Welcome back to LLM Decode 👋
AI is moving out of the chatbot window and into the software where work actually happens. Claude is now working directly inside Google’s core productivity apps, while Meta and Sierra are building a standard that could let personal AI agents interact directly with businesses.
Here’s what matters today.
Claude Just Moved Into Your Google Workspace
Anthropic has launched Claude for Google Workspace in public beta, bringing Claude directly into Google Docs, Sheets and Slides. A sidebar lets users work with Claude alongside the file they already have open, rather than constantly switching between an AI chatbot and their work.
In Docs, Claude can draft, rewrite and restructure content. In Sheets, it can create formulas, build tables and analyze data. In Slides, it can create slides using existing layouts and help refine presentations.
Anthropic is also launching separate Google file connectors. That means users can start inside Claude, provide a Google file, and have Claude create or edit the document, spreadsheet or presentation while respecting the user's existing Google permissions.
Why it matters
AI is moving into the workflow: Users don't have to copy-paste information between apps and chatbots.
Claude is competing directly with Gemini: Anthropic is putting its AI inside Google's own productivity ecosystem.
Context becomes more valuable: AI can work with the actual document, spreadsheet or presentation rather than a pasted excerpt.
Human control remains built in: Users can review changes before Claude applies them, making the workflow more practical for professional work.
AI Agents Are Getting a Direct Line to Businesses
Meta and Sierra have introduced the Personal Agent Protocol, an open standard designed to define how personal AI agents interact with businesses. The initiative is being developed with partners including Shopify, Stripe, Walmart, Genesys, Instinct and Rocket.
Today, an AI agent often behaves like a human user: it opens a website, clicks through pages and fills out forms. The new protocol aims to create a more direct connection between the agent and a company's systems, potentially allowing tasks to be completed faster and more reliably.
The protocol is designed around authentication, permissions and business-defined controls. Customers can decide whether their personal agent has read-only or write access, while businesses can determine which actions their systems allow. The first version of the specification is expected later this month.
Why it matters
AI agents could become the new customer interface: Instead of visiting ten websites, users could ask one agent to handle the work.
Shopping could become agent-first: Product discovery, purchases, returns and support could increasingly happen through personal AI.
Brands will need to become “agent-ready”: Companies may need APIs, permissions and systems designed for AI customers.
Marketing could change dramatically: Winning the customer may eventually mean being recommended by their AI agent, not just appearing at the top of a search page.
4 AI Workflows to Try
Work Directly Inside Your Files: Instead of copying a document into an AI chatbot, use an AI assistant where the document already lives. Ask it to analyze, rewrite or restructure specific sections while preserving the original context.
AI-Powered Report to Presentation: Give AI a report or research document and ask it to identify the key findings, build a narrative and turn those findings into a presentation. Review the structure before finalizing the deck.
Design for an “AI Customer”: For an ecommerce or service business, map what an AI agent would need to know: products, prices, availability, returns, policies, reviews and customer permissions.
Build an Agent-Ready Workflow: Identify one customer journey that currently requires multiple website visits. Document the information and permissions an AI agent would need to complete it, then look for opportunities to expose that information through APIs or structured data.
Ready to level up your AI skills?
Explore Our Courses →That’s it for today.
The AI space doesn’t slow down - and neither should your thinking.
See you in the next drop.