Gartner just put a number on the thing every CIO has been quietly worrying about: $234 billion. That's how much enterprise application spending is exposed to agentic arbitrage between now and 2030 — roughly 20% of all enterprise SaaS spend, simply up for grabs.
This isn't a prediction about some distant AI future. It's a market correction already underway. SaaS valuations have shed approximately $300 billion over the past 18 months as investors price in the disruption. What took them a year and a half to digest, your enterprise software contracts are just starting to feel.
The question for every technology and business leader right now isn't whether agentic AI will reshape enterprise software spend. Gartner says it will. Deloitte says 74% of companies expect to use agentic AI at least moderately within two years. The question is whether you're positioned to capture the upside — or absorb the cost of the wrong bet.
What "Agentic Arbitrage" Actually Means
The term is worth unpacking before we go further, because it's the mechanism behind the $234 billion figure.
Traditional enterprise software was designed for humans. A finance team member logs into an ERP, opens a treasury platform, pulls from a banking portal, and reconciles invoices across three systems. It takes hours. The software vendor charges per seat — per human who does that work.
Agentic AI breaks that logic entirely. An AI agent can be given a goal, reason through the steps required to achieve it, call APIs across multiple systems, complete the work overnight, flag exceptions, and surface only what needs a human decision. The vendor is no longer selling accounting software seats. It's selling reconciled invoices. Or not — because an agent can now do it without the traditional interface at all.
That gap — where an agent completes tasks across multiple systems, bypassing traditional user interfaces — is what Gartner calls agentic arbitrage. When an agent can do the work without "using" the software in the traditional sense, the per-seat model begins to collapse.
This is not hypothetical. Talking to operations leaders at large enterprises, you hear the same thing: teams are already running agents that touch five or six SaaS systems in a single workflow. The software vendors are getting the API calls. Their per-seat revenue is getting squeezed.
Three Pressure Points on the Enterprise Software Stack
The Deloitte analysis of this shift identifies three structural pressure points that matter to both technical and business leaders.
Pricing models are broken for an agentic world. Per-seat and per-user pricing assumed humans were the unit of work. That assumption is now wrong. One person supervising a team of agents can produce what previously required dozens of licensed seats. Meanwhile, some vendors are adding AI usage fees — token consumption, inference charges — on top of existing licensing costs, creating unpredictable bills. Organizations that locked in multi-year seat-based contracts are starting to see the mismatch. The shift is toward hybrid models that combine outcome-based components with usage metrics, but most enterprise contracts haven't caught up yet.
The user interface is becoming a control tower, not a workspace. When agents do the work, the application's UI stops being a workspace and becomes a dashboard for reviewing outputs and setting goals. Employees spend less time inside applications and more time at the meta-level — deciding what the agents should pursue. This raises a question that no SaaS vendor has cleanly answered yet: where does that control layer live? Inside one vendor's walled garden? In a neutral orchestration layer above the entire stack? Whoever controls that surface controls how agents are directed — and that's where real enterprise leverage will sit.
Orchestration is the new battleground. Large enterprises won't run on a single vendor's suite of agents. They'll run dozens of agents from incumbent suites, cloud AI platforms, purpose-built AI companies, and internal builds. The critical question isn't who has the smartest individual agent. It's who owns the layer above the agents — the orchestration plane that routes work, enforces policy, manages authentication, and ensures compliance. That orchestration layer is where the next generation of enterprise software lock-in will emerge.
How the Major Vendors Are Responding
The incumbents aren't sitting still. Every major enterprise software provider is racing to become the orchestration layer rather than just an orchestration target.
Workday is building a service called Sana — a natural language front door where employees log in and interact with enterprise data through conversational AI. A payroll manager asks what drove variance in compensation costs this quarter; Sana reasons across HR, finance, and workforce data to answer. Workday's bet is that its trusted data relationships with risk-averse enterprise clients are a durable moat — that enterprises won't route sensitive HR and finance data through a neutral AI orchestration layer they don't control.
Salesforce has moved aggressively with Agentforce, positioning itself as the agent platform for customer-facing workflows. Oracle launched AI Agent Studio to let enterprises build agents that run on top of its application suite. ServiceNow is embedding agentic capabilities into its IT workflow products. SAP is integrating AI agents directly into its ERP suite.
