SaaS Shifts to Agent Work Unit Metering

The Breakthrough

Salesforce Agentforce achieved an $800 million run rate (up 169% year-over-year) by metering 2.4 billion agentic work units—actions like updating records, summarizing cases, or executing workflows—via Flex Credits, decoupling pricing from human seats.

What Actually Worked

• Salesforce bills agent actions through Flex Credits and agentic work units; each action (update record, summarize case, answer inquiry, run workflow) draws from a shared meter, supplementing but not replacing seats.
• Microsoft Copilot Studio uses Copilot credits with tiered rates for features like answers, generative answers, agent actions, grounding in the graph, flow actions, and premium reasoning; a 100-seat company faces scaling costs for runtime credits.
• ServiceNow Action Fabric meters operational work (provision access, escalate incident, onboard, open change request) through governed pathways with identity, permissions, and audit, claiming workflow substrate value.
• Fair agent licenses require visible/transparent meters, forecastable usage, non-billable failed work, governed paths for third-party agents, distinctions between read/draft/write/approve/execute, settable caps, exportable usage data, and fixed rate cards.
• Negotiate upfront: confirm seat coverage for agents, independent agent entitlements, third-party agent paths, credit-consuming actions, failed action billing, fixed rates, exportable logs, departmental caps, and seat reductions if agents replace human work.

Context

Seat-based SaaS pricing proxied human work value via logins and clicks, generating predictable revenue; AI agents break this by using systems (read CRM, update tickets, trigger workflows) without seats, forcing vendors to rethink models amid exploding agentic workflows (e.g., one developer used 8 billion tokens monthly). Vendors like Salesforce, Microsoft, ServiceNow erect toll booths on agent actions, while SAP's 2026 API policy restricts unsanctioned AI sequences, prioritizing platform control. Builders risk economics capture without understanding incentives; operators must negotiate before usage embeds, as leverage vanishes post-adoption.

Notable Quotes

• "Agent Force pricing uses flex credits and agentic work units. You update a record you summarize a case you answer an inquiry execute a prompt run a workflow whatever. Each one draws from the same meter."
• "I was talking to a developer just this past week who has used 8 billion tokens in the last month. Eight billion in a month."
• "The seat was always a proxy for human work and value created. And the agent license is becoming a meter for that same unit of value except now that it's been delegated."

Content References

Full checklists, eight-vendor breakdown (Zendesk outcome pricing, HubSpot per-resolution, Workday agent system of record, Atlassian Rovo credits), policy pushback language, operation taxonomy, and cost dashboard appear in linked Substack post.