GoodData joins agentic AI development mix with Agent Builder
Automatic connections to relevant data through a context layer and a no-code development interface aim to simplify building context-aware, cutting-edge applications.
With the launch of Agent Builder, GoodData is the latest vendor to release a framework for customers to develop and manage agentic AI applications.
As agentic AI has become the focus of many enterprises' development initiatives, numerous data management and analytics vendors have introduced toolkits designed to simplify building AI tools that can autonomously generate insights and carry out business processes.
Data platform vendors Databricks and Snowflake, hyperscale cloud vendors AWS, Google Cloud and Microsoft, and even one-time analytics specialists such as Domo and Qlik are among them.
GoodData, a longtime analytics vendor based in San Francisco, is doing the same with Agent Builder. Released on April 22, Agent Builder is a dedicated environment for creating and scaling custom agents that are context-aware based on GoodData's semantic and Context Management layers, and governed so they can be trusted to perform as intended.
Given that the development environment turns GoodData's semantic layer from a reporting asset into a feeder for trustworthy AI-powered insight generation, Agent Builder is significant for the vendor's users, according to Michael Ni, an analyst at Constellation Research.
"For business leaders, the real win is speed and reuse," he said. "GoodData users can take what they've already defined in their semantic layer and turn it into working agents in minutes, instead of starting from zero every time."
Mike Leone, an analyst at Moor Insights & Strategy, noted that GoodData is often used by independent software vendors (ISVs) to embed analytics in their own products. Agent Builder's framework will, in addition to developing agents for their own analysts and business users, enable ISVs to embed purpose-built agents across customer environments without having to build from scratch each time.
"Not having to rebuild the agent for every new customer environment is the part that I think actually matters, and it's exactly the kind of work their customers want to offload," Leone said.
Enabling AI development
Despite the emphasis many enterprises have placed on AI development over the past few years and the efforts of data management and analytics vendors to simplify agent development and deployment, most AI initiatives continue to fail.
For business leaders, the real win is speed and reuse. GoodData users can take what they've already defined in their semantic layer and turn it into working agents in minutes, instead of starting from zero every time.
Michael NiAnalyst, Constellation Research
While the reasons vary, disorganized data estates that make relevant data difficult to discover and poor data quality that results in bad AI outcomes are among the most common.
Key to enabling access to relevant data is GoodData's semantic layer, according to Rosta Striz, a principal product manager at GoodData. In addition, Agent Builder is designed to enable users to develop trustworthy AI tools by assigning ownership of agents to administrators rather than engineers, who can't be in the loop for every change, and it's built for analytics processes and workflows rather than question-and-answer interactions.
Regarding data quality, GoodData's context layer -- which goes beyond the semantic layer to define and organize both structured and unstructured data -- provides governance, predictability, auditability and security, Striz continued.
"We worked closely with enterprise customers on analytics projects where grounding, relevance and data quality were the primary success metrics," he said. "Everything we've shipped in the AI layer since flows from what we learned there. We didn't build another agent framework. We built a platform where agents can work with data in a secure, governed, and performant way."
Leone likewise cited the significance of GoodData's context layer. Because GoodData provides a semantic layer, the vendor's agentic AI development framework is set up to deliver relevant data to agents once it's in GoodData's platform, according to Leone.
"GoodData is genuinely strong because their semantic layer has been running in production for the better part of two decades, so metric definitions and access rules are already enforced before any agent touches the data," he said. "That track record is hard to match for vendors whose semantic layers were built in the last two years to ride the AI wave."
However, data quality -- freshness, anomalies, and schema changes -- is something that GoodData customers will likely need to address in their data warehouse or with data observability tools, Leone continued.
Meanwhile, despite providing a framework for AI development, GoodData isn't making it easy to build agents, according to Ni. However, like Leone, he noted that its focus on semantics and context to make agents trustworthy is key to enabling organizations to more successfully build agents and agentic networks.
"By tightly coupling agent behavior to its semantic layer and analytic logic, it reduces the risk of inconsistent or ungoverned outputs that plague more open-ended frameworks," Ni said. "By enforcing consistent definitions, permissions and enabling data products, GoodData ensures agents are grounded in governed, decision-ready data rather than raw, uncurated inputs."
Assuming the data used to inform agents is trustworthy, GoodData's Agent Builder enables the building and managing of agent networks. The agentic AI development environment features the following:
Code-first and no-code options for development, including a natural language interface, prebuilt templates and APIs that enable users to configure the roles, skills and knowledge of agents and what they can access.
Automatic connections to contextually relevant enterprise data through GoodData's governed semantic layer, an AI Memory layer that empowers agents to understand organizational rules and terminology, and an AI Knowledge layer for context-aware analytics.
A structured reasoning framework that plans tasks, selects tools, executes steps and adapts for agents, including guardrails and audit capabilities.
Centralized control over an agent's skills, personality, knowledge and permissions.
Built-in tracing, performance monitoring and usage analytics so administrators can observe agents throughout their lifecycle.
"[Agent Builder] came from … building real, high-impact AI use cases with customers across analytics, embedded products and business workflows," Striz said. "What became clear is that you can't deliver AI experiences at scale, across different users, products, and tenants, without a central infrastructure layer to configure and govern those experiences. … It's the infrastructure the agentic future runs on."
With Agent Builder now available, GoodData's product development focus is on adding depth to prebuilt agents, providing customers with more tools to customize agents, enabling users to extend agents beyond GoodData's environment to other applications workers use to do their jobs, and deepening governance and observability of agents, Striz continued.
"The common thread is the same bet we've been making from the start: governed, enterprise-grade agents as the default experience, not an advanced option," he said.
Competitive standing
While Agent Builder is similar to other agent development frameworks in that it aims to provide tools that help users more successfully build AI tools, one way GoodData's development environment stands apart is its pricing, according to Leone.
Most analytics vendors price their products on a per-user basis, Leone noted. GoodData, however, offers multi-tenancy through per-workspace pricing, which could help customers better control AI development costs.
"GoodData's two-tier model with unlimited users and multi-tenancy built in keeps that math from breaking, and that's a combination per-seat competitors can't easily match without rewriting their whole pricing strategy," Leone said.
However, adding more transparency to Agent Builder would further benefit GoodData users, Leone continued.
"I'd push them [to give] the people who consume the answers a clearer view into how the agent got to them, since that audience is broader than the developers," he said. "A business user should be able to see what context an agent pulled and why it answered the way it did without ever opening a developer console. That's the move that could attract buyers beyond GoodData's traditional builder audience."
Ni, meanwhile, noted that one of the biggest benefits of Agent Builder is that GoodData users don't have to add AI development capabilities from third parties.
"The value to customers is that they can extend their existing analytics into agent-assisted and automated decisions and actions without introducing another platform," he said. "Yes, Agent Builder can drive incremental revenue, but more importantly, it protects GoodData's core by keeping it embedded as analytics shifts toward execution."
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management.
