Mania Math: The AI Arbitrage Play Has a Fatal Flaw

Musical accompaniment for this week's newsletter.

$11.38. That's the median amount companies spend per employee per month on AI. About the same as a Spotify Premium subscription. Or a cheap beer at an airport bar. Or a gallon of gas in the U.S. by the fall. 

The top 1% of companies? They're spending $7,500. Per employee. Per month.

That's not a gap. That's a different sport entirely.

The stats from this TechCrunch piece frame it as an AI ambition story — the bold vs. the cautious, the visionaries vs. the laggards. But I think they buried the real headline. Because when you lay that number next to the recent layoffs from Oracle and ServiceNow, a different story emerges.

Oracle just wrapped up 30,000 layoffs — while posting record cloud revenue. ServiceNow cut hundreds of employees and didn't even try to bury the reason. They called it "real AI efficiencies." In the press release. On purpose. Not a hot-mike moment. That's new. 

For years, the AI displacement conversation was a vibe. That's over.

What changed this week isn't the math. What changed is who's saying it out loud. ServiceNow's Bill McDermott once pledged no layoffs. Oracle's Larry Ellison has spent the last two years calling AI demand "insatiable" while simultaneously spending like a hyperscaler and cutting headcount like a company under margin pressure. Both things are true. Neither one gets to hide behind the other anymore.

New BambooHR data puts a finer point on it: Half of American workers fear AI will take their jobs. Now it seems that their bosses agree with them. After Oracle and ServiceNow file it in their earnings reports, every board in America has a template. Expect more corporate slash and burn.

But here’s the thing: what are these companies getting for $7,500 PEPM token spend? Are they getting disproportionate value per employee? The current pricing environment is essentially VC-subsidized, burning investor capital to acquire enterprise customers at below-cost pricing. Same playbook as Uber, AWS in its early days, every SaaS company that ever ran negative gross margins to hit growth targets. Once OpenAI and Anthropic go public, that won’t last.

That’s why it seems the assumptions underlying the AI vs. employee arbitrage math rest on a faulty premise: what a human costs today vs. what a token costs today. The real cost of AI deployment also includes infrastructure, integration, oversight, error correction, security, and compliance. So it seems there’s a lot of ready-fire-aim going on with employers right now. The guns-blazing AI leaders aren’t questioning their assumptions that seem more grounded in mania than reality. 

Tell me how this ends. Probably ends with a messy, multi-year recalibration that nobody calls a failure. The efficiency gains get quietly revised downward. The headcount math gets quietly revised upward. It gets reframed as “AI-human collaboration was always the plan.”

That’s the beauty of the tech industry. No one ever fails. You just recast the narrative and tell everyone that the missteps were the mission all along.

What else is going on this week?

Are Toxic Workplaces Winning? Looks That Way.

In an environment where AI layoffs are surging, you might think that toxic leadership is winning. And, unfortunately, new research from Businessolver in its 2026 State of Workplace Empathy Report seems to validate this view The report, released today, is a sobering read.

They surveyed 300 CXOs (including CEOs, CHROs, CTOs, and CFOs) and 1,000 employees across 6 industries and found:

• 70% of CXOs in toxic cultures report significant financial growth, compared to 36% in non-toxic organizations — nearly 2 times the rate of growth in past 12 months
• 40% of employees say their workplace is toxic, rising 18 percentage points year over year (YOY), alongside 33% of CEOs, up 25 points YOY
• And yet, 98% of CXOs say they're empathetic leaders, rising to 100% among CXOs in toxic organizations. 

In the AI era, it has felt like a crisis has been brewing in employee experience that will ultimately impact employee retention and business performance. This report indicates that there’s some there there.

Oracle Wins $396 Million Federal HR Systems Overhaul

The Office of Personnel Management is sticking with the incumbent as the agency moves forward with a plan to modernize human resource systems across the government. OPM has picked Oracle for the 10-year, $395.8 million Federal HR 2.0 contract that will cover more than two million federal employees. Oracle faced challengers such as Workday, IBM, SAP and Economic Systems Inc. Oracle may have the headline number, but expect this initiative to throw off a lot more integration and update opportunities for other HR tech vendors.

Webinar: Why Your Partner Marketing Content Isn't Working — And How to Fix It Now

Partner marketing is the only GTM motion where you're expected to serve three customers at once — your company, your partner, and the end buyer — with a content budget built for one. That's the gap. And most companies don't see it until partners stop engaging, co-sell stalls, or a competitor shows up better-prepared in the channel. Join us at 1 p.m. Central on Thursday, June 25, as we break down what going wrong and how to make it go right. Register now!

