Why AI Won’t Replace Developers: Thoughtworks CTO on Legacy Code Modernization

Thoughtworks chief technology officer believes that although artificial intelligence is driving a fundamental transformation in the tech industry, we must retain developers.

Laycock recounts a story where an AI‑generated app was praised for its speed, yet within 24 hours the software was compromised because it lacked basic security considerations, highlighting the limits of AI‑only development.

She stresses that the biggest challenge facing the industry is legacy code; as AI‑generated code proliferates, the amount of technical debt will only worsen.

AI does help repay technical debt and aid cloud migration, but it also increases demand for engineers who can think deeply and solve problems.

Artificial Intelligence and Legacy System Modernization

Many organizations bet that AI agents will become increasingly intelligent, hoping retrieval‑augmented generation (RAG) will accelerate model and tool improvement, yet we are still far from understanding AI’s long‑term impact.

Legacy system modernization remains the biggest challenge for most enterprises, compounded by fragmented knowledge across hundreds of applications.

Both business and technical teams struggle to comprehend code, and the more AI‑generated code there is, the harder it becomes to understand.

The Real Cost of Trying to Replace Developers with AI

Laycock observes that many companies aim to reduce developer headcount by deploying AI agents, but the cost of such agents can reach tens of thousands of dollars per developer per year, far exceeding the $100 per‑developer cost of tools like GitHub Copilot.

There is no proven economic model for large‑scale AI deployment, and the security of AI tools—such as malicious npm packages or credential‑stealing backdoors—remains uncertain.

She warns that using AI agents without proper oversight can lead to infinite loops and wasted token spend, urging a balanced approach that leverages AI’s strengths without discarding human expertise.

CodeConcise: Solving Legacy Code Mysteries

Thoughtworks is building a generative‑AI tool called CodeConcise Legacy Assistant, which indexes code, provides contextual windows, and adds a conversational AI layer to help clients understand their systems.

The tool is not a silver bullet; it must be combined with deep knowledge of mainframe code to migrate to cloud‑native environments.

Laycock emphasizes a “thin‑slice” interactive method that identifies exploitable domains, creates seams, and iteratively removes dead code.

Recruiting: More Developers for AI Experiments

While AI code generation is still nascent, solving legacy problems is a more impactful use case for generative AI in enterprises.

AI can aid in adjusting data architectures to support AI applications, but it will not magically solve complex challenges without human involvement.

Thoughtworks is crowdsourcing hypotheses about the future of the tech industry and advocates a product‑thinking mindset throughout the software development lifecycle to test and validate AI solutions.

“Artificial intelligence is permeating all of these areas, which is great, but we must remember it is not deterministic.” – Rachel Laycock