The world of AI-powered coding saw major updates this week. Enterprise tools led the charge, with Microsoft hosting an AI Agents Hackathon featuring workshops on their Semantic Kernel framework for team-based AI systems. Meanwhile, Google revealed major upgrades to Gemini Code Assist, letting it build apps from product descriptions and convert code between languages automatically. A new Kanban board helps developers track these AI agents' progress on complex jobs.

Developers are adapting fast – Salesforce reports 28% of new code is now AI-generated, and Shopify made AI mandatory for all employees to speed up work. Tools like Agentforce show why: they helped Secret Escapes build customer service bots in two weeks instead of six months. Low-code platforms let even beginners create AI helpers through simple conversations.

Schools and researchers aren’t left out. Clarivate launched AI Agents for academic work, helping write papers and manage data while keeping human oversight. Their new Agent Builder tool lets schools create custom AI helpers without coding.

Security challenges came into focus as Astrix released tools to protect non-human identities (NHIs) used by AI agents. Since these bots need special access to work, hackers could steal their permissions to cause big breaches. The OWASP framework now includes NHI security guidelines to address these risks.

While AI speeds up work, quality checks remain key. Meta faced criticism for gaming AI benchmarks, and tests showed even top tools like Devin only completed 15% of complex tasks. Experts remind teams to always review AI-made code for errors before use.

The big takeaway: AI agents are becoming essential teammates, but need careful management. From classrooms to big companies, these smart helpers are changing how we build software – making it faster but requiring new safety practices.