Zendesk and Sierra argue AI agents should be billed per result, not per seat

# What they said At HumanX in Amsterdam, Zendesk CEO Tom Eggemeier and Sierra co‑founder Clay Bavor made the case that per‑seat software pricing is declining and that customers should pay for AI agents only when those agents complete a measurable job. Eggemeier's headline line: "Unequivocally: seats are dead."

# How outcome pricing works in practice

Sierra already charges only when an agent resolves an issue or makes a sale. Bavor gave a concrete example: if resolving a case via an AI agent costs a company €1 but the same case costs €10 in a traditional call centre, the customer saves €9 each time. He also said Sierra originates more than $1 billion of new mortgages per month for customers such as Rocket Mortgage, serves almost half of the Fortune 50, and works with one in three of the world's largest banks.

OpenAI has introduced outcome‑based pricing for some customers as well, indicating a wider industry shift toward paying for results rather than access.

# What counts as an outcome Zendesk defines outcomes jointly with each customer. That definition will vary by customer and use case. Examples Eggemeier gave include a problem solved without human intervention or an AI agent closing a specified revenue gap on a sale for a marketplace customer.

# Business and operational implications Eggemeier set out three tests companies will face as AI agents change workflows:

• Treat AI as both a threat and an opportunity. He expects the timing for this shift to be three to five years, significantly faster than the earlier shift to the cloud.

• Send people into the field to guide customers through the change.

# Volume and agent interactions Zendesk expects AI agents to handle more than 80% of interactions within about three years, with humans covering the remainder. Eggemeier also predicted rapid growth in agent‑to‑agent interactions: he expects these to pass 50% within two years.

Bavor said personal consumer agents (such as Meta's Muse) will soon communicate directly with business agents, a capability he put on an approximate six‑month timeline.

# Where companies should watch risk

Eggemeier also described geopolitics of talent and operations: two of Zendesk's three main AI hubs are in Berlin and Lisbon, the gap between San Francisco and Europe has narrowed to weeks, and he worried about lagging regions within the US such as Ohio and Kentucky.

# Takeaway for buyers and vendors