AI-driven product development
Atlassian is especially well-positioned to understand how developers use AI. They’re using that understanding to redesign their own product development lifecycles. The upshot: AI alone won’t do what you need it to; you must redesign your product development processes around context and human judgment.
Goldman Sachs pulls Claude Code
Software ate the world — and that was before AI made it incredibly easy for anyone to make software. But it’s becoming increasingly clear that having everyone write software isn’t tenable for many businesses. Goldman Sachs — which is underwriting Anthropic’s IPO — recently restricted its traders and bankers from using Claude Code. Why? Because agent-built software introduces significant risks.
The “Human-in-the-Loop” Fallacy
It’s become a cliché: have AI do the work but keep humans around to check the output. But that may be a losing strategy, since it focuses valuable human attention on tasks that have diminishing value. Better to reinvent the business altogether around the new capabilities AI enables. (HBR gift link)
Wise AI use
Business leaders are starting to realize AI isn’t a magical technology that replaces human labor and accelerates value creation with minimal input. Like all prior technologies, AI must be used wisely. The real productivity gains are to be had by moving past individual AI use, and that requires redesigning business workflows.
Multi-user multi-agent collaboration
What’s the most effective way for humans to collaborate using AI on shared resources? A team of researchers evaluated different approaches, including peers working through a single coordinating agent and peers working with one agent per user. The results have important design implications: it’s not enough to give separate agents access to shared resources; they also need coordination rules and guardrails.