AI Isn’t Killing SaaS—It’s Taking the Budget

Software stocks sliding into a bear market isn’t a vibe. It’s math.

In the last week of January, the software sector took the kind of hit that resets boardroom conversations: IGV fell 22% from highs, January 29 became the worst single day for software since the Covid crash, and companies beat earnings and still got punished (ServiceNow dropped 11% even after another strong quarter). Microsoft shed $360B in market cap in a day.

A loud explanation is making the rounds: “AI agents will replace SaaS.” It’s a clean story, and it fits into a headline. The reality in U.S. enterprise tech is more uncomfortable—and more useful if you build, buy, market, or invest in software.

AI isn’t replacing SaaS systems of record. It’s starving SaaS growth by redirecting budgets, changing buying criteria, and shrinking seat counts. In this installment of our series on How AI Is Powering Technology and Digital Services in the United States, that’s the shift worth paying attention to—because it’s also where the next wave of durable software businesses will come from.

The myth: “AI agents will kill Salesforce”

AI-assisted building is real, fast, and wildly productive—but it doesn’t replace enterprise software. Tools like “vibe coding” can ship a prototype in hours. I’ve seen teams build internal apps over a weekend that would’ve taken a quarter in 2022. That’s not hype.

What’s hype is the leap from “we can build fast” to “we can replace Salesforce, Workday, ServiceNow, or a regulated system of record.” The missing part is the 98% of enterprise software work that shows up after v1.

The 98% problem: shipping v1 is the easy part

A production-grade enterprise app isn’t a feature list—it’s an operating model:

• Security audits and ongoing monitoring (not a one-time checkbox)
• Compliance requirements (SOC 2, HIPAA, PCI, FedRAMP, depending on market)
• Role-based access, data retention, and governance
• Integrations with dozens to hundreds of tools
• Change management and admin workflows
• Performance and uptime at enterprise scale

This is why U.S. enterprises don’t casually “replace the CRM” because someone built a nice UI in a dev tool. Systems of record survive because they’re where data is created, validated, controlled, and audited.

Memorable rule: AI can help you build software faster. It doesn’t volunteer to be on-call for it.

The reality: SaaS isn’t being killed—it’s being re-priced

The current shakeout is a re-rating of growth, not a funeral for SaaS. Public SaaS growth has been decelerating since the 2021 peak. Every quarter. That trend didn’t start with agents.

The market finally stopped paying “future re-acceleration” multiples for companies showing a different pattern:

• Growth driven by price increases
• Expansion inside existing accounts
• Weak net new logo momentum

That’s not nothing—it’s just not the growth profile that commands premium valuations.

Why 2026 doesn’t feel like 2016

There was a major SaaS drawdown in 2016, too. The key difference is what’s behind the belt-tightening.

• 2016 was cyclical: budgets paused, then resumed.
• 2026 is structural: budgets are growing overall, but AI is taking a larger share.

That changes the question from “when will they buy?” to “what will they buy instead?”

The budget shift behind the “AI crash” narrative

The most important numbers in the current SaaS debate are budget allocation numbers. Jason Lemkin’s framing is blunt and correct:

• AI budgets: up 100%+ year over year
• Overall IT budgets: up ~8%
• App count: flat
• Net new customers: declining
• Seat counts: under pressure

Do the subtraction. If AI spend doubles and total IT is up single digits, the money is coming from somewhere. In many U.S. companies, it’s coming from:

• Fewer new app purchases
• Renegotiated renewals
• Slower expansions
• Tighter seat provisioning

This is also why “AI vs. SaaS” is the wrong debate. Most enterprise buyers aren’t choosing between “CRM” and “agent.” They’re choosing between:

• Another module, another seat bundle, another vendor
• Or GPU-heavy infrastructure, model access, data pipelines, security tooling, AI teams, and governance

AI isn’t eating the product. It’s eating the budget.

The five forces pressuring SaaS—what they mean in practice

If you’re building digital services or SaaS in the U.S., these are the pressures shaping your pipeline right now. The point isn’t to panic. The point is to design for how buyers are behaving.

1) Budget reallocation to AI infrastructure and headcount

Hyperscalers are pouring money into AI infrastructure in 2026—on a scale that pulls spending and attention across the entire enterprise ecosystem.

When leadership teams approve AI initiatives, they often fund them by holding the line elsewhere. That means SaaS teams get asked to:

• Prove ROI faster
• Reduce total cost (not just “improve productivity”)
• Offer pricing that matches usage/outcomes

Practical move: If your product supports AI workflows (data access, approvals, monitoring, security), position it as an AI-enabler tied to the AI budget line item.

