Musical accompaniment for this week's newsletter.
We are in that magical time of year when college students are wrapping up the best four, five, or six years of their lives and about to embark on the great unknown. But before you get about simmering in ennui whilst floating around the swimming pool drinking beer like Dustin Hoffman in The Graduate, you need to check the box on one very important milestone: sitting through the commencement address.
This commencement season, The New York Times cataloged what graduation speakers told the class of 2026 about AI and their futures. TL;DR it ain’t pretty. The range ran from “AI will create more jobs than it destroys” to “learn to work alongside it or good luck.” Which sounds reassuring until you realize that the spread of those takes — from optimists to fatalists, delivered by people who were paid to have something coherent to say — is itself the data point. Nobody actually knows. The speakers don't know. The economists don't know. The vendors selling the tools that are disrupting definitely don't know, but they have a slide for it.
Welcome to the real world, graduates. Nobody knows anything. This is not a recent development.
Those of us who graduated from college in 1991 understand this and have a few thoughts to share. The economy was a dumpster fire in the depths of a post-Gulf War recession. Nobody told us it would be fine. It wasn't fine for a while. But we eventually found our footing. You figure it out. But we were able to do that because the entry-level roles we needed still existed.
The class of 2026 doesn't have that backstop.
A recent survey of 1,500 senior talent leaders found more of them expect AI to increase entry-level hiring than decrease it. Great headline. But entry-level job postings are down 35% since 2023. You can't paper over that gap with optimism and a press release.
The most honest thing said out loud this week wasn't at a commencement. It was Standard Chartered CEO Bill Winters calling the 7,800 back-office workers losing their jobs to AI “lower-value human capital.” He apologized carefully and surgically in what was a textbook non-apology apology. One newspaper called it being “too honest about AI job losses,” which is the most clarifying phrase I've read in months. No one has a problem with the math. We just can't call it for what it is out loud at normal volume.
And yet, there is actual, organized worker resistance to mandatory AI tool adoption inside companies. People looked at the tools, looked at what the tools were for, and said no thank you. Which means the AI sales cycle has a demand-side problem that nobody in the vendor community wants to put in a deck: you can buy the license, but you can't buy the willingness.
I mean, even the Pope is concerned about the risks of AI. He dropped a 42,300-word document on the topic, which is either the most thorough AI governance framework of 2026 or the most extreme example of not reading the room on document length. Either way, when the Catholic Church is publishing longer AI risk frameworks than most tech companies, it's worth at least skimming the executive summary.
I don't have a tidy thesis for you. Neither do the commencement speakers. The 1991 parallel is real. It’s the uncertainty, the sense that the rules changed while you were studying for finals. But we didn't have AI eating the training wheels. That's genuinely new.
What I'd tell the class of 2026 is the same thing I'd have wanted to hear in 1991: the map may be wrong, but that's not the same as being lost. Figure out what humans do that's genuinely hard to automate — judgment, trust, context, the stuff that doesn't show up in cost curves — and get very, very good at it.
Everything else is noise. And there's a lot of it right now.
What else is going on this week?
The two guys who developed OpenClaw are now warning that vibe coding at scale produces vibe slop at scale, and the technical debt is going to be someone's problem. In case that slice of irony isn’t delicious enough for you, there’s more. Companies now have a new AI problem: too many agents.
Enterprises deployed agents quickly, without a governance framework, and now they're drowning in overlapping workflows and conflicting outputs, with no one really sure who owns what. Enterprises bought the dream. Now they're living the nightmare.
Meanwhile, the inference bill is on the way. Nobody budgeted for what it actually costs to run AI at scale across enterprise workflows. As agentic deployments multiply, so does the per-query tab — and most CFOs are about to get an education.
Speaking of CFOs: that Copilot seat, that Claude license, and the dozen point solutions nobody's deduplicating? They're about to land one line item at a time at renewal. That conversation is going to be fun. The math that justified the AI pivot — cheaper than humans — is getting messier, but companies are cutting people anyway. This is no longer purely an economics story; it's an ideology story.
And on the vendor side, 3Sixty Insights makes the case that the GTM motion itself is broken. Pricing models and packaging haven't caught up to how AI actually delivers value, which explains why pipelines look healthy and close rates don't.
The agentic era didn't wait for anyone to be ready. That's becoming increasingly obvious.
Anthropic just raised $30 billion. In one round. And is about to post its first profitable quarter.
Let that land for a second. A year ago, Anthropic was the scrappy underdog to OpenAI. Now it's the more interesting story as both companies sprint toward IPO. The narrative just flipped from “AI moonshot burning cash” to “AI moonshot with a business model.” That's a meaningful shift for the whole sector's credibility, not just one company's cap table.
Zoom out and the Crunchbase data tells the structural story. Capital is concentrating at the top of the AI stack at a proportion that makes every other category look like a rounding error. Six percent of U.S. tech companies raised 70% of the capital last year. That’s some gulp-inducing concentration risk. Even if you believe in the efficiency of markets, it’s hard to see how this ultimately ends well.
I’ve been covering this market long enough to have a pretty good feel for when someone is diagnosing a real problem versus dressing up a pitch.
Opal Wagnac is doing the former.
She’s been in HCM for nearly 20 years, came up as an engineer, and helped build what became UKG Pro. And she has a very specific, very clear-eyed take on what’s actually happening with AI in this market right now — and what buyers keep getting wrong when they walk into an evaluation.
The line that stuck with me: stop looking for your next HCM. Look for your last one. Give it a listen.
