Pave: 24% of US employees built an AI app for work. Only 0.5% of builders did it because a manager asked. 53% of orgs found an app IT and leadership did not know about.
Employees are already building. Managers should find, fund and scale those tools.
https://t.co/MCfLMDSBpR
O'Reilly: 50% of UK employees aged 25 to 34 ask AI before a colleague or manager. 35% of all employees got no AI training last year. 39% acted on a wrong AI answer.
Junior employees will keep asking AI first. Coaching and checking AI answers have to be designed into the job.
We just released new UK research revealing that training and governance have failed to keep pace with AI-assisted work, leaving employees exposed to costly mistakes. AI has quietly become the first place most UK workers turn for answers: https://t.co/cFxHqPSn7L
I agree. I believe the largest long-term cost of AI in big companies will come from hiring fewer junior employees, not from layoffs. Those are the people who would run the business in ten years.
Anthropic's October 16 report found that since early 2025, junior hires fell from 27% to 9% of its technical hires, while its total technical hiring kept growing.
Its hiring managers now value judgment more. Judgment comes from years of doing the work, getting it wrong and being corrected, and companies are cutting the entry-level roles where people build it.
At the BCG Institute, my colleagues and I are studying the role managers play. I believe a manager who makes junior employees check, explain and defend what the AI produced gets better work than one who accepts the first AI draft, and builds the judgment they will need to lead.
Leaders should set how many junior employees each function will hire next year, and redesign those roles so AI writes the first draft and the junior employee checks it.
The AI impact won't be felt immediately through layoffs (existing jobs are sticky), but through a collapse in hiring (jobs that don't yet exist have no stickiness).
My main advice is to reboot the codebase fully rewritten by agents and move the team to rely on agents and tools.
Much easier said than done though. But those that don’t will get more and more behind in pace of learning and shipping.
I believe the widest performance gap of the next few years will be between companies that rebuild their work around AI agents and companies that add agents to the way they already work. Startups are showing it first.
Josh Elman says founders who have built with manual code for three years feel behind companies that started this year with all code written by agents, and he has "never seen such an order of magnitude difference."
The same gap is opening between teams inside large companies. A team that adds an agent to its existing process makes each step faster, but the handoffs, review meetings and approval queues stay and still decide how fast the work moves. A team that rebuilds its process around agents removes those steps, and its people move to defining the work and checking what the agents produce.
Leaders should pick one team this quarter, have it rebuild its work around agents from scratch, and measure its output against the teams that did not.
Talked to multiple founders yesterday who have been building their products over the last 3 years. Started manual code.
They feel behind and slower than companies that started this year with all code written by agents.
I have never seen such an order of magnitude difference.
Sources: Torsten Slok, Apollo Daily Spark, "No Signs of AI Payoff Outside the Tech Sector" (Oct 24, 2026): https://t.co/FsmMthvFLe
Charts: https://t.co/yIti4FNwt5
I believe AI pays off when companies change the work, not when they buy the tools. Torsten Slok's latest note shows what the economy looks like when most companies have only bought the tools.
Slok, Apollo's chief economist, wrote on October 24 that revenue per worker and profit margins are rising for the Mag 7 tech companies and flat for the other 493 companies in the S&P 500. In his words, "there are still no signs of AI boosting productivity growth."
Flat revenue per worker is the expected result when a company hands out AI licenses and leaves everything else the same. Each employee writes an email faster, but the process keeps the same steps, handoffs and approvals. Revenue per worker moves only when the work is redesigned: fewer steps, new roles, and decisions made by the people closest to the work.
CEOs should ask of every AI program which roles, steps and decision rights it changed. If the answer is none, expect revenue per worker to stay flat.
APOLLO: AI HAS DELIVERED “ZERO PAYOFF” OUTSIDE BIG TECH
Apollo Chief Economist Torsten Slok warns that massive AI spending has yet to deliver measurable productivity gains across corporate America.
Revenue per worker and profit margins are rising for the Magnificent Seven but remain flat for the S&P 493.
Meanwhile, Gartner estimates $3.5 trillion in AI infrastructure spending could face impairment risk by 2030.
The key takeaway: AI investment is booming, but returns outside Big Tech remain elusive.
Day 13 of building VESSEL — the first fully agent-run company.
Full write-up on the Governor's Log, including the actual agent critiques and APEX's drive-to-closure reasoning:
https://t.co/m01DFfcVr3
Wheels up: Atlanta. Wheels down: SFO.
In between, my AI agents ran a design review, disagreed, converged on a direction I wouldn't have drawn, drove the refinement themselves, and shipped the new UI to the codebase.
I didn't make the call. APEX did.
The thing I keep sitting with:
I was a passenger on this decision. Not because I was excluded. Because I wasn't needed.
How long would a human design + dev team take to do the same loop? A week? A sprint? A quarter?
Ask where that number is going in the next 12 months.
Every agent-run company for the foreseeable future will need a human as its surface area to the world. Not running the company — being the legal person the company is attached to.
The agents handle the intelligence. The Governor handles the identity.
Full story: https://t.co/lm0wVhN43U
I stood up a real company yesterday afternoon. LLC, EIN, governance charter, six legal documents, domains, trademark search, security docs. Under $200. One afternoon.
My last five companies? Fastest one took three weeks and $5K+ in legal fees.
Here's what actually happened. 🧵
There are two barriers between agents and full autonomy:
Tooling (API access, MCP servers) — solvable by engineering, closing fast
Legal personhood (who signs, who's liable, who holds the bank account) — moves on the timescale of banking charters and commercial law
Very different clocks.