About the news
In early October 2026, Progress released a major update to its Agentic RAG platform centered on a new capability called Smart Agent. Instead of treating every question as a single vector lookup, Smart Agent breaks complex questions into smaller sub-questions, decides which sources are most relevant, retrieves information, checks whether the results are sufficient, and continues iterating until it has enough grounded context to answer.
The release also extends retrieval beyond static knowledge bases. Smart Agent can now reach live business systems such as CRMs, ERPs, and ticketing tools through the Model Context Protocol (MCP), pulling current records into the same answer loop. It includes a built-in web search tool (currently backed by Perplexity and Google Gemini) that participates in the same plan-and-verify process. Alongside these retrieval improvements, Progress added a native Microsoft Teams application and a WordPress plugin for easier content ingestion and governance.
The same platform currently ranks number one on Snowflake’s MADQA Leaderboard, an independent benchmark developed with researchers from Oxford, UNC-Chapel Hill, and Hugging Face. The leaderboard tests whether answers are accurate and properly grounded in real documents rather than simply sounding plausible. Progress holds the top position across both agentic and conventional RAG categories.
Why it matters
Most production RAG systems still rely on a single retrieval step followed by generation. That approach works for simple questions but breaks down when the answer requires information from multiple sources, live systems, or iterative checking. Smart Agent moves the planning and verification logic into the platform itself, so teams no longer have to hand-code multi-step retrieval loops.
By combining indexed knowledge with live business data and web search inside one controlled loop, the system addresses a practical gap: many of the facts an organization needs change by the hour and never appear in a static document store. The MADQA ranking provides independent evidence that the underlying retrieval and grounding quality is competitive, which is especially relevant for enterprises that need answers they can defend.
The fact that existing deployments receive these capabilities without rebuilding pipelines or re-indexing lowers the barrier to moving from traditional RAG to more agentic retrieval.
What to watch next
The immediate question is how widely teams adopt the multi-source planning behavior in production. Success will depend on whether the quality scoring and verification steps reliably surface weak retrievals before they reach users, and how cleanly the MCP connections to live systems behave under real permission and latency constraints.
It will also be worth watching how the MADQA ranking evolves as more systems are evaluated, and whether Progress continues extending the agentic layer into additional enterprise tools. If the combination of planning, live data access, and grounded evaluation holds up at scale, this style of retrieval could become a practical default for organizations that need agents to reason over both documents and operational systems.
Sources
- Official Progress blog: Smart Agent Is Here – Retrieval That Plans, Verifies, and Reaches Beyond Your Knowledgewww.progress.com/blogs/smart-agent-is-here--retrieval-that-plans--verifies--and-reaches-beyond-your-knowledge
- Official Progress blog: Progress Agentic RAG Adds a Smart Agent Plus Native Teams and WordPress Supportwww.progress.com/blogs/progress-agentic-rag-adds-a-smart-agent-plus-native-teams-and-wordpress-support
