smith puts it: "An agent may know what revenue means, but not whether the revenue data it's reading is complete or current, and it will still produce a convincing answer."

From semantic layers to trust layers

Semantic layers, the software that tells agents how to interpret business concepts such as "revenue" or "customer", have become standard infrastructure in enterprise AI stacks. Platforms including Snowflake, Databricks, and dbt already contribute to the Apache Ossie specification, which standardises how those definitions travel across tools. What Ataccama's converter adds is a second dimension: continuously evaluated quality signals from the source systems where data originates, expressed in Ossie-compliant YAML and therefore portable across every environment that consumes the semantic model.

Josh Klahr, Head of Product Management at Snowflake, frames the contribution as a portability problem solved: "Ataccama is extending that portability to data quality, giving organisations a consistent way to make those signals available across platforms rather than rebuilding them for each environment." Ataccama's Model Context Protocol (MCP) Server adds a further layer, letting agents retrieve granular evidence, which quality checks failed, which records were affected, when they need to reason through an issue rather than simply act on a pass/fail signal.

The agentic enterprise and its data liability

The broader implication runs well beyond any single vendor's product release. As enterprises accelerate deployment of agentic AI systems across finance, operations, and customer management, the quality of the data those agents consume is becoming a board-level risk management question. A poorly governed data estate that was tolerable in a human-reviewed analytics workflow becomes a liability when agents are delegating operational decisions at speed.

The convergence at play here spans at least three sectors: the datatech infrastructure layer (semantic standards, cloud data platforms), the AI orchestration layer (agentic frameworks, MCP, foundation models), and the enterprise risk and compliance layer (data governance, auditability, regulatory exposure in financial services and healthcare). Capital in this space is concentrating accordingly, Ataccama's Gartner Leader status across augmented data quality and data governance signals that analysts expect the governance-and-trust segment to grow alongside, not lag, the agentic AI wave.

For cross-sector investors, the strategic read-across is clear: the value of any agentic AI deployment is increasingly bounded by the trustworthiness of its data substrate. Vendors that sit between the enterprise data estate and the AI orchestration layer, owning the quality, lineage, and semantic context that agents rely on, are positioned to capture meaningful margin as agentic workforces scale. Ataccama's open-source contribution to Apache Ossie is as much a standards-positioning play as a product release, aiming to make its quality signals the default trust mechanism in a specification that Snowflake, Databricks, and dbt are already treating as infrastructure.