Data Tech 2026

AI coding tools are transforming how data and analytics teams build software, but speed without structure has a hidden cost. Unstructured AI-assisted development produces pipelines that fail unexpectedly, agentic systems that make decisions no one can trace, and codebases where intent is lost the moment the chat window closes. This session makes the case that spec-first engineering, defining system intent, constraints, and behavior before the AI writes a line, is the discipline that separates reliable AI-assisted development from fragile, untraceable output.

Attendees will see a live demonstration using GitHub's open-source SpecKit framework applied to a healthcare medication alert system, showing how structured workflows transform AI into a disciplined engineering partner. Key takeaways include how to encode requirements as AI context, why your spec is your audit trail, and how to know when structure is non-negotiable. Whether you are building data pipelines, ML systems, or agentic AI workflows, you will leave with a concrete and repeatable approach you can apply immediately.