Coupa acquires Tonkean. Procurement software absorbing automation capability is a logical move and a signal that the workflow layer is becoming a feature of larger platforms rather than a standalone category. (Press Release)
Viktor raises $75 million Series A. Polish/German startup builds AI coworkers that live in Slack to allow companies to create reports, dashboards, apps, campaigns, code and recurring automations across their existing business systems. (FinSMEs)
Humanly raises $25 million Series B. The Seattle startup’s pitch is helping companies hire faster and stay fully staffed, which lands a bit differently in a week like this one. (Press Release)
Findd raises $21 million. The Provo, Utah-based provider of an AI-native workforce management platform focuses on frontline visibility and labor tracking. (FinSMEs)
RemotePass raises $17.4 million Series B. The London-based global payroll and contractor management for distributed teams is another example of the international employment infrastructure space’s continued momentum. (FinSMEs)
Blink raises $17 million. Frontline connectivity remains a durable problem: there are more frontline workers than desk workers globally, and they've historically been the last to get technology investment. (FinSMEs)
Zendesk commits $100 million to startups & VC program. This is strategic positioning as much as investment. Zendesk wants to be the enterprise CX platform that the next wave of startups builds on or integrates with. (VCWire)
SD Worx subsidiary Protime acquired Dutch workforce planning firm Checks, expanding its position in the Netherlands. (Press Release)
LinkedIn planning to lay off 5% of staff. The cuts to 875 roles across engineering, product, and marketing, despite recording its first $5B+ quarter and 12% revenue growth. It notably did not cite AI as the reason, making it one of the few 2026 tech layoffs where the AI narrative wasn't the cover story. New CEO Daniel Shapero framed it as reallocating toward "highest impact" priorities, which is the kind of sentence that means everything and nothing at the same time. (Reuters)
Workday pops as Q1 results, guidance top estimates. Workday beat Q1 estimates, and the stock responded. That’s a meaningful data point after a stretch of underperformance and persistent questions about whether the platform is losing ground to AI-native alternatives. One good quarter doesn't end the Workday Rorschach test, but it does give the bulls something to point to. (Seeking Alpha)
Intuit to lay off over 3,000 employees to refocus on AI. They will cut 3,200 people (17% of its global workforce) with CEO Sasan Goodarzi insisting "this was not about AI" while simultaneously describing a full resource redirect toward AI across TurboTax, QuickBooks, and Credit Karma. (TechCrunch)
Staffbase eliminates every fifth position. The German employee communications unicorn cut 20% of its workforce. European HR tech has been insulated from some of the U.S. layoff wave, but these cuts suggest the reckoning is arriving on that side of the Atlantic too. (MSN)
New data proves your 'AI employee' is destroying trust faster than it's cutting costs. Companies are deploying AI-facing customer interactions before the technology is ready to handle them well, and customers are noticing. The trust damage from a bad AI interaction compounds faster than the efficiency savings from a good one. (The State of Brand)
OpenAI's consulting deployment goes deep into HR-owned work. OpenAI launched a $4 billion enterprise consulting unit, and the HR function wasn't named as a stakeholder once. What's being redesigned includes hiring pipelines, performance management, and workforce planning: i.e., everything HR thought it owned. (HR Executive)
HiBob expands into Canada with Toronto office. HiBob is planting a flag in Canada, targeting the fast-growing mid-market business segment there. (Markets Insider)
TalentNeuron appoints David Wilkins as CEO. This is the kind of hire you make when you're ready to push harder on enterprise sales and market positioning. (Press Release)
Paychex launches AI platform for an agentic workforce. The incumbent advantage here is data depth and enterprise trust; whether Paychex can move fast enough to make WISE a genuine differentiator rather than a rebrand is the question worth watching. (Press Release)
Cornerstone launches its reinvention, helping to redefine corporate learning. Cornerstone is calling this a full platform reinvention — AI-powered skills intelligence, redesigned UX, new learning architecture — and Josh Bersin covered it favorably as a genuine strategic pivot rather than a press release rebrand. (Josh Bersin)
Deel launches stablecoin salary payouts. For companies with contractors in high-inflation markets or countries with limited banking infrastructure, this is genuinely useful; for everyone else, it's an interesting option nobody will use. (Deel)
Google unveils new Gemini AI agent for personal tasks. Every Work Tech vendor with an AI assistant play just got a better-resourced competitor. (Wall Street Journal)
LinkedIn is quietly killing video reach the same week that Google killed search as we know it. The algorithm pivot is real. LinkedIn video reach is declining while text posts continue to outperform, according to the data. Meanwhile, AI overviews are eating the traffic that content marketers depend on, and every Work Tech company with an SEO-dependent demand gen strategy needs a plan B, and probably needed it six months ago.
Cloudflare CEO: How I choose which employees to replace with AI. This decision framework for determining which roles get replaced by AI is either refreshingly honest or deeply unsettling, depending on your position in the org chart. (Wall Street Journal)
The stack around the agent. Craig Hepburn's piece on the infrastructure around agentic AI — what has to exist beyond the model itself for agents to actually work in enterprise environments — is one of the clearer frameworks floating around right now. (Substack)
Enterprise Agentic Computing — 12 Rules from Geoffrey Moore. The fact that Mr. Crossing the Chasm is writing these tells you something about where enterprise adoption is in the hype curve. (LinkedIn)
And, finally: Do yourself a favor and listen to some Miles Davis. The jazz legend would have been 100 this week.