AI coding assistants shift developer labor from writing to reviewing

The adoption of AI coding assistants is fundamentally altering the developer's workflow, trading the act of creation for the labor of auditing. While these tools accelerate the production of raw code, they have not linearly reduced the total workload for software engineers.

Miengha Behardien, a senior developer at the Cape Town-based custom software firm Khanyisa Real Systems (KRS), reports that the time saved during the generation phase has simply migrated toward reviewing, testing, and validating AI-produced output. "AI has made output a lot faster, but the time I saved writing code has shifted to verifying the code it produced," Behardien told ITWeb. To manage this shift, KRS has implemented "adversarial QA," a process where a different AI model is used to review code than the one that originally generated it to better identify errors. Behardien specifically cited Claude Code as their preferred AI coding companion.

The rise of code bloat

This shift in labor is driven by the inherent nature of large language models, which can perpetuate human errors found in their training data and often produce inefficient results. Behardien highlighted "code bloat" as a primary concern, noting that AI often takes the longest route to a solution rather than the simplest. This tendency toward verbosity can lead to maintainability issues and increased technical debt. According to Sonar's 2026 State of Code Developer Survey, 40% of developers believe AI has increased technical debt by generating unnecessary or duplicate code.

Redefining productivity

These findings challenge the prevailing industry narrative that AI leads to a direct reduction in development time. For businesses, the distinction is critical: faster code generation does not automatically translate to faster software delivery. As the volume of generated code grows, the cost and complexity of quality assurance and engineering oversight may actually increase. The bottleneck has moved from the keyboard to the review screen, meaning the human element of engineering remains the primary safeguard for software stability.

The future of agentic development

In response to these challenges, firms like KRS are evolving their internal processes. Rather than focusing solely on prompt engineering, the company is adapting its junior developer training to emphasize "agentic development" and responsible AI usage. The goal is to ensure that new engineers view AI as a tool requiring rigorous oversight rather than a replacement for fundamental architectural thinking. The industry now faces a broader question of whether AI will eventually optimize the review process or if the burden of validation will continue to scale alongside the speed of generation.