Will your AI feature deliver on time?

How to measure progress reliably and communicate effectively

Is your AI product going to launch on time?

In the software world, we have become good at working backward — defining milestones, creating jiras for each task, and estimating the sizing with increasing predictability.

Based on the jira burn rate, the overall status is bubbled up as: Green (on-track), Yellow (hitting roadblocks), Red (off the road).

But, what is the equivalent approach for measuring progress on AI features? There are significant unknowns in AI/ML development — input features, datasets, models, and config.

In traditional software, a human writes the logic which has got increasingly predictable. AI/ML is inherently unpredictable as the logic needs to be derived from the data. The traditional approach of defining milestones, sizing the milestones, and tracking with color status is less useful (if not completely useless) for AI projects.

Without a proper estimation strategy, there is a risk to lose the confidence of both…