By Damian Mathews & The Last Mile Team
Yesterday at a banking conference in Sydney, Sam Altman said something most people thought they’d never hear from him.
“I’m delighted to be wrong about this. I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened.”
That’s the CEO of OpenAI publicly walking back years of warnings about an AI-driven “jobs apocalypse.” Fortune reported the same week that Anthropic’s Dario Amodei is doing his own softening. Both companies are preparing for IPOs valued in the hundreds of billions. Enterprise customers who hear “AI will replace your workforce” are harder to close than ones who hear “AI will make your team more productive.” Read the walk-backs accordingly.
The aggregate unemployment data supports this softer story. Yale Budget Lab found no meaningful change in unemployment for high-AI-exposure jobs through March 2026. The headline number sits at 4.5%.
And our last two issues told a different story.
In Humans Around the Edge we covered Coinbase cutting 14% of its workforce. In Lower-Value Human Capital we covered Standard Chartered cutting 8,000 support roles by 2030. This week Meta started cutting 8,000 jobs. Intuit cut 17% days ago. Tech layoffs in 2026 have crossed 115,000.
Both can be true. Aggregate unemployment stays calm when cuts are targeted, displaced workers eventually find other roles, and new hiring has collapsed alongside. Goldman Sachs is calling it a “jobless recovery,” with the S&P 500 hitting record highs while monthly payroll gains have collapsed to 55,000. Earnings up. Net hiring down.
Economists are calling the pattern “no hire, no fire.”
Here’s the part most CX leaders are missing. The leading indicator for what AI is doing to your contact center probably isn’t in the layoff headlines, but in your own hiring data.
How many tier 1 reqs has your org opened in the last six months? How many got filled? How many got closed without being filled? How many quietly converted into an AI workflow somebody built in their spare time? If you don’t have those numbers in front of you, you’re operating on the macro narrative instead of your own micro reality.
One thing Altman said landed harder than anything else in his speech. He specifically called out customer support as a category that will largely disappear in its current form, even as he walked back the broader apocalypse framing.
So here’s the strategic question. If your tier 1 pipeline is quietly or gradually closing, where do your senior CX leaders in 2032 come from?
The companies hiring carefully today, training people with AI fluency from day one and building career ladders that don’t depend on tier 1 attrition, are designing their 2030 leadership team right now. The ones in default mode are saving budget today and accumulating a leadership gap they’ll feel later.
The aggregate apocalypse didn’t come. The structural reshape is happening anyway, and it’s most visible in the numbers you’re not putting on a dashboard.
What does your hiring tell you about your future?
— Damian
Here’s what went down this week.
Bleeding Edge
Early signals you should keep on your radar.
Anthropic is closing its $30 billion round at a $900 billion-plus valuation, the largest private AI raise on record. SpaceX’s IPO prospectus disclosed Anthropic is paying SpaceX roughly $1.25 billion per month for Colossus GPU compute through 2029, a $45 billion contract. If the round prices as reported, Anthropic edges past OpenAI on paper while quietly renting its frontier infrastructure from Elon Musk.
OpenAI filed a confidential S-1 with the SEC, targeting a Q4 2026 listing at $852 billion to $1 trillion. The debut could come as early as September, despite OpenAI losing $1.22 for every $1 of revenue in Q1 2026. Altman reportedly told staff that filing is different from being ready, which may be the most honest sentence in any S-1 this year.
Leading Edge
Proven moves you can copy today.
Microsoft published its monthly Copilot Studio update with computer-using agents now extended into multi-step workflows and a redesigned workflows canvas. Computer-using agents reached GA earlier in May with OpenAI CUA and Claude Sonnet 4.5 as the supported models, plus Azure Key Vault credential storage and Purview audit logging. For enterprises that have been blocking screen-acting agents on governance grounds, the excuse window is closing fast.
xAI opened Grok Build’s coding-agent beta to all SuperGrok and X Premium+ subscribers, with Custom Skills and a 25-language Speech-to-Text API landing the same week. Grok Build ships with a TUI, Agent Client Protocol support, parallel subagents, and worktrees, no longer locked to the $300 tier. xAI is shipping more incrementally than ever, and Grok could shake its novelty reputation with enterprise buyers within twelve months.
Off the Ledge
Hype and headaches we’re steering clear of.
Pope Leo XIV released his first encyclical, Magnifica Humanitas, focused on protecting human dignity in the age of AI. The 42,300-word letter was signed 135 years to the day after Rerum Novarum and presented personally alongside Anthropic co-founder Chris Olah. When the Vatican joins the AI safety conversation, governance teams that ignore Leo’s framing may find their boards quoting it back to them.
A GitHub tool called Heretic strips safety guardrails from Meta’s Llama 3.3 and Google’s Gemma 3 in under ten minutes, with four lines of code. The stripped models answer prompts on biological weapons, malware, and child-exploitation content that the originals refuse. Heretic has already produced 3,500 decensored models with 13 million downloads, and the open-weights safety story looks increasingly indefensible.
