Embed AI in the development lifecycle
At Globant, a $2 billion IT services firm, Juan José López Murphy, head of AI, describes the challenge bluntly: “The way software is built has fundamentally changed. If we don’t upskill, we risk becoming obsolete.”
Globant’s strategy is to embed AI not just into coding, but into every step of the software development lifecycle — planning, testing, monitoring, and deployment. Developers are trained to be fluent in all major large language models (LLMs) — ChatGPT, Claude, Gemini, Bedrock — and to use proprietary agents like Globant GPT for client projects.
The training itself is layered. Basic prompting covers clarity of objectives, context framing, and avoiding “context rot.” Context rot occurs when conversations with AI agents grow too long or unfocused, causing the system to lose track of instructions and generate lower-quality output. Developers are trained to recognize when to reset or reframe prompts so the AI can stay aligned with goals.
This story originally appeared on Computerworld
