GS Shop has launched an MD curation system it built with AWS. When an MD enters the planning goals and conditions, AI analyzes sales records and customer behavior data to pick products, then helps set up and register the promotion. During a one-month trial in October, preparation time dropped 70% compared with before.
Elon Musk, CEO of SpaceX, cited high operating costs as the reason for reducing usage limits on the Grok AI agent. He said that although the number of users has grown, the cost of providing the service is far too high compared with its revenue.
LangChain has released Managed Deep Agents in public beta, a service that lets developers run AI agents directly in the cloud. Developers only define the agent's instructions and tools, while LangSmith handles server setup and operations. You can create a single folder and deploy with a few commands.
In the preview version v0.2.1-alpha.1 of DeepSeek Harness, the plugin management screen now has an "Create plugin with agent" option. Selecting it sends an existing draft to Creator Mode, and nothing runs until the user submits a request. Harness is still in developer preview.
Meta's personal AI agent, Muse, buys and chooses products on behalf of users, and analysts say this could hurt Amazon's advertising business, which earns money from product ads. Since its launch in September, Muse has been used by more than 10 million people in the United States and Canada in about six weeks.
Only 48% of insights created by AI agents actually helped improve business. Companies surveyed pointed to a lack of business context and scattered data as the main causes. This report looks at what companies are doing to raise the quality of their insights.
Superhuman released Coursework Helper in beta, an AI agent for students inside Superhuman Go. Once students connect their school's Canvas learning management system, it helps them plan the week, rank assignments, and prepare for exams.
AI agents call models many times for a single coding task, so token costs (the smallest units AI uses to read and write text, which determine billing) can climb quickly. This article argues that companies should judge AI spending by the cost of each merged code change, not by total usage.