$40M+
Subscription ARR the analytics serves
5,59,000
recurring subscribers
$700M
Annual revenue
As subscription and services revenue grew into a strategic priority, executives, sales, and marketing needed fast, reliable answers from a large Snowflake estate. Tableau dashboards were the only interface available — but complex questions required moving across multiple reports, queries on large data volumes could time out, and a generic AI assistant had no knowledge of the company’s tables or business definitions. Inferenz built a customized conversational analytics layer on Claude, connected via Snowflake MCP and Tableau MCP, grounded with AI context pages and enriched metadata. Business users now ask questions in plain English and receive HTML visualizations and narrative explanations in a single conversation. Subscription Analytics is the first project in production, with org-wide rollout underway.
Answering a single executive question often meant navigating across multiple Tableau dashboards and reports. A natural follow-up — such as why a particular subscription channel was underperforming — required opening several more, with no ability to keep context across the conversation.
With tens of millions of records flowing in continuously, Tableau queries against the full data volume could time out, slowing the very people who needed answers fastest and undermining confidence in the analytics platform.
An off-the-shelf AI assistant had no knowledge of the company's tables, column definitions, or business logic, so it could not reliably generate correct SQL or route to the right data in the Snowflake estate.
A dashboard could show thatthe number was high or low but not what was driving it. Understanding the root cause required further manual investigation across additional reports — adding time and friction to every leadership decision.
Inferenz built a customized conversational analytics interface on Claude, connecting it to the governed Snowflake estate and existing Tableau visualizations through MCP connectors, then grounded the assistant in the company's own data definitions and business context.
Snowflake MCP connects Claude directly to the governed production Snowflake estate for text-to-SQL queries. Tableau MCP connects it to existing visualizations. Both connectors read from the same trusted data the business already relies on, keeping self-service answers and existing dashboards consistent.
Using Claude Projects, AI context pages describe the business, its KPIs, and how each metric is defined. Instructions tell the assistant how to interpret the company's specific data model, so responses reflect how the business actually uses its numbers rather than generic assumptions.
Table and column comments were added across the production schema, generated efficiently using Snowflake Cortex. Every data point carries a description the assistant can reference, ensuring MCP queries route to the correct tables and views rather than guessing from column names alone.
Business users ask questions in plain English and receive answers as HTML visualizations paired with narrative explanations — the number, a chart, and the reasoning behind it — all in a single conversation. Follow-up questions remain in context, eliminating the need to navigate between reports.
Subscription Analytics is the first Claude Project taken to production, with a feedback loop that checks response accuracy and refines context over time. Because Claude Projects can be shared across an organization, the same grounded experience is being extended to executives, sales, and marketing beyond the initial analyst team.
For analytics
Plain-English questions, HTML visualizations, and explanations in one conversation.
Snowflake + Tableau
The assistant reads the same governed data the business already trusts.
In company data
Table and column comments route every query to the right tables and views.
Self-service
Shareable Claude Projects extend access to executives, sales, and marketing.
Whether you’re starting with data modernization or exploring AI copilots, we’re here to help.
Contact Us
