Musical accompaniment for this week's newsletter.
Last Monday morning, about 60,000 Microsoft employees opened an email with a link to their exit package. Two days later, Microsoft told Wall Street it had spent nearly $40 billion on capex in a single quarter. Stop me if you’ve heard this one before: The money is moving from people to processors.
The offer is going out to roughly a quarter of Microsoft’s 223,000-person workforce. The package is similar to Microsoft’s standard severance, not some sweetened golden parachute.
Here’s the thing: Microsoft expects 3,000 or fewer to accept. When it ran a voluntary retirement program this spring, about 30% of the roughly 8,750 eligible employees took the deal, and the company said the results reduced the size of the layoffs it announced in July. Voluntary first. Involuntary after. That’s the sequence.
Meanwhile, the server side of the ledger is having a banner year. Azure grew 48% as capex jumped to $39.4 billion. The GPUs are coming in the front door. People are being ushered out the back.
Josh Bersin makes the case that this isn’t a layoff but a new model of management. Giving people a choice and real support beats a 6 a.m. calendar invite from HR. He’s also said that companies taking the heavy-handed approach to AI-driven layoffs risk missing out on AI productivity gains. Both things can be true. A kinder process is still a process with a target.
All this happened the same week that Gallup reported U.S. employee engagement falling below 30% for the first time in 15 years. Imagine being one of the Microsoft employees who sticks around. How focused do you think they’re going to be for the next few months? That uncertainty is an engagement tax.
And it may not even buy what leadership thinks. As Vishal Sood argues, AI productivity has a multiplayer problem. Individual AI gains don’t automatically add up to team gains, especially when the team spends Q4 wondering if one or all of them are next on the chopping block.
Nonetheless, voluntary exit programs are about to become the playbook. They’re cheaper than RIFs, kinder on headlines, and easier to explain to a board. If you sell workforce planning, engagement, or internal mobility tools, your buyers are about to need help with the 95% who stay. As my colleague Tony Spangler has been saying a lot lately, it’s feeling a lot like 2009 again. Employee engagement is in the tank and a lot of employers haven’t quite gotten the memo.
Should I stay or should I go? At Microsoft, 60,000 people are asking. Most of them are staying. The question their bosses should be asking is what kind of employees they’ll be by spring.
What else is going on this week?
Enterprise AI vs. SaaS Finally Has an A/B Test
Elsewhere on the earnings front, there were other incumbents shuddering. ServiceNow built an empire by convincing enterprises to rip out their clunky old help desks. This week, Wall Street decided ServiceNow might be the clunky old help desk. Shares fell about 23% on a two-cent EPS miss and softer backlog guidance. That made it the company’s worst day ever. Two cents. A quarter of its market cap.
Here’s what the selloff got wrong, and what it accidentally got right. The “wrong” part is easy. On the earnings call, the CFO said early renewals pulled into Q3 drove about half the beat, at Q4’s expense. That’s a calendar problem, not a collapse. Subscription revenue still grew 21%.
The “right” part showed up with Serval raising $150 million at a $3 billion valuation, triple the $1 billion it fetched in December. Serval builds AI agents that do IT and HR service work, and tech pundits are already framing it as a ServiceNow replacement. Serval’s CEO said Lucid Motors migrated from ServiceNow to Serval in 45 days.
Google also saw its stock price tank after its earnings report (even though cloud revenue was one of the bright spots. Query growth hit its fastest pace in five years, yet search revenue growth slowed to 12% from 17% the prior quarter.
Then SAP quietly addressed the elephant in the server closet: pricing. It’s swapping pricey AI models for cheaper ones, and it runs more than 70 models and expects over €2 billion in efficiency gains over two years. Because SAP charges for completed tasks rather than computing resources, every cheaper model widens SAP’s spread. Your bill doesn’t shrink. Their margin grows. SAP told Reuters the goal is “healthy margins as scale grows.”
Enterprise software value has moved to the task, and whoever controls task pricing controls the margin. That’s the opportunity and the risk.
Not every incumbent is getting eaten. ADP beat and raised, growing revenue 6% to $5.5 billion and raising full-year guidance. ADP gets paid for getting paychecks right, not for logins. Compliance, trust and distribution don’t care how many seats you have. Boring has always been a good business. Now it’s a valuable bit of security.
