Design an AI data analyst for a company with a 4,000-table warehouse. A business user asks "why did EMEA revenue drop last quarter?" and the system must find the relevant tables, write and run SQL, analyse the results, produce charts, and explain what it found in business language.
Wrong answers here are worse than no answers: a confidently-presented number becomes a slide, and the slide becomes a decision.
Build the architecture on a canvas: place the components, configure them, connect them into a data flow, and write a short reason for each one. The AI reviewer grades your design against a rubric written specifically for this problem.
How do you get from 4,000 tables to the handful the SQL agent sees? Be specific about what is indexed and how it is retrieved.
A generated query runs successfully and returns a number that is wrong. What in your design catches that before the user sees it?
The analysis step runs Python that a model wrote. Describe the environment it runs in and what it can reach.
Which requests require a human before the answer is returned, and how is that decided?
Minimum 9 components · needs a wide desktop screen