The rally and what triggered it
Enterprise software had a loud day on July 27. ServiceNow closed up roughly 7 percent, and the broader software and IT services group rose about 6 percent, according to market-mover data from TradingKey. Coverage from 24/7 Wall St put Salesforce up around 7 percent, ServiceNow up 8 percent, and Workday up 10 percent on the session, framed as a rotation back into enterprise software after months of AI-driven skepticism. The proximate cause was a run of Q2 results that beat the low bar the market had set.
The setup makes the move more interesting than a single earnings pop. Investors had spent much of 2026 pricing in the idea that autonomous agents would hollow out seat-based subscription revenue. When SAP, ServiceNow, and Workday instead reported expansion and agent monetization, the thesis met its first full earnings season and did not hold. For technology leaders, the stock reaction is secondary. The signal that matters is that the incumbents are converting agents into revenue rather than watching agents cannibalize them.
How brutal the drawdown actually was
The rebound only makes sense against the damage that preceded it. Industry analyses of the first half of 2026 estimate that close to 2 trillion dollars in market value came out of the enterprise software sector, with Salesforce reported down roughly 190 billion dollars from its peak and Adobe down about 160 billion. Those are not recession numbers. They are the market repricing a specific fear, that agents from Anthropic, OpenAI, and the platform vendors would eat the workflows subscription software was built to serve.
Some of the operational data fed the fear. Reporting through the year noted Atlassian's first-ever decline in enterprise seat counts and Workday itself cutting about 8.5 percent of its workforce, an awkward look for a company that sells workforce software. We take those data points seriously, because they show the disruption is real at the margins. The question the July results answered is whether the disruption is broad displacement or a narrower squeeze on weak products, and the evidence points to the latter.
The Q2 scoreboard for incumbents
The numbers under the rally are what give it substance. SAP reported cloud revenue up roughly 22 to 24 percent, with cloud backlog climbing, and reaffirmed the direction of its business despite absorbing acquisition costs. ServiceNow beat on subscription revenue and operating margin and raised full-year guidance, citing large-deal momentum. Workday pointed to agentic AI approaching 500 million dollars in annualized revenue and more than 4,000 clients using at least one agent it built, and Monness Crespi upgraded the stock to Buy with a 150 dollar target after a roughly 45 percent drawdown this year.
Read together, these results describe monetization, not erosion. The agents these vendors ship run inside the systems of record customers already depend on, which is why adoption shows up as expansion revenue rather than churn. That is the mechanism skeptics underweighted. An agent that books a journal entry in Workday or resolves an incident in ServiceNow deepens the customer's dependence on that platform, and the vendor prices the agent on top of the seat it was supposed to replace.
Why the incumbent moats held
The pattern separating winners from losers is about moats. Vendors with broad distribution, deep workflow lock-in, and proprietary data can point their AI at problems no independent agent can touch without that context. A standalone agent can draft an email, and it cannot reconcile your subledger without the subledger, the permissions, and the audit trail that live inside the incumbent's platform. That is why analysts arguing the thesis is too simplistic keep landing on the same conclusion, that weak and easily replicated software is under real pressure while entrenched systems of record are compounding.
For buyers, this reframes where the genuine agent threat sits in your portfolio. The point tools with shallow data and thin workflow, the single-purpose apps a well-built internal agent could replicate, are the ones exposed. The core platforms are harder to dislodge precisely because their lock-in now cuts in their favor. If you have spent years resenting that lock-in, the uncomfortable truth this quarter is that it is also what makes their AI credible in production.
The threat that is real: budget reallocation
The more accurate version of the SaaS risk is financial, not architectural. Every dollar routed to AI infrastructure, tooling, and headcount is a dollar not spent on another Salesforce seat, another Workday module, or another ServiceNow add-on. Subscription software is being starved of incremental budget as AI line items expand, and that pressure is real even where displacement is not. This is the mechanism CIOs should be modeling, because it hits net revenue retention through slower expansion long before any agent replaces a product outright.
That framing changes how you should run renewals. The question is no longer whether a vendor is good, it is whether each seat and module still earns its place against an agent alternative you could build or buy. Steady growth from an established platform is a baseline to justify against, not a reason for autopilot procurement. The vendors that keep expanding through this will be the ones that make their agents indispensable inside the workflow before your finance team reallocates the budget elsewhere.
What to do with the rebound
For technology leaders, the July rally is a useful reality check on a thesis that had hardened into consensus. Agents are not clearing the field of enterprise software, and the incumbents that own the workflow are turning AI into expansion revenue. That should temper any plan built on the premise that you can rip out your systems of record and replace them with a thin agent layer this year. The switching cost and the data gravity are still where they were.
It should not, though, make you complacent about spend. Run the portfolio in two buckets, the entrenched platforms where AI deepens the moat and consolidation makes sense, and the replaceable point tools where an internal agent could carry the load. Push your core vendors to prove agent value inside the contract, and redirect the savings from the second bucket toward the AI capabilities that actually move your numbers. The thesis failed its Q2 test, and the discipline it forced on your software budget should survive anyway.
