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 9 of 76
Morgan Stanley's AI Agents Rest on Data Built Around Clients
Morgan Stanley's wealth division now has an AI agent that handles service requests like changing a beneficiary, and it works because years ago the division organized its records around the client, not the account.
· 2 min read
HR Software's Built-In Limits Become Company Policy by Default
When HR software allows only one critical role in a succession plan, that limit becomes company policy, and AI now makes it cheap for a team to write its own rule on top.
· 2 min read
A Paywall That Hands Out Its Own Key Fails Against AI Agents
Claude got Georgia's paid vote counts by asking the state's dashboard for the free token it gives every visitor, which shows how little a gate protects when the page hands out its own key.
· 2 min read
Radio's AI Co-Host Works Because It Argues With a Human
A Los Angeles station's AI co-host is winning listeners where an earlier AI clone of a human DJ did not, because it has opinions of its own and a human host who steers it.
· 3 min read
Robots Taught by Workers Could Reach Toyota's Small Suppliers
Toyota's humanoid robots learn by watching skilled workers, which could make automation pay for small suppliers and turn the best person on a line into the robot's teacher.
· 2 min read
Supply Chains Lose Weeks to Internal Arguments, Not Bad Forecasts
AI agents that make a stock decision across every department at once show that supply chains lose most of their time in the argument between teams, which makes writing down the trade-off rules the new work.
· 2 min read
When Every Rival Gets AI, the Savings Go to Customers
Flat margins and sales per worker outside Big Tech do not prove AI fails: in 1990s retail, when Walmart's rivals copied its systems, the gain reached shoppers as lower prices.
· 2 min read
A Refund Can't Cover What a Flawed AI-Aided Report Costs
A Deloitte study for Western Australia has a cost total wrong by A$260m on a A$166,000 fee, so a refund cannot cover the risk and the buyer has to add up the totals before paying.
· 2 min read
AI Buyers Who Can Switch Providers Are Getting Big Discounts
OpenAI is cutting prices for individual big customers to pull them from Anthropic, which shows that what a company pays for AI now depends on how easily it can leave.
· 2 min read
Pennsylvania's AI Ad Label Won't Protect a Misleading Ad
Pennsylvania's new law asks whether an AI-made ad misleads customers about what they are buying, so a label cannot fix a picture that oversells the product.
· 3 min read
AI Is Shrinking the Time Firms Have to Install Updates
AI found a 14-year-old flaw in the software behind most websites, and attacks were seen before the fix was out, so how fast a business installs updates now matters.
· 2 min read
Company AI Only Knows What Staff Wrote Down
An assistant that answers from a company's chats and emails mirrors the people who write the most, so you should check who is missing from the record before acting on its answers.
· 3 min read
Canva Trained Its Image AI on Menus and Posters, Not Photos
Canva's new image model learned from finished posters, menus and social posts instead of photographs, which points to where AI images for everyday work are heading.
· 2 min read
AI Turned One Stolen Vendor Key Into Attacks on 2,250 Firms
Hackers used AI agents to try a stolen backup-vendor key on more than 2,250 companies, which shows that every outside app you approve in Microsoft 365 is a key someone else can lose.
· 3 min read
Big AI Gains Come When Companies Redraw the Work
Only 9% of big-company CEOs call AI agents transformational, and early electric factories suggest why: the gain came only after the factory layout was redrawn around the new power.
· 2 min read
Banks Will Give AI Agents Their Own Keys, Not Customer Passwords
Banks blocking AI agents that log in with a customer's password is a replay of the 2015 fight over money apps, which ended with banks issuing limited access instead of passwords.
· 2 min read
Uber Hands AI Agents to 30,000 People Who Know the Work
Uber's own two-week experiment showed that the hard part of automating a job is knowing the job, and giving all 30,000 employees an agent puts the people who know it in charge.
· 2 min read
When AI Speeds Up the Work, the Wait Moves to Decisions
At software firms using AI coding agents, engineers wrote about 30% more code but finished work did not rise, because the wait moved to the people who have to approve it.
· 3 min read
AI Conversation Logs Are Becoming a Company's Password File
LiteLLM's poisoned update most likely began with login tokens read from its own AI logs, which shows that any record of staff conversations with AI now holds the keys to the company.
· 2 min read
The Password Was Never What AI Agents Needed
1Password now signs AI agents into your accounts without showing them the password, much as Apple Pay stopped shops seeing card numbers, but a signed-in agent can still do whatever the account allows.
· 2 min read