Is AI just another tool, or something that requires you to manage it? In this episode of the Scrum.org Community Podcast, Dave West talks with Darrell Fernandes about his paper, “The Hidden Management Gap in AI Adoption.”

Darrell explains why AI sits in a gray area: more than a compiler or a database, but certainly not a human. Because it infers, and may not give the same answer twice, working with it well calls for skills that have traditionally belonged to managers, and that most individual contributors were never given the chance to build.

Drawing on the work of Michael Watkins and Patrick Lencioni, Dave and Darrell discuss:

• Why you remain accountable for AI output, even when you didn’t write it
• The “automation deference” failure mode and how polished output can hide hallucinations
• How regulation and organization size shape where AI governance gaps show up
• Why review bottlenecks can’t simply be escalated to senior leaders
• One simple practice you can start today: writing a RACI before you engage AI in a workflow

Read the paper: https://www.scrum.org/resources/hidden-management-gap-ai-adoption

Why are forecasts so often late, or short of what was planned? In this live Ask a Professional Scrum Trainer session, Professional Scrum Trainer Ludwig Harsch answers audience questions on estimation and planning for Scrum Teams. He draws on his training and consulting work to explain why estimation should be tested and adapted to each team's context, not treated as a universal rule.

Ludwig explains how teams can move from unreliable hour estimates to relative sizing. He discusses when Planning Poker is worth the effort and when a lighter approach fits better. He also covers how to forecast when support work, context switching, and external dependencies eat into capacity. He makes the case for giving stakeholders projected dates with probabilities rather than fixed promises, and explains how Monte Carlo simulations can support that conversation.

In this episode:

• Moving from hours to relative sizing, and what "size" should mean
• Using Planning Poker and facilitation to build shared understanding
• Knowing when to stop estimating
• Using AI in estimation without losing team discussion
• Choosing a planning horizon, including for regulated products
• Handling support work, dependencies, and shifting priorities
• Comparing velocity, throughput, and capacity-adjusted metrics
• Why Scrum Teams commit to the Sprint Goal, not a fixed number of backlog items
• Advice for newly formed teams still collecting data

Tune in for some great insights on Estimation and Planning in Scrum.

Matthew Hodgson, founder of Zen X Machina, joins Dave West to unpack the Raptors project, an experiment that started by accident during digital transformation work and grew into a working model for agentic teams. Hodgson explains how he built a cross-functional AI team, complete with a Scrum Master agent named Al, and what it took to get the agents to self-organize, adapt sprint by sprint, and catch their own mistakes. The conversation digs into why explicit governance files matter more with agents than with humans, where agents still need human oversight, and why traditional governance models move too slowly for agentic AI. Hodgson also previews ideas from his book, Evolve, on adaptive governance for modern product operating models.

Access the whitepapers that cover the experiment and learnings.