Agentic Arbitrage: What Gartner’s Forecast Does and Doesn’t Say

What agentic arbitrage means

Agentic arbitrage is Gartner’s name for an effect on enterprise software: AI agents completing work across several systems, so people have less need to use each application’s interface.

Gartner’s release defines it as AI agents completing tasks across multiple systems, which reduces the need for users to interact with multiple traditional software interfaces. Its analyst George Brocklehurst, a Managing Vice President at Gartner, says agentic systems deliver outcomes directly, bypass user-experience-heavy applications and make the software invisible. The release does not describe software as going away. It says SaaS will not be destroyed and will emerge in a different form.

What Gartner’s agentic arbitrage forecast says

Gartner estimates that up to $234 billion of enterprise application spending is exposed to agentic arbitrage between now and 2030.

The release, dated 1 July 2026, adds that by 2030 this will account for roughly 20% of enterprise application SaaS spending. The wording matters in two ways. A ceiling of up to $234 billion is not a central estimate, and exposed describes risk rather than realised loss. The release headline uses at risk in the same sense.

Claim

Up to $234 billion of enterprise application spending is exposed to agentic arbitrage between now and 2030, roughly 20% of enterprise application SaaS spending by 2030.

Claim type

Forecast (analyst estimate)

Original source

Gartner press release

Publication date

1 July 2026

Research or measurement date

Not stated in the release.

Sample and population

Not stated in the release. The spending named is enterprise application software and SaaS spending.

Method or definition

Method not stated in the release. Gartner defines agentic arbitrage as AI agents completing tasks across multiple systems, reducing the need for users to interact with multiple traditional software interfaces.

Limitations

A forecast to 2030, not a measurement. The release does not state whether $234 billion is annual or cumulative, or the base spending behind the 20%. No published measurement of this forecast was found as of 29 October 2026.

Editorial interpretation

Editorial analysis: cite it as a Gartner estimate of exposure with its 2030 window, not as spending that will be lost.

What the release does not say about agentic arbitrage

The release leaves several questions open, and some published coverage fills those gaps with statements Gartner does not make.

As published, the release text does not describe a method, a sample or a base year. It does not state the total spending that the 20% refers to, and it does not say whether $234 billion is an annual or a cumulative amount.

On the term itself, Gartner uses it in the release, and a Gartner research abstract dated 30 June 2026 carries it in its title, one day earlier. The full research is paywalled and was not reviewed, so the abstract adds no method. The release does not claim that Gartner coined the phrase, and this article does not claim it either. Gartner’s own materials also vary in wording: the release headline says at risk, its body says exposed, and the abstract summary describes up to 20% of spend as exposed to displacement. None of these is a measurement of spending that has already moved.

A CIO article published on 2 July 2026 that quotes the same Gartner analyst does not add a method or a base figure either.

The table compares statements seen in some published coverage with what the release says.

How agentic arbitrage connects to seat-based pricing

Gartner’s explanation for the exposure is that agentic arbitrage weakens the link between how many people use enterprise software and how much its vendors earn.

The release headline says agentic arbitrage breaks the traditional SaaS seat-license model. In the release, Brocklehurst says agentic systems break the link between user growth and revenue growth for many enterprise software vendors. It names vendors defending legacy dashboards and seat-based models as facing an existential threat, and it describes a revenue opportunity for vendors that enable agentic cross-domain workflows.

Gartner describes the overall shift as “less an apocalypse and more of a metamorphosis,” and says SaaS will emerge in a different form.

This is Gartner’s reasoning rather than a measured result. This article does not test it against seat-pricing data, which would need separate, verified evidence.

What Gartner tells vendors and buyers to do

Gartner’s advice splits by audience: vendors are told to move from interface-based value to outcome-based value, and buyers are told to check how agents can use the software they already own.

For vendors, the release says incumbents must move from interface-based value to outcome-based value, embed agentic capabilities at the point of execution, and capture and retain customer-specific knowledge, not just data. It says AI-native startups and service providers can act as the agentic layer across enterprise systems. It also notes that some vendors already offer autonomous end-to-end workflow execution, which today typically needs heavy services engagement.

For buyers, the CIO article reports advice from the same analyst. Organisations should check whether an agent can do through a system’s API everything a person can do through its screens, and whether the vendor’s terms permit it. Vendor terms can restrict third-party autonomous use, so contracts deserve as much scrutiny as the technology, and agent permissions are best negotiated into agreements now. The article also describes Gartner’s Knowledge Retention Rate, an organisation’s ability to retain what AI systems learn from its operations, and says the key contract question is who owns what the system learns from the customer. Gartner also advised treating agent autonomy as an explicit governance decision: where agents may act independently, who authorises it, and how often permissions are reviewed.

Claim

Buyers should check whether an agent can do through a system’s API everything a person can do through its screens, check vendor terms on autonomous use, and treat agent autonomy as a governance decision.

Claim type

Opinion (expert guidance reported by a news outlet)

Original source

CIO

Publication date

2 July 2026

Research or measurement date

Not applicable. The guidance comes from an interview.

Sample and population

One Gartner Managing Vice President, George Brocklehurst, speaking to CIO.

Method or definition

Interview reported in a news article. No method is described.

Limitations

One analyst’s guidance as reported by one outlet. It has not been tested or measured here.

Editorial interpretation

Editorial interpretation: these are practical checks for software buyers, not measured outcomes.