Meta published the prompts SMBs actually run through its Muse agent. Three patterns, and what each one means for agency workflows and reporting.

ChatGPT Ads reached 89 countries, a new survey shows consumers will let AI shop but not pay, and voice commerce is hitting a trust ceiling. What agencies should change this week.

Tiger Data's new PRISM index in pg_vectorsearch makes Postgres vector index rebuilds dramatically cheaper. What it changes for RAG builds, and where it does not fit yet.

Language was chosen once at the phone menu. New streaming voice models detect it every turn, and most agency call flows are still built around the old assumption.

Vendors are shipping MCP servers fast. The question that matters for agencies is not whether it integrates, but which human seat the agent inherits — and what that identity can touch.

Google's Demand Gen cost adjustment beta can now run 60 days, and the English and UK help pages disagree. What to change in your reporting automation.

Typeface finds 49% of marketers plan to cut SEO spend as AEO takes the budget. Plus agent traffic up 4.7x, Peec AI's ARR jump, and TELUS-Parloa.

Prompt-prefix caching is now automatic at the model API. The savings left sit in semantic caching: matching the same question asked a different way. How to size it, pick the threshold, and dodge the failure modes.

Voice quality is table stakes now. The week Decagon, Amazon and Microsoft all shipped the fix for dead air mid-call, and the acceptance tests agencies should run before promising a resolution rate.

Last week Google did for enterprise AI budgeting what most agencies still have not done for themselves: it installed a meter. At Gemini at Work 2026 on October 8, Google Cloud introduced a single “universal” Gemini agent that plans work,…