The plumbing for agentic commerce is settling into place. Following the Chain of Trust demonstration and Stripe’s claim that activity went vertical in September, the checkout map has matured. The pipes exist to move value between autonomous entities. Yet the monetization map — how vendors extract value when software agents replace human seats — remains in a state of flux.
This realignment affects the $234 billion in enterprise software spend that Gartner identified on July 1, 2026, as exposed to “agentic arbitrage” by 2030. As George Brocklehurst, the firm’s managing vice president, noted: “Agentic AI changes the economics of software. Agentic systems deliver outcomes directly, bypassing traditional user experience-heavy applications and making the software invisible. This breaks the link between user growth and revenue growth for many enterprise software vendors.”
Currently, four competing architectures are vying for dominance.
The per-action credit model is exemplified by Salesforce Agentforce, which charges $0.10 per action and $0.15 per voice interaction via Flex Credits at $500 per 100,000. The per-conversation or per-resolution model remains common: HubSpot charges $0.50 per conversation, Zendesk targets $1.50 to $2.00 per automated resolution. The industry is chasing outcome-based pricing, where startups like Ema are raising capital on the promise of charging for results rather than activity. And there is the hybrid seat-plus-consumption model, gaining ground as Monday.com demonstrated when it shifted from per-seat to AI credits in May 2026.
Salesforce currently offers all four of these models simultaneously. That is not confidence. It is hedging. By maintaining legacy conversation models alongside new credit-based systems and traditional seat pricing, the company is avoiding a definitive bet on how agentic value will be measured. This indecision carries a cost. The Meta Business Agent, priced at $2.00 per million tokens, holds a 40-to-50x price advantage over the legacy per-conversation models still lingering in the enterprise stack.
The ambition to move beyond these granular metrics is best captured by SAP CEO Christian Klein. During the company’s Q2 2026 earnings call on July 23, Klein argued that AI provides a unique chance to completely reset the price level through outcome-based value. “There is nothing you can relate to when you come from the system of record pricing,” Klein said. “And that is, in my eyes, a unique chance also for SAP to do a reset and really price outcome value based.” That remains aspirational. SAP is moving toward AI units and agent-based pricing, but true outcome-based commercial availability is not yet generally available. The industry is stuck between the desire for value-based pricing and the reality of needing a measurable unit of work.
This transition carries structural risk. Analyst-estimated data suggests AI company gross margins sit at 50-60% compared to 80-90% in traditional SaaS. When an AI agent resolves a service ticket for $0.62 compared to $7.40 for a human — as reported in 2026 McKinsey service operations data and cited in our prior coverage — the vendor loses the ability to charge for the human seat that previously managed that ticket. To survive, vendors must capture a portion of that efficiency gain. They have not yet agreed on the mechanism.
The market is testing whether it will settle on the raw efficiency of token-based pricing or the perceived value of outcome-based contracts. As we move further from 2024-era models, the question is which of these architectures can sustain enterprise margins when the software itself becomes invisible. The monetization map is still being drawn in real-time. The only certainty is that the per-seat era is ending.