Sierra prices the agent by the outcome it resolves

What it does

Sierra, the conversational-AI company founded by Bret Taylor — formerly co-CEO of Salesforce — and Clay Bavor — formerly of Google — has spent the past year putting customer-facing agents into production at enterprise brands, and its platform is the most complete picture yet of where the agent category is heading. The pitch starts with a pricing decision: Sierra charges for outcomes — resolved conversations — rather than seats or raw usage. The argument is that when the vendor is paid only when the customer’s problem is actually resolved, the vendor’s incentives line up with the customer’s, and everything else about the product follows from that alignment. The agent handles the full arc of a customer interaction: answering with grounding in the company’s policies and systems, taking actions — refunds, plan changes, the next step in troubleshooting — where policy allows, and escalating to a human where it does not, all in the brand’s own voice rather than a generic assistant’s. Around the agent sits the Agent Experience platform, the build-test-deploy-improve loop that the company positions as the difference between a demo and a deployment: author the agent’s policies and workflows, evaluate its behavior against real scenarios before it ships, trace every conversation afterward, and tune the boundaries as you learn. The trust architecture is the product: guardrails define what the agent may do, escalation paths define where it must stop, and every action the agent takes is logged and auditable. Sierra’s framing is the one worth carrying away: a customer-facing agent is not a chatbot with better prose — it is a product surface with the same discipline you would bring to any surface that meets customers, tested before release, observable in production, and bounded by explicit policy. WeightWatchers and SiriusXM are among the deployments the company names.