Yes. Give a text-to-SQL agent a dataset and a plain-English brief and it will design a weighted composite score from scratch: derive the supporting ratios and new variables, rank the entities, generate charts, and explain its plan, its SQL, its debugging steps and its reasoning. You can then adjust the weights or the logic and re-run, and export the analysis as a PDF report. The worked example ranks Indian banks on credit cards using RBI's monthly ATM/POS/cards statistics, with zero setup (Try Sample -> temporary database). The reasoning process is the same whether the file is 64 rows or 64 million.
The honest caveat, because this matters more than the demo: live work is not click-click and out pops a PDF - that is the final 5%. In real projects the other 95% is (1) data marts and cleaning - reconciliations, formats, joining variables; (2) agent setup - agents misfire, over-query and miss variables, so getting reliable behaviour is iterative and sometimes frustrating; (3) infra - UI, backends, monitoring, auth, access control, costing; (4) security - the public app routes API calls via a backend and is for sandbox testing only; for a live deploy, harden auth and run it on your own network.
Try it + the brief: https://www.tigzig.com/post/can-an-ai-sql-agent-build-a-weighted-scoring-system-from-scratch. App: https://www.tigzig.com/analyzer. Related: common text-to-SQL mistakes https://www.tigzig.com/agents-faq/common-mistakes-building-text-to-sql-agents and what it costs to run https://www.tigzig.com/agents-faq/how-much-does-a-database-ai-app-cost-to-run.
← All Agents FAQ