Welcome to Humans in the Loop! 🙏
Thank you for being an early subscriber. You clearly have great taste!

I built this for smart, busy executives outside the Silicon Valley bubble who keep thinking: 👉 "I know AI matters… but can someone just explain it in plain English?"

Got thoughts? Good, bad, brutal — send them my way: ha.nguyen@ember.new

OpenAI stopped waiting for you to figure it out yourself. This week the company officially launched the OpenAI Deployment Company, a new venture designed to embed specialist engineers — called Forward Deployed Engineers — directly into businesses to rebuild workflows around AI. To staff it from day one, OpenAI acquired applied AI firm Tomoro, bringing roughly 150 engineers who have already done this kind of work at companies like Tesco and Virgin Atlantic. The venture is backed by a who's-who of money: TPG, Goldman Sachs, Bain Capital, McKinsey, Capgemini, and 15 others, at a reported $10 billion valuation.

Here is what this means for a CEO who did not go to engineering school: OpenAI is no longer just a software vendor. It is now in the professional services business, and it is coming after McKinsey and Accenture's lunch at the same time. If your board is asking why your AI pilots are not turning into real savings, this is the answer — deployment is the hard part, and OpenAI just decided to own that layer. The question is whether you want OpenAI's engineers redesigning your core processes or your own team doing it first.

• Enterprise now makes up more than 40% of OpenAI's revenue and is on track to match consumer revenue by end of 2026.
• OpenAI's new Frontier platform is already helping Oracle, State Farm, and Uber manage AI agents company-wide.
• OpenAI's own data shows 'frontier firms' use 3.5x as much AI per worker as typical firms — up from 2x just a year ago, and the gap is widening fast.

Looking ahead, both the OpenAI and Anthropic deployment ventures are racing to sign their first big non-tech reference customers — your industry may be next on their pitch list.

--ML

Anthropic — the maker of the Claude AI — is reportedly in early talks to raise $30 to $50 billion at a valuation north of $900 billion, which would put it ahead of OpenAI's last private price tag of $852 billion. That is a stunning jump from its $380 billion valuation just three months ago. The fuel: Anthropic's annualized revenue has rocketed from roughly $9 billion at the end of 2025 to over $44 billion by May 2026, with enterprise clients accounting for more than 80% of that haul. More than 1,000 large customers each spend over $1 million per year on Claude.

For a non-tech CEO, here is what to take from this: the AI race is no longer just about which model is smartest. It is about which company can lock in the most enterprise contracts at scale. Anthropic already has eight of the Fortune 10 as stable clients, Claude runs on all three major cloud platforms (AWS, Google, and Microsoft Azure), and a potential IPO is reportedly targeted for as early as October. If you have not evaluated Claude for your workflows, your competitors probably have.

• Google pledged up to $40 billion in Anthropic in April 2026; Amazon has committed up to $25 billion.
• Anthropic's gross margins improved from 38% a year ago to over 70%, quieting fears that AI companies can't turn a profit.
• Anthropic also launched a joint enterprise deployment venture backed by Blackstone, Goldman Sachs, and Hellman & Friedman, mirroring OpenAI's move almost simultaneously.

Looking ahead, an Anthropic IPO in Q4 2026 would be the most closely watched tech listing in years and could reshape how investors value every AI company in your portfolio.

--ML

At its annual Sapphire conference in Orlando this week, SAP unveiled what it calls the 'Autonomous Enterprise' — a vision where AI agents built into SAP software can close the books, reforecast a quarter, or fill a headcount vacancy without a human clicking through screens. The centerpiece is a new unified SAP Business AI Platform that merges three separate products into one, powered by a 'Knowledge Graph' that gives AI agents a live map of your entire business. Joule Studio, SAP's tool for building these agents, is available now.

If you run SAP for finance, HR, or supply chain, this matters immediately. Your vendor is betting that AI agents will replace dozens of routine steps your team does today. The good news: SAP says users can just describe what they want in plain language and Joule figures out the rest. The awkward news: SAP quietly updated its API policy to block outside AI agents — like Microsoft Copilot, which 77% of SAP's own enterprise customers already use — from talking directly to SAP data, which is raising eyebrows industry-wide.

• SAP deepened partnerships with Anthropic, AWS, Google Cloud, Microsoft, NVIDIA, and Palantir — basically every major player at once.
• SAP also agreed to acquire data lakehouse company Dremio, and committed over €1 billion to build a frontier AI research lab from its Prior Labs acquisition.

Looking ahead, if you are an SAP shop, ask your account team right now how the new API policy affects the Microsoft Copilot or other AI tools your teams already rely on.

