Ask a business question and get an answer with its evidence attached, prepare read-only SQL that is validated before it runs, interpret the dashboards you already own, and find the data faults behind a wrong number.
Most analytics capacity is spent on requests that are individually small and collectively enormous: a number someone needs, a query someone cannot write, a dashboard nobody trusts, a report that goes out every Monday. VDF analytics agents take those five shapes of work as separate jobs, run them against your own warehouse over a read-only path, and return the query and the rows alongside the answer so an analyst can check the work rather than redo it.
An open question becomes a stated interpretation, the query that tested it, the rows it returned, and the caveats that limit it — a finding an analyst can sign rather than an assertion they must verify.
Queries are written against the live schema, checked for joins that fan out, filters that silently drop nulls, and scan cost, then explained in plain language before anyone runs them.
Metric definitions, refresh times and dashboard lineage are read directly from the BI layer, so a figure that disagrees with another one can be explained instead of argued about.
Each agent has its own SEO page with use cases, governance notes, expected outputs, FAQs, and related tools.
Turn a business question into a finding with the query, the rows and the caveats attached.
Explore agent Tier 2Prepare read-only SQL against the live schema, validated and explained before it runs.
Explore agent Tier 2Explain what a KPI means, what moved it, and why two dashboards disagree.
Explore agent Tier 2Produce the recurring report on schedule, with commentary on what actually changed.
Explore agent Tier 2Find the missing values, duplicates and inconsistencies before a report inherits them.
Explore agentThese agents are sequenced by how settled the question is. Exploration comes first, the query is next, the metric layer interprets what the query produced, reporting makes it repeatable, and data quality runs underneath all four.
The Data Analyst Agent converts a vague request into testable statements, identifies which tables can answer them, and states up front what the available data cannot show.
The SQL Analyst Agent drafts read-only SQL against the real schema, flags join fan-out and cost risk, and returns a plain-language reading of what the query will actually count.
The BI Agent compares the result with the governed metric definitions and existing dashboards, and explains any difference in terms of filters, grain or refresh timing.
The Reporting Agent packages the output on a schedule with a change commentary, while the Data Quality Agent watches the underlying tables for the faults that would quietly corrupt it.
An analytics agent that can modify data is a data-loss incident waiting for a bad prompt. These agents connect through read-only credentials, are scoped to the schemas their role is entitled to see, and log every statement they execute, so a question asked at nine and a report published at ten can both be traced back to the rows they came from.
They are specialised agents for analytical work: exploratory analysis, read-only SQL preparation, BI and KPI interpretation, recurring report production, and data-quality checking — each grounded in your own warehouse and BI definitions.
No. They are provisioned with read-only credentials and can only issue read statements. Anything that would write, alter or drop is outside what the connection permits, not merely discouraged by instruction.
No. They read the metric definitions and dashboards your BI platform already governs and explain them. The semantic layer remains the source of truth for what a metric means.
See these agents applied to your operations — governed, on-premise, and orchestrated together.