2) App fatigue and vendor consolidation

Most CIOs want fewer apps. The “best-of-breed” sprawl created integration debt, security risk, and admin overload. AI made that sprawl feel even more expensive because it adds another layer: data readiness and governance.

Practical move: Sell a “replace 3 tools” narrative with a credible migration plan. If you can’t, become the platform’s preferred extension—integrate deeply, don’t sit adjacent.

3) Seat counts get squeezed as work gets automated

This is the quiet revenue killer for seat-based SaaS.

If AI automation reduces headcount or changes job design (fewer coordinators, fewer SDRs, fewer tier-1 support agents), the number of paid seats falls even if the company is “growing.” You may still deliver value—your pricing model just stops matching it.

Practical move: Shift from seat-based pricing to:

• Consumption pricing (per workflow, per call, per document, per ticket)
• Outcome-based pricing (per resolved case, per qualified meeting, per invoice processed)

These models align with how AI value is measured: output, not logins.

4) “Growth” driven by price increases is fragile

Price increases can sustain revenue—until customers get a credible alternative.

AI is creating alternatives in two ways:

• Better automation reduces the perceived need for add-ons.
• Faster internal tooling makes “good enough” workflows easier to build for non-core needs.

Practical move: If you raise prices, pair it with an AI capability that measurably reduces time-to-outcome (not a chatbot feature). Bring numbers: time saved, tickets avoided, cycle time reduced.

5) AI makes older SaaS UX feel dated

User expectations changed faster than most roadmaps. Once teams get used to natural language interfaces, summaries, suggested actions, and automated workflows, a 2019-style dashboard-and-forms app feels slow.

But bolting on a chat window won’t fix the problem. The real upgrade is redesigning workflows around:

• Asking for intent, not fields
• Automating routine steps
• Generating drafts (emails, tickets, notes, reports)
• Explaining actions for auditability

Practical move: Pick one workflow that drives daily usage and rebuild it “AI-first.” Don’t spread effort across 30 half-features.

What U.S. SaaS leaders should do next (founders + revenue teams)

The winners in 2026 are the companies that get paid from AI budgets while reinforcing systems of record. Here’s an operating checklist that I’ve found actually holds up in enterprise conversations.

Reposition around outcomes, not seats

Replace “per user per month” messaging with a clear economic promise:

• “We reduce onboarding time from 14 days to 3.”
• “We cut tier-1 tickets by 25% by auto-resolving common issues.”
• “We shorten quote-to-cash cycles by 18% through automated approvals.”

Outcome language isn’t marketing fluff. It’s how you stay funded when budgets get reallocated.

Make your product part of the AI stack

Most enterprises are assembling an AI stack with familiar categories:

• Data access and quality
• Governance and policy
• Security and privacy controls
• Workflow automation
• Observability and evaluation

If you can credibly sit in one of those categories, you’re not “another app.” You’re infrastructure for AI-powered digital services.

Protect the system-of-record advantage

If you own the data layer—or you’re deeply embedded in where data is written back—you’re durable.

If you’re “just a UI,” you’re exposed to agents that can hop between tools.

Design goal: Make your app the trusted place where AI outputs are approved, logged, governed, and auditable.

Plan for slower growth and higher proof

The era of effortless expansion is over. Assume:

• Longer sales cycles
• More security reviews
• More procurement scrutiny
• More pricing pressure

The teams that win are the ones who treat this as a product and go-to-market redesign problem, not a “wait for the market to bounce” problem.

What this means for the broader “AI is powering U.S. digital services” story

The most useful reframing is simple: AI isn’t deleting enterprise SaaS. It’s forcing SaaS to earn its keep in a new budget reality. That’s how AI ends up powering technology and digital services in the United States—by pushing vendors to automate the hard parts: compliance, security, integration, workflow execution, and measurable outcomes.

If you’re a founder or operator, don’t waste cycles arguing with the “agents will kill SaaS” crowd. Spend those cycles answering two questions your buyers already care about:

1. Are you capturing AI spend—or are you funding it?
2. Can you prove value without relying on seat expansion?

Teams that can answer “yes” and “yes” will still build very large businesses. The bar is higher, and honestly, that’s healthy.

What are you rebuilding first: your pricing model, your core workflow, or your data-layer advantage?