Salesforce made waves in February 2026 by introducing the Agentic Work Unit, or AWU, as a new way to measure and potentially price AI agent activity. The Metrics Brothers, Ray "Growth" Rike and Dave "CAC" Kellogg dig into whether the AWU is a legitimate step toward outcome-based pricing, a vendor-specific adoption metric, or just another awkward intermediate measure in the long, messy history of software pricing models.
Episode Highlights:
- Tokens are not business metrics. Ray and Dave open by drawing a clear line: tokens measure text processing and compute consumption, not business outcomes. Nobody walks into a board meeting announcing they processed 14 billion word fragments, and enterprise buyers should not be priced on that basis.
- What Salesforce is actually trying to do with the AWU. Defined as "one discrete task accomplished by an AI agent," the AWU is Salesforce's attempt to bridge the gap between low-level compute metrics and business outcomes. The hosts debate whether it is a pricing metric in waiting, an AI adoption signal, or simply the best available approximation of work performed by agents.
- A short history of bad pricing units. From CPU counts to MIPS to kilocharacters to gigabytes, Ray and Dave trace the long, humbling history of software vendors searching for a pricing metric that maps to value. The Soviet chandelier analogy makes an appearance, courtesy of Appian CEO Matt Calkins.
- Activity versus outcome: the core tension. Ray argues the AWU is directionally right but fundamentally an activity metric, not a true economic outcome. Dave is more skeptical that outcome-based pricing can scale broadly, pointing out that customers say they want value-based pricing until a vendor actually tries to take a cut of the upside.
- Vertical AI applications have the clearest path. Both hosts agree that verticals with well-defined, countable outputs, such as cases resolved in customer support or claims processed in insurance, are best positioned to price on outcomes and may not need the AWU at all.
- The AWU needs a new name, and probably a new definition. Ray and Dave close with the observation that just as NRR took nearly a decade to emerge as a standard SaaS metric, meaningful AI metrics will take time to mature. The AWU, as currently defined, is a Salesforce-specific construct and unlikely to become an industry standard.
If you are a B2B SaaS or AI-Native software operating executive, this conversation on one of the first agentic AI metrics to measure work activity is a great listen.
See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.
