Brighte reposted this
We're running teams of AI agents that coordinate work between themselves. This is Claude agent teams - the next chapter in the AI playbook we're writing at Brighte Our engineering teams have been building with AI across our operations for months now. Only recently we started working with Claude agent teams, a capability Anthropic released on 5 February this year. Here's the simplest way I can explain it. Most AI coding tools use "subagents" - a main AI agent delegates tasks to helpers, they work in isolation and report results back. Agent teams are a different thing entirely. Multiple AI instances share a live task list, communicate directly with each other, and coordinate the work between themselves. Less like a boss giving instructions. More like a small team collaborating in real time. Photo attached to explain the diff and source article in comments 👇 Our Staff Software Engineer Chris Young , has been leading the charge on this at Brighte. He recently showed me a team lead agent coordinating four other agents in parallel - backend, frontend, QA, and code review - across a process that previously had people in each of those roles. It was pretty cool tbh seeing this live in action. We are excited enough about the impact that we made the decision to move faster. We cut back engineering delivery workload to 50% and gave the teams clear bandwidth to focus on building out agent teams across our codebase. This dedicated time is about going deeper and faster, so we can 2-3x engineering capacity over the coming months. Where does this end up? Some of our codebases will reach a point where agents run end to end with no human in the loop. But not all of them. Stripe, who operate in regulated payments and credit like us, have written about this same reality. Some workflows in financial services need human judgment and that's not changing. We're being deliberate about which is which. Because as I've said before, people drive technology outcomes. Not the other way around. If you're exploring agent teams or similar approaches, I'd love to hear what you're seeing. What's working? What's not? in comments below 👇 This is post three in our weekly series on the AI journey at Brighte. We're sharing what we're building, what we're learning, and what's actually working. Let's keep writing this playbook... #AgenticAI #Fintech #CleanEnergy