The month AI agents ran through everything

One theme tied October together: the word “agent” stopped being a demo and became a default setting. AI agents showed up in the checkout, in the ad account, in the project tools and in the frontier models — often as something switched on for you rather than something you chose. Here is what actually moved across eCommerce, marketing, delivery and AI this month, and why each one matters.

A $17B merger, and marketplaces push back on AI

The biggest commerce story was a mega-merger. The quieter, more telling one was platforms drawing hard lines around AI content and counting the cost of the build-out.

• DoorDash is buying Instacart for about $17B. Announced 29 October at $65 a share, a 26% premium, with closing expected in late 2027 or the first half of 2028. Instacart’s share of US delivery spend has slipped from 26% to 18% in five years while DoorDash’s grew from 44% to 59% — consolidation in a market that is getting harder to win alone. — AP via ITV
• eBay cut about 1,000 jobs — roughly 9% of its workforce, and its second round of layoffs this year. — Retail Dive
• Etsy drew a line on AI. Most seller listings now need at least one authentic, non-AI image, and AI-generated recipes and patterns are banned — a marketplace betting that “made by a human” is still a selling point. — CNBC
• Fanatics added OpenAI to “Shop the Look.” A pilot of the outfit-curation tool showed 10% higher engagement — AI recommendation moving from the big players down to category retailers. — Retail Dive
• Amazon’s Q3 showed what the AI build-out costs. Free cash flow was negative $42.4B, and 2026 capex was held unchanged — a reminder that the agent era is being paid for up front. — CNBC
• Social checkout is getting closer. Meta is testing direct checkout inside its Messenger business agent, powered by Shopify — the storefront moving into the chat window. — ALM Corp

The ad platforms hand the controls to AI

Meta, Google and Microsoft all pushed more of the ad account into automation this month — and quietly shifted more of the responsibility onto you.

• Meta changed a lot in one month. From 6 October it began testing agentic actions in its business assistant, which can create campaigns, change targeting and adjust budgets from a chat prompt. On 15 October it announced Ads Super Burst, First Impression Ads and wider Reels Trending Ads, and on 26 October Threads became a default ad placement globally. — Good Morning Co.
• Meta’s late-October deadlines carry real risk for advertisers. Product browsing switched on for existing ads on 28 October unless you opted out, and new Commercial Terms took effect on 30 October — making advertisers responsible for what Meta’s AI changes in their ads. Worth reading before it changes something for you. — Amax Agency
• Google pushed more automation too. Google Ads is moving eligible Search campaigns to AI Max, and from 12 October began pulling coupons from advertiser sites into campaigns automatically. Its September spam update finished rolling out on 8 October. — Digital B
• Microsoft Advertising dropped Max CPC as an option for new campaigns using automated bidding strategies on 1 October — one more manual lever quietly retired.

Every project tool now ships an AI agent

If you run delivery, this is the shift to watch: the agents are now inside the tools your team already uses, and the pricing model is changing with them. It is exactly why I built out a project and delivery practice — the hard part is governing these agents, not buying them.

• Every major tool is now selling AI agents. Built-in agents now exist in Asana, Jira (Rovo), ClickUp, monday.com, Planner, Notion and Wrike — and pricing is shifting to credits, so “how much does it cost at 10x usage?” becomes a real question. — Epicflow
• monday.com held Elevate in NYC on 28 October, announcing “monday Frontier,” which pairs managed services with forward-deployed engineers. Gartner also named it a Leader in collaborative work management for the fourth year running. — UC Today
• Atlassian held Team ’26 Europe in Amsterdam on 6–8 October, and on 27 October let engineers call a coding agent from Jira through Rovo Chat — the project tool and the IDE starting to blur. — UC Today

Google’s biggest model month, and a $740B build-out

The frontier labs and their suppliers all moved at once — and the scale of the spending behind it came into focus.

• Google had the biggest model month. It launched Gemini 4 Argon globally, then Gemini 4 Carbon two weeks later for long-horizon engineering and cyber defence, with Carbon in government pre-release testing. It also introduced Genie 4, a real-time world simulator, a SynthID Detector for spotting AI content, and a “Gemini agent” for work on 8 October. — Google
• Hardware and open models moved too. Nvidia’s RTX Spark laptops launched from $2,399, and Meta released Muse Spark Open, an open-weight model for Macs — capable AI edging onto local devices. — AI News Archive
• The spending is staggering. Alphabet, Amazon, Meta and Microsoft have outlined over $740B in capex and raised $326B in debt to fund the build-out — the bill behind every “free” agent feature you were handed this month. — Dentro

The takeaway

Put the four stories together and the pattern is clear: agents are being switched on inside the tools you already pay for — your checkout, your ad account, your project board — and the terms are quietly shifting the responsibility onto you. The practical move is the same every time: before you let an agent near your budget, your listings or your customers, know exactly what it is allowed to touch, and check what changed by default this month.

If any of this bears on a decision you’re weighing — an ad-account change, a store that needs to be readable by shopping agents, or a delivery team drowning in new tools — get in touch and we’ll talk it through. See you next month.
— Deepak