Insights archive
Every analysis we have published, newest first. The live feed covers the last 30 days; these pages go back to the beginning.
1504 stories · page 29 of 76
Gartner: Over 40% of AI Agent Projects Will Be Canceled
AI tools that analyze company data often sound confident while answering the wrong question, and Gartner predicts more than 40 percent of AI agent projects will be scrapped by 2027 because businesses never taught the AI their internal rules and judgment calls.
· 2 min read
Bersin: AI Agents Could Cut HR Headcount 30-50% by 2030
A new report from The Josh Bersin Company projects HR departments could shrink by 30 to 50 percent in headcount by 2030 as AI agents take over administrative tasks, but both Bersin and Gartner say full replacement is unlikely because someone still has to manage the agents and fix them when they fail.
· 2 min read
Meta Tests Robots to Replace Data Center Workers
Meta is testing robots from Kinova, ABB, and Watney Robotics to power cycle servers and swap cables, and one worker says a successful rollout could take over up to 80 percent of some jobs as the company's AI spending climbs past 70 billion dollars a year.
· 2 min read
Giving AI Agents Names Makes Staff Miss More Errors
A Boston Consulting Group study of more than 1,200 managers found that giving AI agents human names and job titles causes employees to catch fewer mistakes and pass more bad work along to colleagues.
· 2 min read
AI Agent Traffic to Websites Grew 7,851% in 2025
AI agent traffic to websites grew 7,851 percent in 2025, and businesses that simply block this traffic risk turning away real customers along with attackers.
· 2 min read
Meta Launches Muse Image, Priced at $0.01 a Photo
Meta launched Muse Image, an AI tool for creating and editing pictures, through a developer service priced at one cent per image, cheap enough to make bulk ad and catalog image production affordable, though the same technology already faces privacy complaints over generating images of real Instagram users without their consent.
· 2 min read
ChatGPT Can Now Take Direct Actions on Websites
OpenAI added support for a new standard called WebMCP that lets ChatGPT search, book, and buy directly through actions a website exposes, instead of clicking through pages like a person would.
· 2 min read
Anthropic Launches Standard for AI to Control Machines
Anthropic released the Model Hardware Standard, letting AI agents like Claude directly operate lab and factory equipment such as robotic arms and microscopes, cutting setup time from months to hours.
· 2 min read
Salesforce Puts CRM Data Directly Inside Claude Chatbot
Salesforce and Anthropic launched Claudeforce, a plugin that lets salespeople manage Salesforce records directly from inside the Claude chatbot, as Salesforce fights investor fears that AI agents will make traditional business software obsolete.
· 2 min read
School Districts Keep Humans in Charge of AI Hiring Tools
K-12 HR leaders are using AI to speed up hiring paperwork, records cleanup, and pay calculations, but they are deliberately keeping people in charge of the final decision, a defensive move that mirrors the legal trouble companies like Workday now face over AI-driven hiring discrimination claims.
· 2 min read
Claude, Codex Ran Unclaimed Code Inside Fortune 500s
Security researchers found that AI coding assistants such as Claude, Codex, and Hermes will automatically fetch and run software listed in a website's machine-readable guide file, and when they registered a few of the missing package names themselves, real Fortune 500 companies triggered the trap within an hour.
· 2 min read
Photoshop Now Lets You Draw AI Edits on Photos
Adobe added a drawing-based AI editing tool called Markup to Photoshop, along with Lightroom-style lighting controls and a simplified prompt-only editing mode, as it races to compete with Canva and Google's AI image tools while raising subscription prices.
· 2 min read
Plaud Launches AI Earbuds With Built-In Cell Connection
Plaud has launched Plaud One, $249 earbuds with a cellular-connected case that can record and transcribe meetings and calls even without a phone nearby, entering a booming but legally risky market for AI note-taking hardware.
· 2 min read
OpenAI Starts Showing Ads to Free ChatGPT Users in India
OpenAI will show ads to ChatGPT's free and low-cost Go users in India, a country with over 100 million weekly users, as the company races to build new revenue streams before a planned stock market listing.
· 2 min read
US Worker Survey Finds Just 3% Lost Jobs to AI
A new survey of 1,250 US workers finds only 3% say they lost a job to AI since 2023, while company layoff announcements increasingly blame AI for cuts, showing a gap between what workers experience and what employers say.
· 2 min read
China Launches Youth Job Programs as AI Reshapes Work
China is rolling AI into every sector of its economy while youth unemployment sits near four times the adult rate, and its new five-year plan bets that job programs and internships can close that gap before it widens.
· 2 min read
Relay Launches Voice Typing App, Mic Ships 2027
Relay, a London startup founded by two former Nothing employees, launched voice dictation software today that can also act inside other apps, with a dedicated microphone accessory arriving in early 2027 for $150 plus a $15 monthly fee.
· 2 min read
Anthropic Adds Built-In Browser to Claude's Desktop App
Anthropic has built its own web browser directly into the Claude desktop app so its Cowork assistant can log into websites, click through forms, and pull data from portals that have no software connection, all inside a separate window that never touches your own tabs or passwords.
· 2 min read
Most Firms Using AI See No Productivity Gain Yet
A wave of 2026 surveys, including one covering nearly 6,000 executives, finds that most companies now use AI but only a small fraction have seen it actually improve productivity, and the difference comes down to redesigning how work gets done rather than which tool gets bought.
· 2 min read
MIT Test: Priciest AI Model Performed Worst on Invoices
A real-world test found the most expensive AI agent setup solved fewer invoices correctly than a mid-priced one, showing that paying more for AI does not guarantee better results.
· 2 min read