Atlassian AMP: New Protocol Tracks AI vs Human Code Authorship

What happened

Atlassian has launched a new set of offerings aimed at giving enterprises clearer visibility into which parts of their codebase were written by AI versus human developers. The centrepiece is Agentic Multiplayer Protocol (AMP), described by the company as the framework governing how humans and AI agents collaborate across its platform — giving agents an identity, scoped authority, shared context and tasks, and reviewable results.

The launch also includes Teamwork Graph, a capability that indexes source code down to the level of individual functions, symbols and classes, making it easier for developers to search and trace work across both Bitbucket and GitHub. Jamil Valliani, Atlassian's head of AI products, said the inability to distinguish AI-generated from human-written code — work that in practice is rarely one or the other, but usually a blend — is a significant barrier to enterprises scaling their AI adoption further.

Why it matters

This is fundamentally a story about the plumbing of AI adoption inside software organisations. As coding agents become embedded in everyday development workflows, enterprises are discovering that the hard part isn't getting AI to write code — it's knowing what it wrote, why, and whether it can be trusted, audited or rolled back. AMP addresses that gap by formalising agents as accountable participants in a shared workspace rather than invisible contributors, which is a meaningful shift in how organisations govern human-machine collaboration at scale.

For technology and transformation leaders, the move signals that attribution and provenance are becoming as important as capability when it comes to operationalising AI. Without that visibility, compliance, quality control and even basic debugging become harder, which in turn slows adoption — the exact friction Atlassian is positioning this to solve.

The Renascence take

The interesting signal here isn't the tooling itself but what it reveals about where AI adoption typically stalls in large organisations — not at the point of capability, but at the point of trust and accountability.

Most coverage of enterprise AI tools focuses on what the AI can do; the real friction point is almost always what happens after — who is accountable when something breaks, and who gets credit, blame or context when it doesn't. Treating AI agents as identifiable "team members" with scoped authority and reviewable output is a service-design principle as much as an engineering one: it mirrors how good organisations handle delegation and escalation with human staff. Leaders rolling out agentic AI should ask not "can it code" but "can we explain, six months from now, exactly what it changed and why" — because that question, not raw productivity, is what actually unlocks sustained adoption.