Brainverse

Gartner asked 350 executives at $1B+ companies how they paid for AI. About 80% cut staff. The cuts had zero correlation with returns. The firms that profited kept their people and ran the work differently. Full breakdown: youtu.be/AB1BEuOdILQ

Roughly 4 in 5 companies cut staff to pay for AI. Gartner just found the cut was the wrong bet. The savings was real. The return on the AI was not there behind it. The firms that got returns made the opposite call. Almost nobody made it.

The White House just put AI agents in a federal enforcement order. On June 2 it told the DOJ to go after agents that reach data without permission, using fraud laws that have been on the books for years. There's no new crime here, the order just says those laws cover agents now. A Gravitee survey of 900-plus companies found 88% had an AI agent security scare this year (State of AI Agent Security 2026, via VentureBeat). Usually it's a bot running on way more access than anyone meant to give it. The way through is agents that run as a team, each one scoped to what its job actually needs, with a person who owns the result.

Source: Careerminds 2026 survey, 600 HR pros. careerminds.com/blog/cost-of-a…

Two out of three companies that cut staff for AI are already hiring them back. About 1 in 3 spent more on the rehire than the AI ever saved. Over half said they were babysitting the tool. They didn't replace a worker with an AI team. They replaced one with one tool nobody checked.

20% of companies just took 74% of all the value AI created this year. The ones winning didn't point AI at cutting costs. They pointed it at speed. And at AI speed a lead like that compounds, so a slower competitor doesn't catch up, they fall further behind every cycle. Jeff breaks it down, every number sourced. Watch: youtu.be/IoHGtz0r5Ek #AI #agenticAI

20% of companies took 74% of all the value AI created this year. PwC counted it. The winners didn't aim AI at cutting costs. They aimed it at growth, did the same work in a fraction of the time, then reloaded the freed-up hours into 10x more output. IBM: fast movers pull further ahead every cycle. Menlo: AI-natives already hold 63% of the app-layer market, up from 36% a year ago. A rival stuck at human speed isn't behind. They're uncatchable. The prize was never saving labor. It's velocity. We run 100+ agents in production. Full breakdown in the replies. #AI #agenticAI

Full breakdown, every source linked: brainverse.ai/?utm_source=x&…

The 19-of-20 came from one of the companies building this, quoted by CBC. And Gartner found ~80% of firms that cut staff for AI got no return on it. They didn't replace a worker. They removed one and hoped.

19 of 20 is a solo-agent number. One model, a real task, nobody checking the work. An AI company just admitted its own agents miss the bar that often. Companies are still firing people to run them like that. Put a person over the agent and the number changes.

Anthropic just raised $65B at a $965B valuation. Most valuable AI startup on earth, past OpenAI. The model the smartest money on the planet paid $65B for? You can rent it for about $20 a seat. The model was never the moat. The gap is the team you build on top of it. We run 100+ AI agents in production as coordinated teams. Follow for daily AI news.

Only 13% of people at big companies can actually work well with AI agents. Out of 14 skills they measured, that one came in dead last. Most companies see that and try to train the whole staff into operators. We think that's backwards. What's the one task where your team still babysits the AI the whole way through?

65% of companies running AI agents have already had a security incident caused by one of their own agents. That number's from a Cloud Security Alliance report this year. Most of them can't actually shut a bad one off, either. Only about 1 in 5 has a clean way to pull an agent out when it starts misbehaving. We wrote up why an agent you can't turn off is a liability instead of an asset, and what it takes to keep one under control. #AIagents #AIsecurity

In a clinic, the AI mistake that actually hurts someone is almost never inside one document. One tool reads the chart. A different tool reads the lab order. A third drafts the prior-auth. Each one is right on its own file. The harm sits in the gap where they disagree: a dose that fits the chart but not the diagnosis, an order that contradicts an allergy noted three pages back. A smarter model does not fix this. A model that reads one file at a time, however good, cannot catch a contradiction it never holds both sides of. What catches it is agents that hand the work to each other and keep the whole case in shared memory, not one chatbot per document. That is how we run our own shop, 100+ agents that cross-check each other instead of working solo.

Full video: youtu.be/EOsKohC-aSU?si…

Half a billion dollars on AI in a single month. Reportedly. A company turned Claude on for everyone with no usage caps. People started burning tokens to top an internal leaderboard. One exec said his staff used it to check the weather. The bill was never the real story. Nobody owned the AI or decided what it was for. That is what scattered tools with no structure actually cost you. The opposite is an AI organization, agents with real jobs and an owner for the work. We run ours that way every day. Full breakdown on YouTube, link below.

99% of CEOs say they're planning AI layoffs. Most are solving the wrong problem. The companies actually winning with AI don't have fewer people. They have people who each run a team now. We run 100+ agents in production. Not one replaced a person.

43% of enterprise AI projects are about to fail. Full breakdown: brainverse.ai/insights/enter…

Anthropic just raised $1.5B from Goldman, Blackstone and Apollo to build what's basically the McKinsey of AI: embed engineers inside companies and rebuild their workflows around AI agents. Full article: brainverse.ai/insights/anthr…