Thoughtworks on AI‑augmented engineering

Thoughtworks is telling software teams to treat artificial intelligence as part of the delivery system, not just a faster autocomplete tool. (thoughtworks.com) In its “AI-first software engineering” report, Thoughtworks says the biggest changes are happening across the whole software lifecycle, not only in code generation. The report says teams are using AI for error detection, knowledge transfer, legacy modernization and exploratory testing as well as coding. (thoughtworks.com) Thoughtworks ties that message to its September 25, 2025 reading of the 2025 DevOps Research and Assessment, or DORA, report. Chris Westerhold, the company’s global practice director of engineering excellence, said organizations seeing returns from AI start by rebuilding the platforms, workflows and team structures that AI depends on. (thoughtworks.com, thoughtworks.com) The basic idea is simple: an AI coding assistant predicts likely code the way a phone predicts the next word, and an AI agent can take a longer chain of actions inside developer tools. Thoughtworks says those systems work better when teams already have clean code, solid internal platforms and clear review practices. (businesswire.com, thoughtworks.com) That framing reflects a shift in Thoughtworks’ public guidance over the last two years. In April 2024, its Technology Radar said AI tools were spreading across testing, documentation and refactoring; by April 2025, the company was emphasizing supervised agents in integrated development environments and warning against complacency with AI-generated code. (nasdaq.com, businesswire.com) The newer reports also narrow the measurement question. Thoughtworks says leaders should look beyond raw productivity and track process outcomes such as throughput, quality, developer experience and whether teams can transfer knowledge and detect errors earlier. (thoughtworks.com, thoughtworks.com) Thoughtworks is not arguing that AI replaces engineers. Its report says engineers spend more value on prompt design, architecture decisions and checking machine-generated output, while weak systems can become “a faster way to create chaos” when AI increases delivery speed without fixing instability. (thoughtworks.com) The company’s own April 2025 Radar used similar language from Chief Technology Officer Rachel Laycock, who said companies need a pragmatic approach to adoption and still have to invest in software delivery basics such as observability and data management. (businesswire.com) The through line in Thoughtworks’ recent research is that AI changes software engineering most when teams change how they build, review and run software. The tools may write more code, but Thoughtworks’ advice is to fix the system around them first. (thoughtworks.com, thoughtworks.com)