The strategic logic is the same across all of them: if agents are going to complete work, those agents should run inside our platforms, using our data, subject to our governance frameworks. That way, agentic arbitrage flows through us rather than around us.
Newer entrants are betting on a different outcome — that orchestration will eventually commoditize the underlying application as just a data source, and that a neutral orchestration plane sitting above the entire enterprise stack is where the value accrues. That battle is still in early innings.
What This Means for Enterprise Leaders
The IDC research director framing is worth holding: "The SaaS apocalypse is overrated, but disintermediation is real." Software won't disappear. What disappears is the assumption that users will always interact with it through traditional interfaces, at a per-seat cost.
For CIOs and CTOs, this is a vendor relationship and architecture question. The software stack that made sense in a world of human-driven workflows may not be the right stack for agent-driven workflows. More importantly, the vendors with durable capabilities — unique data, trusted workflows, deep compliance frameworks — will survive and evolve. Vendors whose only differentiator was a well-designed user interface face genuine disruption.
The most important architectural question to be asking vendors right now: what is your MCP (Model Context Protocol) or agent integration story? How do agents interact with your system — through APIs, through native agent frameworks, through your own orchestration layer? Vendors without clear answers to that question are behind.
For CFOs and business leaders, this is a contract and cost structure question. Multi-year seat-based contracts signed in 2023 or 2024 may be over-indexed on seats that agents are making redundant. The time to renegotiate is before renewal, with data on how much of the licensed capacity is now being handled by agents. Simultaneously, watch for vendors who add AI usage fees on top of existing licensing — this is becoming a significant and often unexpected cost driver.
For COOs and department heads, this is an operating model question. If an agent can complete end-to-end invoice reconciliation, contract review, customer onboarding verification, or expense processing — the question isn't just "do we adopt it?" It's "how does our team structure change when agents handle the execution layer and humans own the judgment layer?" That org design question is lagging well behind the technology deployment.
Three Things to Do in the Next 90 Days
Based on conversations with enterprise leaders navigating this shift, three actions stand out as high-leverage in the near term.
Audit your software contracts against actual usage. Before your next renewal cycle, run a utilization analysis. Which licenses are being accessed by humans versus being called via API by agents? Which seats went from 80% utilized to 30% utilized as agents took on more routine tasks? That data is your negotiating leverage — both for cost reduction and for pushing vendors toward outcome-based or consumption-based pricing.
Map your orchestration strategy. You will not run your enterprise on one vendor's agent suite. Decide now whether your orchestration layer is going to be a hyperscaler product (Microsoft, Google, AWS), a neutral AI platform, your existing enterprise suite's native orchestration, or a custom internal layer. Each choice has different cost, governance, and vendor dependency implications. Not deciding is itself a decision — usually an expensive one.
Build your data foundation for agents. This is the most consistent gap in enterprises that attempt to deploy agents at scale. Deloitte's analysis is direct: access, observability, lineage, and governance are the prerequisites to deploying agents safely. Most enterprises overestimate how ready their data is. An agent that can't trust the data it's acting on becomes either useless or dangerous. Investing in data readiness is the highest-leverage infrastructure investment you can make ahead of agentic deployment at scale.
The Bottom Line
Gartner's $234 billion figure is not a death sentence for enterprise software. It's a market signal about where value is migrating — from application access to agent-completed actions. The vendors who understand that are racing to become the orchestration and governance layer. The ones who don't are at genuine risk of becoming commodity data sources.
For enterprise leaders, the strategic imperative is clarity: understand how agents interact with your software stack today, know what your contracts actually reflect versus what your utilization shows, and decide where your orchestration layer sits before someone else decides it for you.
The software stack your enterprise runs on in 2030 will look meaningfully different from today's. The decisions you make in the next 12 to 18 months will determine whether you're positioned to capture the efficiency and cost advantage — or whether you're paying legacy seat-based prices for a world that no longer matches how your enterprise actually works.
The enterprise AI landscape is moving fast. If you found this useful, share it with the CIO, CFO, or operations leader in your network who's navigating software contract renewals right now.
Follow on LinkedIn or X/Twitter for more enterprise AI analysis.
