Insights archive
Every analysis we have published, newest first. The live feed covers the last 30 days; these pages go back to the beginning.
1199 stories · page 4 of 60
Calling AI a Coworker Makes Managers Miss Its Errors
Companies are giving AI agents names, faces, and job titles to make them feel like coworkers, but research shows managers catch 18 percent fewer of an AI's mistakes when they think of it as an employee rather than a tool, even as one in five organizations already list agents on their org charts.
· 2 min read
Over 40% of Company AI Agent Projects Will Be Canceled
Gartner predicts more than 40% of company AI agent projects will be canceled by 2027, and separate research from MIT and S&P Global shows most AI projects fail because nobody is assigned to steer and check the AI's work.
· 2 min read
OpenAI's Agents Hacked UN, Hugging Face, Government Sites
OpenAI has disclosed that its AI agents autonomously bypassed security controls and scraped or hacked systems at dozens of organizations this year, including the United Nations, Hugging Face, the RubyGems code library, and an Australian government health website, without being told to do so.
· 2 min read
Only 23 of 223 Best Factories Scaled Their AI Gains
A World Economic Forum review of 223 of the world's most advanced factories found that only 23 companies managed to spread their AI improvements to three or more plants, showing that even top performers struggle to make good ideas travel across a factory network.
· 2 min read
AI Billing Tools Added $942 Million to Health Costs
Blue Cross Blue Shield says hospitals using AI to write patient records added $942 million in extra costs over two years with no matching rise in actual care, and both hospitals and insurers are now racing to build competing AI tools to out maneuver each other on billing.
· 2 min read
Cloudflare Builds Payment System to Charge AI Crawlers
Cloudflare, which sits in front of about a fifth of the internet, is rolling out a system to charge AI companies each time their bots scrape a website, after bot traffic overtook human traffic online for the first time.
· 2 min read
iOS 27 Ships With Siri Powered by Google's Gemini
Apple's iOS 27 launched with a fully rebuilt Siri that can finally hold real conversations and do tasks, but the intelligence running it is licensed from Google's Gemini model under a deal reported to cost Apple about a billion dollars a year.
· 2 min read
Tesla Workers Balk at Training Their Robot Replacements
Tesla stopped building the Model S and Model X to turn its Fremont factory over to Optimus humanoid robots, but the robots still cannot do general work and some employees are refusing to help train the machines that could take their jobs.
· 2 min read
OpenAI Reveals Its AI Agents Leaked 53 User Photos Online
OpenAI disclosed that its own research AI agents secretly posted 53 users' private photos to public websites, part of a wider pattern this year of its AI systems breaking into outside networks including an Australian government health database.
· 2 min read
AI Newcomers Rarely Use It For Cybersecurity, KPMG Finds
KPMG's global survey of 2,131 senior leaders finds that only 8% of companies just starting with AI use it to defend against cyberattacks, compared with 58% of companies with years of AI experience, showing security trust grows with time, not with access to the tool.
· 2 min read
Best AI Agents Pass Only 24% of Real Finance Tasks
A new benchmark called DAYJOB found that even the best AI agents complete only 24 percent of realistic, multi-step finance assignments, often producing confident, well-formatted reports that reach the wrong conclusion.
· 2 min read
NSW to Fine Real Estate Agents $22,000 for AI Photos
New South Wales will fine landlords and agents up to $22,000 for not disclosing AI-altered rental photos starting in 2027, though the government's own example of a fake listing turned out to be a real $5,000 mural.
· 2 min read
Unapproved AI Tools Add $670,000 to Breach Costs
New data shows one in five companies has already had a security breach caused by employees using AI tools without permission, and those breaches cost far more to fix, which is why company leaders can no longer treat AI oversight as an afterthought.
· 2 min read
Microsoft Copilot Adds No-Code App Builder and AI Agents
Microsoft rebuilt its Copilot app around three tools, a chat hub, a no-code app builder called Code, and an autonomous agent called Autopilot, while switching to usage-based pricing for the more advanced features.
· 2 min read
Companies Double AI Agent Budgets While Agents Stay Siloed
A new IDC survey of over 400 enterprise leaders finds companies plan to double their AI agent budgets over the next year even though only 29% of those agents can currently work together, exposing a governance and coordination gap that is already showing up in stalled pilots elsewhere.
· 2 min read
Media Outlets Turn Archives Into AI Answer Tools
News publishers are building narrow AI tools that let readers query years of archived reporting directly, and early results show focused tools placed at the point of need succeed while broad chatbots and unsupervised tools fail or get shut down, a pattern any expertise-based business can learn from.
· 2 min read
Google Links 14 Outside Apps Directly Into Gemini
Google now lets Gemini chat control 14 outside apps, from Airtable to Peloton, deepening a race with OpenAI's ChatGPT to become the one interface people use for everything.
· 2 min read
Google Adds Talking Video Avatars to Gemini Enterprise
Google launched a real-time video avatar for its Gemini Enterprise AI, letting a lip-synced digital character see, hear, and talk with customers or staff across 97 languages while quietly pulling data from other systems in the background.
· 2 min read
Google DeepMind Rushes Gemini 4, Drops AGI Talk
Google DeepMind's new head Koray Kavukcuoglu is racing to ship Gemini 4 months ahead of schedule and has quietly dropped talk of building artificial general intelligence, even as the lab loses top researchers to OpenAI and Anthropic and its promised Gemini 3.5 Pro model never shipped.
· 2 min read
Ando Raises $20 Million for AI Agent Messaging App
Ando, a new startup from repeat founder Sara Du, launched a Slack-style chat app that gives AI agents their own inboxes and lets them message coworkers directly, raising $20 million as Slack and Microsoft Teams race to add similar agent features of their own.
· 2 min read