GoodData’s Context Management aims to make AI trustworthy

While those suites simplified AI development, they did not improve the high failure rate of AI projects. They provided simple frameworks for development, and often included support for MCP, but did not address the data that provides AI tools with contextual awareness.

In response, numerous vendors so far in early 2026 have introduced capabilities aimed at improving the accuracy of data retrieval pipelines so they feed AI tools with more contextually appropriate data. Among them, Databricks launched Instructed Retriever to provide an alternative to traditional retrieval-augmented generation, MongoDB released new embedding and reranking models, and vendors including Pentaho and Insightsoftware have added semantic layers.

GoodData has provided semantic modeling capabilities since 2020 and modernized its semantic layer with AI capabilities in 2025. With the release of Context Management, GoodData is augmenting semantic modeling with additional capabilities, including governance and observability to improve data retrieval for AI.

A combination of customer feedback and an understanding that context is key to successful AI development provided the impetus for developing Context Management, according to Peter Fedorocko, GoodData’s field CTO.

“As customers began using our AI capabilities more heavily, they naturally wanted more control over how answers are generated — how to guide, refine and tune AI responses,” he said. “That demand highlighted the need for stronger context management. At the same time, we believed … that successful enterprise AI deployments depend more on robust context management than impressive demos.”

Context Management is a contextual layer that is designed to make AI-powered analytics tools more accurate. Its capabilities include the following:

• Semantic modeling  that consistently defines metrics, dimensions of data and business logic  so that agents, dashboards and APIs use the same definitions to access relevant data.
• Automatically applied governance to ensure that agents and other AI tools access and use data according to preset guidelines to prevent misuse and potentially unsafe actions.
• Knowledge grounding that bases every response on governed data and makes outputs traceable so they can be trusted.
• Observability that tracks prompts, inputs, outputs and expenses to make AI-powered analytics more transparent and auditable.
• Guidance to help define how AI tools should behave, including setting consistent terminology and priorities.

Mike Leone, an analyst at Omdia, a division of Informa TechTarget, like Ni noted that Context Management is a significant addition for GoodData users.

“It formalizes what every analytics vendor is scrambling to figure out right now, which is how to make AI answers trustworthy in production,” he said. “For GoodData’s embedded customer base, specifically, where you’re shipping AI-driven insights inside someone else’s product, governed context is a requirement to get there.”

Meanwhile, the most valuable feature of Context Management is semantic modeling, Leone continued.

“Everything else falls apart without it,” he said. “If your metrics aren’t defined once and enforced everywhere, governance and the rest of the pillars fall apart.”

While Context Management is an important addition for GoodData users developing AI-powered analytics tools, it also has significance for GoodData from a competitive perspective, according to Ni.

He mentioned that many analytics vendors are adding semantic layers. What they don’t offer, however, are capabilities integrated with the semantic layer that govern how AI systems use definitions in production.

“Semantics answers the question, ‘What does this number mean?'”, he said. “Context goes further. It answers which definition should be used in this situation, how should AI apply it and what actions are allowed once that answer is produced.”

Leone likewise noted that, while the concept of Context Management isn’t unique, the composable, embeddable way it makes such capabilities available could be a differentiator for GoodData.

“Every analytics vendor is trying to figure out governed AI with varying degrees of success,” he said. “Where GoodData has an edge is the delivery model. … Those are the things that resonate with product teams building AI into their own applications.”