You have an idea for a generative AI project, but you’re struggling to begin. The most effective teams begin with rapid prototypes, and today we're going to discuss how our team is doing it with our newest project, Ground Crew.

All modern software engineering methodologies, such Agile and Lean Startup, are designed to mitigate risk. We want to evaluate feasibility quickly before making additional investments, and in generative AI we evaluate a few key questions:

Are the models smart enough?
How much data investment is required?
Will the solution be useful?

You can begin answering these questions very quickly in generative AI. The primary steps are:

1. Prepare the data.
2. Process the data with prompts.
3. Qualitatively evaluate the results.

Work through these steps quickly to evaluative feasibility, and defer process or optimization activities. Don’t invest in user stories, JIRA tickets, metrics, project plans, architecture diagrams, or design mockups, or anything that doesn’t help you answer these three questions. If you are not getting answers in the first two weeks, you are almost definitely wasting energy.

We're going to discuss how we did rapid prototyping for project Ground Crew.