Fixing Agile for Machine Learning explores why traditional Agile frameworks struggle in data science and AI—and what to do instead.

Agile was built for predictable software delivery. Machine learning is anything but predictable. Models fail, data shifts, experiments dead-end, and “done” is never binary. When teams force ML work into classic Scrum rituals, the result is frustration, fake estimates, and broken trust with stakeholders.

This podcast reframes Agile for the realities of machine learning. We dive into:

• Why user stories, velocity, and sprint commitments break down in ML
• How to shift from delivery-centric planning to learning-centric execution
• Redefining “done” for experiments, models, and data
• Separating research from production without losing momentum
• Making data quality, bias, and model risk first-class Agile concerns
• Communicating uncertainty without losing stakeholder confidence

Whether you’re a product manager, Agile leader, data scientist, or executive trying to scale AI responsibly, this show offers practical guidance for building ML teams that learn faster, ship smarter, and stop pretending uncertainty can be planned away.