Analyst-First System Boosts Proactive Enterprise Analytics with Verified Knowledge

Traditional conversational analytics systems assume users already know what to ask, which is a barrier for non-experts facing complex enterprise data. Current "proactive" tools offer limited insights, often relying on pre-curated metrics or query logs that aren't available for new datasets. A new production analytics system inverts this paradigm to an "analyst-first" approach. It introduces two core architectural innovations: first, a pluggable "domain-expert skill" abstraction, which are self-contained, folder-based subject-matter packs that can be automatically selected and integrated into the agentic pipeline, schema explorer, and report engines. These skills provide context-specific prompts, references, and report templates. Second, an offline knowledge-compilation loop proactively analyzes the dataset using DuckDB, performing critic-gated convergence and data-validating joins. This process generates durable schema knowledge that drives standing expert reports with re-verified metrics and suggested questions, effectively closing a proactive loop where insights lead to questions, and clicks launch verified deep dives, all before a user types a query.