From AI Copilots to Agent Swarms
AMD has surpassed its target for AI-driven productivity gains, according to an account from Andrej Zdravkovic, a senior vice president at the company. AMD had aimed for a 25 percent productivity boost over two or three years; one year later, it reports a 30 percent overall boost through AI.
AMD began developing AI systems for code generation, testing automation, bug analysis, and code review in 2024, with an objective of 25 percent AI-generated production code by 2027. The company tracks the share of source code generated by AI, counting only code that passes all reviews and testing and ships in the final product. AMD says it crossed the 20 percent mark at the beginning of this year, is progressing toward 50 percent across its entire codebase, and that some software components now have more than 80 percent AI-generated code.
Agentic AI is used across the lifecycle, AMD says: analyzing and triaging problem reports, debugging and generating code changes, creating unit and integration tests, and preparing architecture summaries, change reviews, and test results for engineer approval before release. AMD says it uses multi-agent workflows through agentic harnesses such as Codex and Claude Code while developing internal multi-agent systems.
The company cites its Radeon Software eXperience (RSX), a user interface for configuring and monitoring graphics driver behavior. In October 2025, AMD began using AI agents to automatically debug and fix reported RSX issues; out-of-the-box tools resolved only 6 percent. By June 2026, AMD says, the figure reached 75 percent. AMD attributes the increase to refining agent objectives rather than retraining underlying models, to a learning loop, and to advances in models and agent run-times.
Looking ahead, AMD describes collaborative AI agent swarms that independently identify and develop solutions, guided by humans on what to solve, then prepare ranked options with validation results and performance metrics for engineer review. AMD says its goal is to empower its workforce with AI, not to reduce headcount, and that it is investing in AI education and training.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
The impact of AI on software development has been both profound and ever-evolving. Last year, I wrote about AMD’s plans to use AI not just for generating new lines of code, but also for other steps in the software development lifecycle (SDLC), such as triaging problems, debugging code, and testing the software. At the time, we were hoping for a 25 percent productivity boost from AI use over the course of two or three years. But with each new release, the capabilities of large language models (LLMs) improve dramatic