Coding agents have been adopted by more than 90% of development teams — and yet something strange is happening: productivity gains have plateaued. In a candid conversation at Atlassian Team in Amsterdam, Minghu, head of engineering at Atlassian, explains exactly why, and what needs to change next.
If your organization has invested in AI coding tools and isn’t seeing the returns you expected, you’re not alone — and the reasons may surprise you. The bottleneck isn’t where most people think it is. In this Techzine TV interview, Minghu shares what Atlassian has learned from its own internal data, customer surveys, and the development of its AI-powered SDLC platform.
This is a conversation that challenges some of the most common assumptions in enterprise software development today — from how code review actually needs to work at scale, to whether human-in-the-loop is helping or hurting your teams, to what it really takes to get quality AI-generated code into production.
What you’ll discover in this interview
- Why near-universal coding agent adoption hasn’t translated into the productivity gains teams expected — and where the real numbers stand
- What Atlassian discovered when it looked at the actual bottlenecks across the full SDLC — and why it completely shifted its strategic focus
- The uncomfortable truth about code review at scale when AI is generating ten times as many PRs as before
- Atlassian’s vision for what AI-facilitated code review should actually look like — and how far along it already is
- When human oversight genuinely adds value, and when it quietly becomes a liability
- The role of risk estimation in deciding what should be auto-approved and what needs human eyes
- The surprising impact that shared context has on AI output quality — including specific numbers Atlassian has measured
- How the Teamwork Graph connects your codebase to your business intent — and why that changes what AI agents can do for your team
- What the AMP multiplayer protocol means for developers moving between their IDE, Jira, and Confluence
Why this interview matters
Minghu brings a rare combination of internal product development insight and real customer data to this conversation. Whether you’re leading an engineering team, shaping your organization’s AI strategy, or simply trying to understand where enterprise software development is heading, this interview offers a grounded, data-informed perspective on what actually moves the needle.
Watch the full interview below to find out what Atlassian believes comes after the coding agent era — and what your team should be focusing on right now.