Our news
-

The upgrade is not a toggle
Google is mid-September progressive auto-migration of Search campaigns still on campaign-level broad match or standalone ACA into AI Max. If you never clicked the feature, your settings may already have changed. This week’s job is a settings audit, not a yes/no toggle.
-

The catalog is the campaign
OpenAI named Shopify its first ecommerce partner and shipped a ChatGPT Ads app that syncs catalog plus Pixel commerce events into Ads Manager. Campaign creation is selecting from a feed. Staff it like another creative channel and you buy the wrong OS.
-

The ad is the associate
Google’s Sep 16 holiday update puts Business Agent inside eligible U.S. YouTube Demand Gen ads. The unit becomes an in-stream product associate, not a click-out to a landing page. Score it like CTR + LP conversion and you will misread the beta.
-

Projects are the runtime
Anthropic relaunched Claude Code Projects on September 17 as a coordinator plus parallel cloud threads, with merge conflicts as the honesty layer. The product you buy is the multi-agent runtime and shared project state, not a prettier file tree.
-

One Claude is the workspace
Anthropic merged Claude chat and Cowork into one Claude and shipped beta Docs and Slides on September 16. That is a SKU and export-path decision for marketing and ops teams, not a cosmetic tab cleanup.
-

The CRM is the ads console
OpenAI named HubSpot its first CRM partner for ChatGPT Ads on September 16. If you cannot trace closed-won back to ChatGPT Ads spend inside the CRM, you bought another dashboard, not a measurable channel.
-

Live is two SKUs
Gemini 3.8 Live and 3.8 Live Extended Thinking are a dual voice-agent SKU (scale and cost versus high-complexity speak-and-reason), not one Live model. Buy the wrong mode and you inherit the wrong latency, tool-calling, Workspace surface, and enterprise preview limits.
-

The mix model is not a report card
Google’s Meridian upgrades (agentic data-quality help, branded query volume, GeoX GA) pull audit, brand signals, and causal geo priors into one measurement stack. Teams still treating MMM as a quarterly slide deck will miss that the same loop should change budget this week.
-

The reasoning stack is the SKU
Salesforce Koa is not another Agentforce job pack. It is a CRM-specific reasoning model Salesforce owns and runs inside its trust boundary, post-trained from NVIDIA Nemotron on synthetic CRM scenarios. The Agentforce question just moved from which named role you buy to which reasoning stack sits under that role.