Funding, Acquisitions, and IPOs

• SpaceX agrees to buy AI coding agent Cursor for $60 billion. A little fresh paper can do what xAI couldn't: become relevant in the one AI category that actually has enterprise revenue. (Wall Street Journal)
• Nuvei to acquire Payoneer for $2.75 billion, creating a leading global platform for local and cross-border commerce. Cross-border payments consolidation continues. (Press Release)
• DeepSeek becomes China's most valuable AI startup after $7.4 billion fundraise. The model that briefly crashed Nvidia's stock earlier this year just raised enough money to make the "Chinese AI is underfunded" narrative obsolete. (Wall Street Journal)
• Genspark.ai extends Series B to $485 million at $2.6 billion valuation. Another week, another AI agent platform round with a "b" in it. The valuation says investors are still betting that someone other than OpenAI and Anthropic wins a vertical. (FinSMEs)
• Factorial raises $150 million Series D at $2.5 billion valuation to expand across Europe. Barcelona's Factorial is the rare HR platform that actually rebuilt its product around AI rather than bolting a chatbot onto a SaaS dashboard. (FinSMEs)
• Ent raises $100 million in seed funding. This SF startup is already deployed with Global 2000 customers across hospitality, financial services, and defense to detect insider risk, govern AI usage, prevent data loss, stop last-mile threats, and investigate incidents. (FinSMEs)
• Voice AI startup Bland raises $50 million after being rejected by 180 investors. Enterprise phone automation has gone from "IVR hell" to VC darling in about 18 months. (Fortune)
• As AI agents become employees, NewCore emerges with $66 million to give them identities. NewCore is betting that identity infrastructure for non-human workers is the unsexy-but-necessary plumbing of the agentic era. (TechCrunch)
• Arcade.dev raises $60 million to secure AI agents. Security for AI agents — looks like this is a trend. (Wall Street Journal)
• Apis Partners invests $50 million to accelerate BIPO's global expansion. BIPO is building multi-country HR and payroll infrastructure for Asia-Pacific and beyond, signaling continued appetite for non-Western global workforce platforms. (BIPO)
• Convey raises $38 million in Series A funding. This SF enterprise AI platform enables non-technical operators to build and manage digital teammates that execute business operations autonomously. (FinSMEs)
• Jedify raises $24 million to help companies arm AI agents with context on their business. AI agents that don't know your company's policies, history, and quirks are just expensive autocomplete. Jedify's bet is that the bottleneck isn't the model; it's the organizational knowledge layer sitting around it. (TechCrunch)
• Orbio raises $21 million in Series A funding. The Madrid, Spain HR/workforce analytics startup is riding the pre-agentic infrastructure wave where data and insight tools are racing to become action tools before the agents make them redundant. (FinSMEs)
• Asana acquires StackAI. Buying an AI workflow automation startup signals that work management platforms know their core product is under threat. If AI agents can orchestrate tasks end-to-end, the manual project board starts to look like overhead. (FinSMEs)
• Payslip closes new funding. The Irish payroll orchestration player sits in the multi-country payroll middleware layer, which is being squeezed from above by Deel and Rippling and from below by newer agentic entrants. (FinSMEs)
• Cegid acquires Shine to build Europe's first AI-powered financial and HR hub for SMBs. Cegid is doing the European SMB version of what Rippling did in the US: collapsing the wall between HR, finance, and IT into one platform. (Press Release)
• PartsSource acquires SkillNet to build unified clinical asset and workforce intelligence platform. This acquisition is about collapsing the distance between "do we have the equipment" and "do we have the trained staff to use it," which is a very specific but very real operational problem in clinical settings. (Press Release)
• Warren raises €10 million in seed funding. The Ghent, Belgium-based startup is developing a digital workplace pension and financial coaching platform. (FinSMEs)
• HeyMilo AI raises $6 million. AI interviewing is a crowded category and one to watch as AI screening becomes table stakes and differentiation gets harder. (FinSMEs)
• Applauz closes $2.75 million CAD growth financing to accelerate its AI-native employee recognition platform. The "AI-native" positioning is doing some heavy lifting here. (Press Release)
• HR SaaS 2026 Q1 public markets update. Kyle Getsiv's quarterly read on public HR SaaS is one of the better scorecards for understanding which platforms are holding valuation multiples and which are getting repriced as AI challengers close the gap. (Substack)

