← All work
Enterprise GenAI

Shipping generative analytics before the playbook existed

Text2SQLPythonEnterprise AI
Reduced quarterly business-review reporting time by 70%.

The problem

Business reviews depended on analysts translating recurring questions into SQL, assembling the results, and rebuilding the same reporting views. Natural-language analytics could shorten that loop, but enterprise adoption required more than a model demo: generated queries had to fit existing data access, review, and compliance rules.

What I shipped

GPT-based Text2SQL and Python analytics tools for quarterly business-review workflows. I led the product through engineering, legal, and compliance, turning an emerging model capability into a production workflow rather than a standalone experiment.

What was hard

The model was only one dependency. The product also needed a clear boundary between generation and execution, a reviewable output, and enough transparency that users could trust the analysis without treating fluent text as proof of correctness.

The result

The tools cut quarterly business-review reporting time by 70% and established an early path for shipping generative AI inside an enterprise environment.

Running untrusted agents in production →Platform operations as agent-executable skills →Rules for the 80%, models for the tail →Right-sizing a fleet with a seasonal load curve →Grounding a sales assistant in renewal knowledge →Building the operational view from zero →Scaling content governance with policy and data →