Enterprise SaaS and AI Narrative

Programming note: We've just released our thematic report: Google's TPU and the New Economics of AI Deployment. It's an 18-page analysis deconstructing how Google is breaking Nvidia's monopoly on AI compute by exploiting the market's shift to inference. Access a free copy here.

The Feature, Not the Future

Since the beginning of the AI boom, the enterprise software playbook has been simple: when in doubt, say "AI agent." The market has rewarded any company that could convincingly tell a story about building an autonomous, intelligent workforce that would revolutionize productivity.

But as the hype cycle matures, cracks are starting to show. That's why last week, Workday's numbers brought renewed focus to Wall Street's growing skepticism. And the market was not impressed.

Shares of Workday, the HR and finance software giant, plunged nearly 10% after it reported earnings. On the surface, the numbers were fine. The company even completed a $1.1 billion acquisition of an AI learning platform, Sana, signaling its commitment to the new paradigm. But the stock fell because of a single, brutal detail in its guidance.

After a year of touting its AI momentum, Workday's forecast for next year's subscription revenue barely budged from its previous estimate. The irony is that, by some measures, the AI push is working. Workday can point to impressive adoption metrics: analysts noted that 35% of new deals now include AI components, and 75% of customers are using its Illuminate platform. But the market looked past the feature adoption and focused on the forecast for forward-looking backlog, which fell short of expectations. The market had been promised a revolution, and what it got was a rounding error.

This is the AI hangover. The market is beginning to differentiate between an "AI feature" and an "AI Agent."

An AI Agent is the grand promise: a fully autonomous digital employee that can reason, plan, and act on your behalf. An AI feature is what is actually being shipped: a smarter search bar in your HR portal. Both are useful, but only one justifies a nosebleed valuation.

Workday CEO Carl Eschenbach tried to put a positive spin on it, noting on the earnings call that AI products contributed over 1.5 percentage points of annualized revenue growth. This was meant to be a sign of progress. But for a market that has priced in a technological upheaval, 1.5% isn't exactly a signal of a transformative AI future.

Workday is not alone in this purgatory. The market is reacting not just to Workday's specific numbers, but to a broader pattern of disillusionment. A similar story has been playing out at Salesforce, which offers a perfect case study of the paradox at the heart of the enterprise AI transition.

Here is the central conflict, a dynamic we examined recently in a deep dive. To sell the AI future, Salesforce must prove its AI agents are powerful enough to replace human workers. But its entire business model is built on selling software seats to those very same human workers. The most damning evidence that this cannibalization is real comes from Salesforce itself. In 2025, the company cut 4,000 customer support jobs, with CEO Marc Benioff bluntly stating he "needs less heads" because his own AI is more efficient. This single move validated the market's deepest anxiety, and it's a key reason the stock has slumped 30% this year.

Benioff's admission isn't isolated; it aligns with broader anecdotes from Silicon Valley. For instance, Paul Graham shared on X about a non-software company using AI to run with only 6 employees—versus the 16 they'd need without it, equating to a 2.7x productivity multiplier. While this validates AI's potential to "need less heads," it also exposes the irony for Salesforce: the very technology they're championing could be eroding their core business model faster than it can replace lost revenue.

This leaves the old guard of SaaS fighting a war on three fronts. First, they must spend billions on AI R&D and acquisitions to avoid being disrupted by AI-native startups. Second, that spending squeezes their margins and produces, for now, only incremental revenue gains. And third, they have to do all this while championing a technology that is actively working to obsolete their profitable, per-seat subscription model—a model that Gartner predicts will see 40% of its spending shift away by 2030.

The market has spent a year buying the hype. Now, it appears to be demanding the math. And for incumbents like Workday and Salesforce, the math suggests the agentic future is still a long way off.

More on AI Narrative

• Are Microsoft AI Copilot-Powered 'Frontier Firms' Worth the Cost? (UC Today)
• How Meta's AI Narrative Wins Over Wall Street (ARPU Video)

On Our Radar

Our Intelligence Desk connects the dots across functions—from GTM to Operations—and delivers intelligence tailored for specific roles. Learn more about our bespoke streams.

WhatsApp's AI Lockdown

• The Headline: Meta is changing WhatsApp's API terms to ban third-party general-purpose AI chatbots like ChatGPT and Copilot, effective January 2026, making its own Meta AI the exclusive chatbot on the platform. (TechRadar)
• ARPU's Take: Meta is leveraging its most powerful asset—WhatsApp's massive user base—to give its own AI a fighting chance. By kicking out superior and more popular competitors like ChatGPT and Copilot, it's creating a walled garden where its own unpopular AI can be force-fed to billions of users without competition.
• The Product Question: This move confirms that for major platforms, controlling the default AI experience is a non-negotiable strategic priority. This creates a critical strategic debate for other "super-app" platforms: do they follow Meta's lead and build a closed AI ecosystem to favor their own models, or do they pursue an open, "app store" model that allows for user choice and competition, potentially ceding the primary user relationship to a third-party AI?

Micron's Geopolitical Pivot to Japan

• The Headline: U.S. chipmaker Micron will invest $9.6 billion to build a new high-bandwidth memory (HBM) chip plant in Japan, with significant government subsidies, to meet AI demand and compete with market leader SK Hynix. (Reuters)
• ARPU's Take: This is a shrewd move by Micron to use Japan's industrial policy as a weapon in the memory wars. By tapping into massive Japanese subsidies, Micron is funding its catch-up bid against SK Hynix while simultaneously diversifying its manufacturing footprint away from the geopolitical flashpoint of Taiwan. It's a two-for-one strategic win.
• The Operations Question: This move demonstrates that the race for critical AI components is now deeply intertwined with industrial policy and geopolitics. For supply chain leaders, this sets a new standard for resilience, where de-risking operations now involves actively aligning manufacturing investments with national subsidy programs in geopolitically stable regions, even if it adds complexity to the global footprint.

P.S. Tracking these kinds of complex, cross-functional signals is what we do. If you have a specific intelligence challenge that goes beyond the headlines, get in touch to design your custom intelligence.

You received this message because you are subscribed to ARPU newsletter. If a friend forwarded you this message, sign up here to get it in your inbox.