Industry Notes

• Anthropic taps TCS to scale its enterprise AI deployments. When the lab that built Claude starts routing enterprise sales through one of the world's largest IT services firms, the "AI is a product" era is officially giving way to the "AI is a managed service" era. This is how frontier models get embedded into the Fortune 500 at scale. (TechCrunch)
• Databricks releases general AI agents for businesses. Databricks stepping into the agentic space with a general-purpose product is the kind of move that should make ServiceNow, Salesforce, and every other "we're building agents" vendor nervous. (Wall Street Journal)
• UKG adds agentic orchestration to connect workforce intelligence with frontline execution. UKG is one of the few legacy WFM platforms with enough data depth to make agentic orchestration actually useful. (Press Release)
• ZipRecruiter expands AI-powered job search with Claude integration. Here’s another signal that the job board layer is racing to add intelligence before it gets disintermediated by AI agents doing the job search autonomously. The question is whether a smarter search fixes a structural problem. (Press Release)
• Indeed's AI-powered sourcing assistant helps employers hire over 30% faster. Indeed has had a rough narrative cycle lately between the free job posting visibility cuts and the competitive pressure from LinkedIn. It’s doubtful that a 30% faster-hiring claim is the kind of number that would reverse the story. (Press Release)
• Greenhouse launches new AI capabilities. Greenhouse is one of the ATS incumbents with the most to lose — and gain — from AI-native recruiting competitors. The new capabilities announcement is light on specifics but heavy on timing, arriving as Ashby and newer entrants apply pressure from below. (Press Release)
• Employ adds VONQ AI screening to JazzHR and Lever. Dropping AI screening into both of these brands is the kind of feature-parity move that keeps buyers from defecting rather than winning new ones. (ERE)
• PrismHR announcements. They unveil AI intelligence layer Prism Intelligence (Pi) and embedded assistant Prisma, as well as launching PrismHR Global to simplify international expansion for service providers and SMBs. 
• HR Tech in the age of AI agents: what changes, what endures. The argument that the problems HR tech solves aren't going away, but the mechanisms are being rebuilt from scratch, is one of the more useful frames for where this market is actually headed. (Substack)
• This week in CyberBreaches: Oracle warns of a security bug that hackers abused to breach 100+ companies. ServiceNow tells customers a bug left some of their data exposed to the internet. 
• Is AI ending enterprise partner careers? Salesforce may be just the start. The system integrator model is directly in the crosshairs of agentic AI. (The State of AI)
• Payscale flight risk report shows new hires out-earning tenured workers in the AI job revolution. Compression is the word nobody wants to say out loud. That math eventually shows up in retention data, and apparently it already is. (Press Release)
• U.S. small businesses report lowest hiring plans in 6 years. SMB hiring sentiment is a leading indicator for the broader labor market, and if the smallest employers are pulling back, the “strong labor market” narrative has some structural holes. (LinkedIn News)
• Lance Haun on the mid-career slump. The structural stall that hits experienced professionals in their 40s is a human story that intersects with every AI displacement narrative, because mid-career workers are both the most affected and the least served by current reskilling infrastructure. (LinkedIn)
• Coders were tech's MVPs. Now they're fighting to stay in the game. The answer isn't "learn to prompt" — it's more complicated, and this piece doesn't flinch from that. (Wall Street Journal)
• Does remote work make employees happier? Here's what the evidence says. Happiness at work is multivariable, and "where you sit" is one of the less interesting variables once you control for everything else. (Wall Street Journal)
• Work from home is here to stay — even if some CEOs don't love it. The question is less "will it stay" and more "who gets to decide the terms." (Wall Street Journal)
• Humans are more important than ever in the AI era, says Satya Nadella. Every tech CEO eventually gives this speech, usually right before a layoff announcement. To be fair to Nadella, he's been more consistent on this than most. (The State of AI)
• Everyone's AI is a 'coworker' now. That's a branding disaster. This piece makes the case that the coworker framing is obscuring real product differentiation — and it's right. (The State of Brand)
• Lattice promotes Sophie Hurcombe to chief people and operations officer to lead people + AI transformation. (Press Release)

Worth Reading

• Meghan Gendelman, Canva's B2B CMO, is right that brand is now infrastructure. The harder build comes next. The argument that brand isn't a campaign but a foundational layer for how a business operates is one of the better reframes for what modern B2B marketing is actually doing. (The State of Brand)
• Systems of record won the SaaS era — clearinghouses will win the agents era. One of the better strategic frames for what happens to enterprise software when agents can pull data from anywhere: the value shifts from who holds the data to who connects it. (Clouded Judgement)
• The LinkedIn makeovers that actually get people hired. The findings on which profile changes actually move the needle (headline specificity, skills alignment) are worth a skim if you have clients or candidates in a job search. (Wall Street Journal)
• Eight predictions for the future of higher education. Credentials are getting repriced and the institutions built around them are scrambling. (The New Yorker)