AI agents shift software sales from seats to completed work

The fundamental transaction in enterprise software is changing. For decades, companies bought software licenses and their employees used those tools to complete work. Now AI agents are performing enough of that work independently that vendors and customers alike are rethinking what's actually being purchased.

Salesforce provides a clear example of the shift. The company reported that Agentforce, its platform for building and deploying AI agents, reached $1.5 billion in annual recurring revenue, up more than 240% year over year, according to No Jitter. The company also expanded its Anthropic partnership so sales professionals can access Salesforce data and workflows through Claude without opening Salesforce itself.

Patrick Stokes, Salesforce's president of applications and marketing, distinguished between agents that help knowledge workers and those designed for "autonomous work or work that touches the end customer." That distinction matters commercially. Software that helps an employee do their job creates a different vendor relationship than a platform that performs defined work on the customer's behalf.

Why it matters

This transition forces enterprise buyers to reconsider budget allocation across software, headcount, and services—categories that have historically been separate. When AI can perform work previously done by employees or outsourced providers, the line between buying software and buying services blurs. That creates pressure on traditional pricing models and raises new questions about quality, accountability, and measurement.

From interface access to work delivery

The shift is particularly visible in customer success operations. Companies typically assign dedicated customer success managers only to larger accounts because the economics don't support giving every customer intensive human attention. Smaller customers receive pooled support or digital resources.

AI could make active engagement practical across accounts that haven't traditionally warranted dedicated coverage, including adoption monitoring, risk identification, and renewal support. Forrester analyst Shari Srebnick has noted that advanced organizations are moving beyond asking what to automate and instead rethinking how they'd design customer success if AI were available from the start.

Gainsight, which provides customer success software, launched Atlas, an AI-native service for managing long-tail renewals. CEO Chuck Ganapathi describes the opportunity as work that's important and repetitive enough to systematize but has enough variation to require an agent rather than a script. Customers can build AI themselves, purchase agents, or effectively hire Gainsight to perform the work—or combine approaches.

The pricing challenge

These different delivery methods create pricing complexity. Salesforce's Claude integration demonstrates the issue: a seller can use Salesforce data and workflows without opening Salesforce, yet Salesforce can still charge for that activity through consumption-based pricing. The underlying data, permissions, and workflows retain value even when the application interface isn't used.

Salesforce acquired m3ter this year, a company whose technology supports usage and outcome-based pricing. Consumption offers one way to charge when software rather than a person uses an application, but it measures usage rather than value. Tokens, API calls, and agent actions don't indicate whether useful work was completed, done well, or what it was worth.

Outcome-based pricing addresses this gap by tying payment to completed work or results. Gartner expects outcome-based pricing to account for 40% of AI-agent vendor customer-service revenue by 2029. However, Gartner also points to the difficulty of defining and measuring outcomes. A resolved service request doesn't necessarily mean a satisfied customer.

Accountability remains complex

Performing more work doesn't mean vendors assume responsibility for everything that follows. Customers still control the policies, data, and permissions systems operate under. Buyers need clarity on what the system will do, how they'll assess quality, and what happens when it fails.

Boston Consulting Group's Akash Bhatia, speaking at the recent AI-Native Services Summit, argued that companies succeeding in this space need domain expertise combined with technology that delivers work efficiently at scale. Emergence Capital, which hosted the event, contends that companies built this way can achieve gross margins above 50% while delivering five to ten times greater speed than traditional services businesses.

These details were first reported by Rob Hilsen in No Jitter.