AI agents are undeniably powerful, but wielding that power responsibly is another story. This month’s picks dive into the real-world dos and don’ts of generative AI: Best practices for using genAI tools for code generation, how to avoid the trap of over-relying on a single tool, and why so many agents fail at understanding business context. And if you get things wrong? One company is offering a new ‘undo’ option for AI mistakes.

Top picks for generative AI readers on InfoWorld

A developer’s guide to code generation
According to a recent survey, 91% of developers are using AI for code generation. But how many are doing it right? This article features real-world talk from experts about how developers can use genAI and avoid its pitfalls.

Multi-agent AI workflows: The next evolution of AI coding
One emerging best practice of the genAI era is: Don’t treat any single tool as the solution to every problem. Just as humans have specific strengths, so do AI coding assistants.

Why AI fails at business context, and what to do about it
As organizations start turning AI agents loose on messy, real-world problems, it’s become clear they tend to miss important business context. Columnist Matt Asay says this is an engineering problem, not a philosophical one. Developers (and others) can solve it by implementing AI with care.

Is the generative AI bubble about the burst?
Developers have a front-row seat on the rollercoaster ride of generative AI. How we use it in our daily work tells us a lot about where the AI boom is headed.

GenAI news bites

• Rubrik unveils ‘undo button’ for AI agent mistakes
• AWS blames bug for Kiro pricing glitch that drained developer limits
• Claude Sonnet 4 upgrade enables full codebase processing in a single request
• Google rolls out AI coding tool for GitHub repos
• Anthropic targets DevSecOps with Claude Code update
• JetBrains previews no-code app builder

More good reads and generative AI updates elsewhere

NIST’s attempts to secure AI yield many questions, no answers
NIST is taking an AI security approach that involves tweaking its current rules to accommodate AI, rather than starting from scratch. Does it go far enough?

Companies are pouring billions into AI—but it has yet to pay off
Eight in 10 companies have reported using generative AI, but just as many have reported “no significant bottom-line impact.”

Why you can’t trust a chatbot to talk about itself
When an AI tool does something surprising, your first instinct is probably to ask it why. But chatbots and other LLMs aren’t capable of introspection, and the answers they give can range from unhelpful to actively misleading.