This Week on the Work Tech Weekly Podcast
In the latest episode of Work Tech Weekly, I sat down with Heidi Barnett, President of isolved Talent Acquisition. We talked about a very 2026 problem: AI made it easy to send a polished application to a hundred jobs, and it made a lot of those applications look alike. That’s why she tells friends who are job hunting to “get human.” A handwritten note is hard to mistake for AI drivel.
It’s a great conversation. Give it a listen.
Transactions
-
Dayforce moves to acquire Ribbon AI to accelerate autonomous recruiting. Ribbon’s voice AI interviewer has completed more than one million candidate interviews, and now it’s getting folded into an HCM suite. The catch: Ribbon customers on Workday, Greenhouse and other ATS platforms have no continuity promise yet. That’s the classic “great partner, until somebody buys them” moment. (HR Tech Feed)
-
Bluewater Learning acquired by NIIT Learning Systems. Dallas-based Bluewater is highly regarded firm that boasts deep partnerships with Cornerstone and Docebo. NIIT is clearly rolling up the implementation layer, after buying SweetRush back in January. In learning tech, the services partners are becoming the acquisition targets. (Press Release)
-
Avante raises $30 million Series A. The Seattle startup builds AI agents for corporate benefits teams, and its CEO’s pitch is that those teams still rely on outdated information, stale spreadsheets and technology built for another era. Benefits is one of the biggest line items any company manages. It’s long overdue for something better than Excel. (FinSMEs)
-
Plato raises $16 million for AI job simulation tech. Plato is building simulated workplaces where AI models practice using software and completing complex jobs before they operate in live systems. We’ve apparently built better onboarding for bots than most companies have for humans. (HR Tech Feed)
-
Mintive raises $7 million. Once agents start doing real work, “who is this thing and what’s it allowed to touch” becomes an identity question that IT and HR will both claim to own. (TechCrunch)
-
Colare secures $2.5 million to fix the costly bottleneck in hardtech engineering hiring. By Colare’s estimate, hiring one mechanical, electrical or embedded engineer consumes roughly 40 hours of senior engineering time. Simulations show up three times in this week’s deals (Plato, Ribbon, Colare), which suggests “skills-based hiring” finally has actual tooling behind it. (HR Tech Feed)
Industry Notes
-
ServiceNow and Google Cloud expand partnership to connect industry workflows with Gemini Enterprise. Every platform is now partnering with every model, and the real bet is being the switchboard rather than the brain. Announced right before ServiceNow’s earnings face-plant, this one got approximately zero attention. (Seeking Alpha)
-
PNC Bank and ADP announce an embedded payroll solution for small businesses. PNC Payroll by ADP will connect payroll, HR and benefits with the banking, payments and cash management capabilities businesses use every day. Payroll is moving to wherever the money already lives, and that’s a distribution problem for every standalone SMB payroll player. (ADP)
-
Anthropic’s IPO filing reveals its strategy for keeping enterprise customers hooked. The S-1 shows 900% net revenue retention and more than 11,900 enterprise customers generating at least $100,000 in trailing-12-month revenue. A Morningstar analyst adds the fine print: there’s no retention data yet on any of its customer cohorts. Stickiness is the pitch. Whether it holds is next year’s story. (MSN)
-
OpenAI can’t raise the capital it needs. This is the awkward counterweight to Anthropic’s IPO week: one lab is pitching profits to Wall Street while the other is still passing the hat. (LinkedIn)
-
HR Tech 2026: 10 observations from the show floor. ERE’s ground-level recap of the state of talent acquisition. In a word: troubled. (LinkedIn)
-
Unleash Paris 2026: what I took home. Anita Lettink’s takeaways from the European side of the conference circuit, where the AI conversation tends to come with more regulation attached. (LinkedIn)
Worth Reading
-
After 28 years walking across the globe, he’s almost home. Karl Bushby set out on foot from the southern tip of South America in 1998 to walk home to England. Your commute complaint has been formally dismissed. (The New York Times)
-
The AI revenue boom: two challenges hiding behind the growth. Chris French pokes at what the top-line numbers aren’t telling you. That’s useful framing in a week when “adjusted” got bandied about quite a bit across Work Tech. (LinkedIn)
-
Mathmeticians are having an existential crisis. A.I. is solving problems that stumped the field for decades. That’s no bueno for some really, really smart people. (The New Yorker)
