Essential reading for your AI journey

AI is not getting more capable slowly. It is accelerating. ChatGPT Business can now draft your emails, create documents, and book meetings. Write actions mark a shift from AI that advises to AI that acts. The businesses that benefit most will not be the ones resisting this. They will be the ones building the fluency to guide, edit, and review AI well. Here is what changed, why the default-off design is the right call, and three steps admins should take this week.

Anthropic has launched the Claude Marketplace in limited preview. Enterprise customers can now use part of their existing Anthropic commitment to buy selected Claude-powered partner tools. Launch partners include GitLab, Harvey, Lovable, Replit, Rogo and Snowflake. Anthropic handles invoicing for partner spend.

OpenAI's new Skills feature lets teams package AI workflows into reusable processes. It is a significant step forward for ChatGPT, but Claude users have had access to this approach for some time.

OpenAI has released GPT-5.4 with stronger document and spreadsheet handling, native computer use, and direct integration with Excel and Google Sheets. Here is what actually changes for professional work.

OpenAI has introduced Lockdown Mode and “Elevated Risk” labels in ChatGPT to reduce prompt injection risk as AI tools connect to the web and apps. For owners and heads of operations, it is a useful signal of where extra guardrails belong before you scale adoption.

AI research tools are only as useful as the sources they’re built on. If the inputs are noisy, biased, or outdated, the output looks confident but drifts away from what you actually need. That’s why the latest improvements to Deep Research in ChatGPT matter: they’re less about “more AI” and more about better governance–clearer sourcing, tighter focus, and easier accountability.