What it is
Sierra is an agentic customer-experience platform — a layer of AI agents that hold full customer conversations across chat and voice and then execute the backend action the conversation is about: processing a refund, saving a cancellation, updating a subscription, taking a payment. Founded in 2023 by former Salesforce co-CEO and OpenAI board chair Bret Taylor and ex-Google executive Clay Bavor, it launched commercially in February 2024. By May 2026 it had crossed ~$200M ARR — up from ~$26M at the end of 2024 — and raised a $950M Series E led by Tiger Global and Google’s GV at a $15.8B post-money valuation. Sierra reports working with around 40% of the Fortune 50, including WeightWatchers, SiriusXM, Sonos, Nordstrom, and ADT.
Why it shows up in CX and support stacks
- The agent does the work, not just the talking. Sierra agents connect to your CRM, billing, and order systems and complete the transaction inside the conversation. The platform is Level 1 PCI-compliant, so an agent can collect a payment without handing the customer to a human. That is the line between a deflection chatbot and an agent that resolves.
- Voice is now the main channel. Sierra’s voice agents have passed text as the primary surface — which matters if your contact-center volume is phone-first, where most AI-support tools are still chat-only.
- You build agents without an engineering queue. Agent Studio plus its “Ghostwriter” tool generate a working agent from existing documents, transcripts, or call audio, so a CX team configures behavior instead of filing tickets to dev. The Agent Data Platform keeps memory across interactions so a returning customer is not starting cold.
Pricing reality
Sierra is sales-led with no public price list, and the model is the pitch: outcome-based billing. You pay when an agent resolves an interaction, saves a cancellation, or completes an upsell — not per seat, not per message. In practice annual contracts land in the $150K–$350K+ range, so this is an enterprise line item, not a self-serve tool. The upside is that cost tracks resolved outcomes rather than headcount; the catch is that “resolution” is defined in the contract, so what counts as a billable success — and what does not — is the single most important thing to negotiate before you sign.
Best for
Enterprise CX and support leaders running high volume across chat and voice who want agents that complete transactions end to end — refunds, billing changes, retention saves — inside regulated, PCI-scope workflows. It is the right call specifically when your bottleneck is resolution rate and containment on complex, system-of-record-backed requests, not when you need a cheap FAQ deflector.
Skip it if your support volume is low enough that a $150K+ floor cannot pay back, if your needs stop at answering questions from a help center (a lighter tool wins on cost), or if you want transparent per-seat pricing you can model without a sales cycle.
Versus the alternatives
The queued head-to-head is Decagon — the other 2023-founded pure-play AI-support-agent vendor and the fastest-growing entrant in the segment; pick Decagon when your business is internet-native and ticket-deflection-led, and Sierra when you need voice plus transaction execution in regulated enterprise workflows. The incumbents are the platforms you already pay for: Salesforce Agentforce, where you would consolidate agents inside the CRM of record, and Zendesk or Intercom Fin, where the AI is bolted onto an existing seat-priced help desk — choose those when staying on one vendor beats adding Sierra’s high-touch deployment on top.
Watch-outs
- “Resolution” is a negotiated definition, and it is your bill. Outcome-based pricing only protects you if the success criteria match real customer outcomes. Guard: get the resolution, escalation, and dispute definitions in writing, instrument them against your own CSAT and reopen rates for a full quarter, and audit a sample of “resolved” conversations before you trust the meter.
- This is a high-touch enterprise deployment, not a plug-in. Connecting Sierra to billing, CRM, and PCI-scope payment flows is a project with security review, not a weekend install. Guard: scope a single high-volume journey first, set a containment and accuracy baseline against your current tooling, and expand only after the agent beats it on a contained pilot.
- Agents acting on payments and subscriptions are a new failure surface. An agent that can issue a refund or change a plan can also do so wrongly at scale. Guard: keep hard limits and human review on irreversible actions — refunds above a threshold, cancellations — and log every agent-executed transaction for reconciliation from day one.