The market narrative says AI agents are about to gut enterprise software budgets. ETR Insights' SaaS Displacement Panel says that isn't happening yet. ETR brought together four senior technology leaders – a director of solutions architecture and a senior director of AI engineering at large technology enterprises, a director of IT at a large industrials enterprise, and a vice president of IT portfolio management at a large business services enterprise – to walk through where AI is actually changing vendor relationships, and where it is mostly changing vendor marketing. Their answers touch Microsoft, Oracle, Salesforce, Workday, Dropbox, Snowflake, Databricks, and dozens of other SaaS names.
AI is not yet displacing SaaS at enterprise scale. Where vendor changes are happening, they remain predominantly SaaS-to-SaaS swaps, and SaaS budgets are still growing rather than shrinking.
The build-versus-buy debate favors buy for most categories. Enterprises are consuming embedded AI from incumbent vendors rather than building custom alternatives, and one panelist reports 20% to 25% productivity gains from embedded AI already in production.
AI-led change outpaces traditional vendor switching in only 2 of the 12 software categories ETR tracked in its SaaS Displacement Drill Down survey: developer tools and data analytics/BI.
Data management is the most disruption-resistant category, and the only one where internal AI builds (20% over the next twelve months) exceed AI assistant adoption (13%), reflecting the structural complexity of enterprise data that off-the-shelf agents can't yet navigate.
Panelists converge on Microsoft as the consensus AI-era winner, split sharply on Oracle, and offer a contrarian bull case for Salesforce anchored in install-base durability. Workday is flagged as lagging in HCM, and point solutions like Dropbox are flagged as absorption risks.
All four panelists say no, not yet, and the gap between the headlines and enterprise reality is wide. "There's definitely not the level of replacement that everyone is advertising of, oh, we're going to be able to do headcount reduction and save all these costs," said a vice president of IT portfolio management. "It's not there. It's too early. It's basically looking at how we can use AI to enhance the overall experience of some of our workers... but in no way is it replacing anything at this juncture." A director of IT at a large industrials enterprise made the same point from a different angle, arguing organizations gravitate toward AI already embedded in tools they depend on rather than building connectors or middleware to replace them. That gap matters for investors: the market has priced a lot of urgency into the AI software narrative, but inside enterprises, risk tolerance, switching costs, security, governance, and integration complexity still govern software decisions.
A senior director of AI engineering confirmed that SaaS budgets are still increasing, not contracting, even as AI gets embedded across analytics and decision-support layers. Across ETR's underlying survey, AI-led change outpaced traditional SaaS-to-SaaS switching in only 2 of 12 software categories tracked, developer tools and data analytics/BI, with vendor-to-vendor competition remaining the primary driver of change everywhere else. As one panelist framed it, the shift underway is real but is reshaping SaaS vendors rather than eliminating the category: "Yes, there is a disruption in the sense that we are changing our direction. Just like what we did when we went from client-server to Web, a similar kind of disruption is happening." AI may even be a near-term spending tailwind rather than a deflationary one, as vendors attach AI features and buyers expand licenses to use them. The better question for this category isn't whether SaaS survives AI. It's which vendors use AI to improve retention, expand wallet share, and defend pricing.
For most categories, yes, and buy won. A director of solutions architecture put it plainly: "Nobody wants to build AI applications from scratch if something is available embedded in the SaaS application itself. These built-in applications with AI features give us a great head start." Another panelist, burned by years of over-customized legacy systems, has no appetite to repeat the exercise: "I think we paid for the sins of the past with a lot of legacy applications that are sort of customized to death. Why would I try and build something that is either equal or better to what is already in the market?" This is where incumbency becomes a strength: embedded AI inherits existing security models, data grounding, user familiarity, and platform maturity, advantages that often matter more to large enterprises than the flashiest AI demo.
