The Agent OS architecture
Sierra’s $15.8B valuation follows a $950M Series E in May 2026. Bret Taylor, the former Salesforce co-CEO and current OpenAI board chair, and Clay Bavor, the former Google Labs leader, founded the company in 2023. Taylor previously co-created Google Maps and served as CTO of Facebook. Bavor spent 18 years at Google, where he most recently led Google Labs. The platform centers on Agent OS 2.0, which provides reasoning, memory, channels, and supervision in one place. The Agent SDK allows developers to define goals and guardrails through declarative programming. For example, a company can ensure orders only return within 30 days of purchase. The SDK supports composable skills, tuning controls for how creative versus deterministic the agent behaves, and CI/CD with GitHub Actions. The platform also includes Agent Studio for low-code workflow design and the Agent Data Platform for cross-journey context. Agent Studio includes "Journeys" for designing workflows in natural language and "Workspaces" for team collaboration. Agent Memory provides persistent context across conversations so an agent carries information from one chat to the next. Supervisory agents monitor the primary agent in real time to ensure factual accuracy and compliance. These supervisors can take subtle corrective actions, such as steering an agent away from mentioning a competitor. Sierra also uses a constellation of models to balance accuracy, latency, and tone, and its Agent Data Platform provides a memory layer for customer context.
Deployment across channels and brands
The platform deploys branded agents across chat, voice, email, SMS, messaging, and ChatGPT. The voice agent replaces rigid IVR menus with natural conversation that handles interruptions and corrections. It parses inputs like order numbers and email addresses via voice and manages payments through PCI-certified infrastructure. The system adjusts to sentiment and tone mid-call. While Nordstrom launched its Nora voice agent in five weeks, Singtel achieved a 70% resolution rate after a ten-week deployment. Cigna reduced patient authentication time by 80% within eight weeks of implementation. Companies like WeightWatchers use these agents to handle nearly 70% of customer sessions. They achieve a 4.6 out of 5 satisfaction score and maintain a consistent brand voice across various digital touchpoints. The company’s customer list includes brands like Wayfair, Gap Inc., ASOS, Rivian, Rocket Mortgage, Nubank, and Chime, which all use the platform for high-volume consumer interactions across various digital channels and multi-channel messaging services. Sierra’s customer base includes more than 40% of the Fortune 50. You likely already know that enterprise deployments require significant technical lift. These projects often involve months of planning.
| Feature | Specification |
|---|---|
| Valuation (May 2026) | $15.8B |
| Voice Languages | 55+ |
| Deployment Window | 4 to 10 weeks |
| Primary Pricing | Outcome-based |
Costs and implementation friction
Sierra is a premium, enterprise-grade, build-it-with-us platform that uses outcome-based pricing where companies pay only when an agent achieves a result like a resolved conversation, a saved cancellation, or an upsell. For lower-value interactions like routing, the company uses blended pricing that mixes outcome and consumption models. Third-party estimates place annual contracts near $150,000, plus a setup fee between $50,000 and $200,000. This high entry point makes the platform unsuitable for businesses that need to go live this month or want to edit the agent themselves on a Tuesday afternoon. Because Sierra engineers tune the agents, some changes wait on Sierra’s internal teams. This lack of day-to-day editing freedom is a significant drawback for teams that want autonomy. Furthermore, the architecture often routes complex failures to humans without the context of what the AI already tried. This leaves human agents to handle frustrated customers without any prior explanation, which contributes to high burnout rates. SQM Group reports that 88% of contact center professionals cite burnout as one of the biggest industry challenges. Integration friction also persists because legacy telephony and proprietary billing systems often lack clean API surfaces. Successful migrations require mapping existing workflows and performing shadow testing to protect KPIs. The definition of a successful resolution is negotiated per contract, not decided by the user in the moment. Can a company truly achieve full autonomy while still requiring heavy engineering support for every change?
