The ai models & agents stories our desk judged worth your time, written so anyone can follow them.
Claude now has add-ons for marketing work, including one that reads your HubSpot customer data. They need a paid plan, and one setting is worth checking first.
Read more: Claude add-ons can now handle your marketing tasks and company dataOpenAI says its Dots assistant should reach more of Europe in a few weeks. The Berlin event also showed how much OpenAI wants businesses to build on its platform.
Read more: OpenAI says its Dots assistant will reach more of Europe soonGitHub now logs 7.4 billion saved code changes a month, mostly from AI agents, and is rebuilding its systems to keep up. Outages have been a recurring problem.
Read more: GitHub now gets 7.4 billion code saves a month, mostly from AITwilio published a guide to AI agents that contact customers first about problems like a late order. Here is how it works and the consent rules to settle first.
Read more: How to build an AI agent that calls customers before they have to call youAI tools that shop and pay for you may need your habits, addresses and card details. Whether privacy fears will stop people from trusting them is still open.
Read more: Privacy worries could slow AI shopping assistants that buy for youStripe now offers virtual cards that AI shopping assistants can use to buy for you, and Visa and Mastercard are reportedly uneasy about it.
Read more: Stripe lets AI shopping assistants pay with one-time virtual cardsGitHub's new Copilot Fusion picks and combines AI models for each coding job. Independent tests found it cost about half as much as a top rival and scored close.
Read more: GitHub's new AI coding tool nearly matches the best for half the costGoogle reported record search growth, but its search revenue missed estimates and shares fell about 10%. Ads in AI answers are the fix, so expect more of them.
Read more: Google gets more searches than ever, but earns less from each oneA New York firm launched a free tool that shows whether AI chatbots name your company or only your brands. Its L'Oréal test found most answers skipped the parent.
Read more: Free tool shows how AI chatbots link companies to their brandsArticul8 released Bivia, a model that makes the small yes-or-no calls inside AI agents, aiming to cut the cost of running them. The test results are the company's own.
Read more: Articul8 launches Bivia, a cheaper AI that makes decisions for agentsChinese AI labs are reportedly leaning on OpenAI's newest model to train their own, which could shape the price and quality of the AI tools you use.
Read more: Chinese AI labs are copying OpenAI's newest model by studying its answersEmpacer, a new AI assistant for small business owners, turns a spoken request into a draft post or reply, and nothing goes out until the owner approves it.
Read more: New AI assistant lets small business owners speak a task and approve the resultAutoScheduler.AI released warehouse software that plans crew sizes and shows why work ran late, so managers can fix the cause instead of just adding staff.
Read more: New warehouse software plans worker schedules and shows why work ran lateMicrosoft says big companies are using AI to run real business tasks, and that its pricing is shifting toward charging for how much work the AI does.
Read more: Microsoft says big companies are putting AI to real workAmazon Quick now lets a second AI model approve routine agent actions on its own, while risky ones like payments and deletions still go to a person.
Read more: Amazon's AI assistant now skips routine approvals, asks about risky onesA consultant says one company ran up about $500 million in Claude charges in a month. The figure is unconfirmed, but the billing lesson applies to any business using AI.
Read more: Company Reportedly Ran Up a $500 Million Claude Bill in One MonthGoogle says shopping and task helpers are coming to the Gemini app's 1 billion users later this year, which could change how people buy online.
Read more: Google wants AI helpers to book, buy and fill out forms for youCircle and Proof showed a test where an AI agent proves who approved a purchase, then pays in digital dollars. It could decide whether stores let AI shoppers in.
Read more: Circle and Proof test a way for AI shopping agents to prove who they work forAnthropic says its Claude AI is helping speed up early medicine research, but any real benefit for patients is still years away.
Read more: Anthropic says its Claude AI is helping speed up new medicine researchCloud security company Upwind raised $400 million at a $3.8 billion value, and says its revenue quadrupled in 2026 as AI changes how companies work.
Read more: Cloud security firm Upwind raises $400 million at $3.8 billion valueMicrosoft is offering voluntary buyouts to more than 60,000 employees, a sign of how AI is changing where tech companies spend their money.
Read more: Microsoft offers buyouts to 60,000 workers as AI costs climbOpenAI added iCloud Mail and Calendar connections to ChatGPT on October 27, 2026, but only some Apple addresses work and privacy questions remain.
Read more: ChatGPT can now read your iCloud email and calendarCopilot is reported to offer 11 Claude models from Anthropic while also competing with Anthropic's own coding tool, Claude Code. Here is what that could mean for buyers.
Read more: Copilot Offers 11 Claude AI Models and Rivals Claude CodeAnthropic says problems with Claude's Gmail and Calendar connections are fixed. If you let Claude read your email or calendar, the status notes may matter.
Read more: Claude's Gmail and Calendar connections had problems, now fixed