The data adds a wrinkle worth holding alongside the buy-side story: in ETR's September 2025 AI Product Series, 84% of organizations building their own AI applications still expected to increase that spending over the next twelve months. Embedded AI is winning the SaaS displacement question, but internal AI application development is a separately protected budget line, not a shrinking one. Panelists weren't anti-AI, they were anti-hype. One vice president is open to embedded AI but wants proof first, not novelty: "We are more attracted to vendors that are not on the bleeding edge, that are not just throwing everything AI. We want to see them integrated successfully." Even so, AI roadmaps are becoming a bigger factor in vendor evaluations, and one senior director reporting 20% to 25% productivity gains from embedded AI said the next step is consolidating AI workflows enterprise-wide: "Now it's time for us to really figure out how do we really consolidate and do this across the entire corporation."
Developer tools and data analytics, where the use cases are proven and organizations have a structural knowledge advantage over vendors. A vice president of IT portfolio management explained why internal builds cluster here rather than in customer-facing applications: "You're building it instead of starting with someone else's box of what they have and trying to put your pieces in." Legacy code conversion is the clearest example in practice: "AI can sit there and probably do 80% of the majority of the work in a conversion of some C++ or Python kind of database from a COBOL database or COBOL code, and get you over there very quickly."
The industrials IT director frames the pattern as a trust curve, the same one that played out with public cloud adoption: skepticism up front, then broad adoption once lower-risk use cases prove out. "People said that, oh, I'm not moving anything to the public cloud. We'll never do that, ever. And then all of a sudden you see different business units... move to the public cloud, and now you have that grand adoption. I think the same thing applies here." Categories like CRM, core HCM, ITSM, and ERP remain far stickier, with agentic features from Salesforce Agentforce, ServiceNow Now Assist, and Workday reinforcing retention rather than triggering switches. One panelist flagged agentic architecture itself as the disruption vector worth watching most closely: "If you look at the trend of what Google and NVIDIA, you name it, even Salesforce, moving to agentic architectures, this is something that we have to watch out for. This has the capability for autonomous workflow execution."
Panelists converged fastest on Microsoft as the consensus winner, citing breadth across the enterprise stack: "They've got their hands in everything. They're dominating the enterprise SaaS ecosystems through all kinds of applications: productivity, security, AI, and copilots." Oracle was the most contested name. Two panelists see it as a loser, strong in AI infrastructure but not a credible cloud-native SaaS competitor, while a third pushed back based on direct use: "We are using them, they're evolving."
Salesforce drew the panel's most contrarian call, with one panelist arguing its install base will prove more durable than the market currently prices in: "I actually think Salesforce is going to be a winner. You just have some resistance in the market, because they've maybe been a little bit slower to adopt than some of the other providers." ETR's CRM category data lends the durability case some support: two-thirds of respondents report no vendor change, but among the minority who are moving, full displacement nearly doubled, from 5% to 9%, while partial displacement declined, suggesting the accounts that do act are becoming more decisive rather than less.
That preliminary April 2026 Technology Spending Intentions Survey (TSIS) data lines up with the panel's read: Microsoft and Salesforce hold strong absolute Net Scores among Global 2000 respondents, while Oracle sits near the bottom of the group. monday.com and UiPath show the sharpest year-over-year Net Score declines, consistent with the panel's broader view that narrower point solutions face rising competitive pressure. On that theme, one panelist flagged Workday as lagging on AI features in HCM relative to Oracle and SAP, and another named Dropbox as a niche point solution at real absorption risk: "There's a high chance of replacement, because it's a point solution and it could easily be absorbed right into any of these big vendors." The broader consensus: "You're going to have these smaller ones that are going to get absorbed by the bigger ones, your Microsofts and Googles." That may be one of the most investable takeaways from the panel: AI may not replace SaaS as a category, but it can accelerate consolidation within it. Vendors with broad platforms, trusted data access, workflow depth, and enterprise distribution look better positioned, while vendors with narrow functionality face pricing pressure, bundling pressure, or outright absorption.