--ML

[ Reported without editorial commentary ]

Google's threat intelligence team disclosed this week that it disrupted a criminal group using an AI large language model to discover a previously unknown vulnerability — a 'zero-day' — in another company's systems. Google said it notified the victim and law enforcement and stopped the attack before damage was done. It did not name the target or the AI model used, but confirmed it was likely not Google's Gemini or Anthropic's Claude. The disclosure confirms what security researchers have warned for months: AI-assisted attacks are no longer theoretical.

For a non-tech CEO, the board-ready version is this: the window between when a software flaw is discovered and when attackers exploit it has shrunk to less than 24 hours in 28% of cases, according to Mandiant's latest data. Meanwhile, Gartner estimates that 40% of enterprise applications will use AI agents by end of year, yet only 6% of organizations have an advanced AI security strategy in place. That math is not great. Ask your IT team this week: do we have a plan for securing AI agents that have access to our core systems?

• IBM's 2026 X-Force report found over 300,000 stolen ChatGPT credentials in infostealer malware — hackers can mine entire chat histories for sensitive business data.
• AI-generated phishing emails now account for 40% of business email compromise attempts, making your finance and HR teams the highest-value targets.

Looking ahead, cybersecurity insurance underwriters are already revising AI-risk exclusions, so check your policy before your renewal — not after an incident.

--ML

On May 7, EU lawmakers reached a political deal to amend the AI Act, pushing some compliance deadlines for high-risk AI systems back by up to 16 months and reducing fines for smaller companies. The package, called the 'AI Act Omnibus,' also extended protections to mid-size companies and simplified documentation requirements. Good news, right? Sort of. The transparency rules for chatbots still kick in this August, and the deadline for labeling AI-generated content was actually tightened — companies now have until December 2, 2026, not six months later.

In the U.S., the picture is messier. Federal AI oversight is minimal under the current administration, but seven states including California, Colorado, and New York have passed their own AI laws — and California's automated decision-making rules require businesses to give consumers pre-use notices and opt-out mechanisms for AI used in hiring, lending, or healthcare. The SEC has also flagged AI governance as a 2026 examination priority for financial firms. The short version for your general counsel: you are now a regulated company for AI purposes, even if you do not feel like one.

• Illinois, Colorado, and New York all have AI-in-hiring laws that require employers to notify job candidates when AI is analyzing their video interviews.
• California and Colorado laws focus on 'consequential decisions' — any AI used in hiring, lending, housing, healthcare, or legal services now comes with notice, opt-out, and bias-testing requirements.

Looking ahead, 40 state legislatures are actively moving AI bills, so your legal team needs a live tracker — this is no longer a 'watch and wait' situation.

--ML

A new Gallup survey of more than 23,000 U.S. employees found that 65% of workers at AI-adopting companies say the technology has improved their productivity, but 27% also say their workplace has changed in 'disruptive ways' — more than double the rate at non-AI companies. The headline job numbers look stable, but BCG research warns that stability 'may mask critical issues' as companies rapidly redesign roles faster than workers can be retrained. Entry-level white-collar workers in their 20s and 30s are most exposed, Goldman Sachs economists say.

For CEOs, the workforce math is getting clearer. PwC data shows workers with advanced AI skills now earn 56% more than peers in the same role without those skills. That premium is a signal: your competitors are paying up for AI-fluent talent, and if you are not investing in upskilling your existing team, you are building a two-speed workforce. The companies winning right now are not replacing people wholesale — they are redesigning jobs so that AI handles the repetitive middle and humans handle judgment, relationships, and accountability.

• BCG says the biggest challenge is not how many jobs AI affects but how quickly companies can reskill and redeploy workers into redesigned roles.
• Cisco used OpenAI's Codex to cut software build times by 20% and save 1,500-plus engineering hours per month — by treating it 'as part of the team,' not a standalone tool.

Looking ahead, if you do not have a workforce AI strategy on the agenda for your next leadership offsite, your 2027 budget process will force it on you anyway.

--ML

Nine in ten retailers say they will increase AI budgets in 2026, according to NVIDIA's annual retail survey, with the focus shifting from chatbots to physical AI in warehouses and supply chains. Retail and e-commerce now lead all sectors with 83% AI adoption in supply chain operations. Meanwhile, hospitals are catching up fast: healthcare and pharma saw the largest year-over-year jump in AI supply chain adoption, leaping from 41% to 65% in a single year. Mayo Clinic alone is mapping more than $1 billion in AI investments across 200-plus projects.

Looking ahead, if your sector is not retail or healthcare, study what they are doing — the playbook is proven and the tools are now affordable for mid-market companies too.

--ML