Sprint planning in an AI-first team looks different — estimation changes, task decomposition changes, and the relationship between a ticket and its implementation changes. What actually works and what to watch out for.
Most AI team metrics measure the wrong things. Lines of code generated, acceptance rates, time saved — these measure AI activity, not AI value. Here's what to actually measure and why.
The toolchain is the easy part. The culture is where AI adoption succeeds or fails. What norms an AI-first team needs, what psychological safety looks like in this context, and the conversations most teams avoid having.
AI doesn't eliminate roles — it changes what each role is for. What senior engineers, juniors, and tech leads actually do differently in an AI-first team, and why the value propositions of each role need re-examination.
Not every AI tool belongs in every team's stack. Here's how to think about toolchain decisions for an engineering team — what each layer does, how the tools interact, and how to avoid the duplication trap.
A practical maturity model for engineering teams adopting AI. Not a framework for consultants — a diagnostic for leads and engineers to understand where they actually are and what the next step looks like.
Everyone's talking about AI-first teams. Most of what gets called 'AI-first' is just 'we have Copilot licences.' Here's the actual distinction — and why it matters for how you structure your team.
Thirty posts across five arcs. Here's the honest synthesis: the biggest surprises, the things I was wrong about, what actually changed in my thinking, and where I think all of this is heading.
When a single agent misbehaves you have one thing to debug. When a multi-agent system misbehaves, the problem could be anywhere across orchestrator, specialists, connectors, and flows. How to build observability that makes the invisible visible.
Connectors, custom APIs, authentication patterns, and the integration decisions that determine whether your Copilot Studio agents actually hold up under enterprise load. What works, what breaks, and what to build carefully.