Sierra logo
Customer Service · Voice Assistants

Sierra

Sierra is an enterprise conversational AI platform that builds customer-facing agents for voice, chat, SMS, WhatsApp, email and ChatGPT. Agents connect to systems of record to resolve requests, and billing follows outcome-based pricing.

Active GDPR compliant Usage Based API available 18+ Verified by Guidaio
Overview

What is Sierra?

Sierra is a conversational AI platform that large organisations use to put autonomous agents in front of their customers. It is published by Sierra Technologies, Inc., a Delaware corporation headquartered in San Francisco and co-founded in 2023 by Bret Taylor — formerly co-CEO of Salesforce, CTO of Facebook and co-creator of Google Maps — and Clay Bavor, who spent eighteen years at Google leading Google Labs, the AR/VR effort and Google Lens. The company runs offices in New York, Atlanta, London, Singapore, Tokyo, Paris, Madrid, Toronto and San Francisco.

The platform is organised into named building blocks. Ghostwriter is an agent that builds agents: it ingests standard operating procedures, call transcripts, whiteboard photographs and audio recordings, or simply a goal stated in plain English, and returns a multilingual, multichannel agent with guardrails already in place. Agent Studio covers the same ground without code, while the Agent SDK lets engineering teams write customer journeys as versionable code, with composable skills, explicit goals and a tunable balance between creativity and determinism. Horizon breaks complex outcomes into steps that unfold over days or months, carrying customer memory across conversations and systems. Insights, Explorer and the observability layer expose every tool call, knowledge lookup and latency figure, and support proactive monitors and multivariate experiments.

Underneath sits what Sierra calls a constellation of large language models — frontier, open-weight and proprietary — routed per conversational turn and wrapped in supervisory layers meant to reduce hallucination and block abuse. The company claims 59 languages across text and voice channels, availability around the clock, and partnerships with 40% of the Fortune 50; named customers include Rocket Mortgage, Singtel, SoftBank, SiriusXM, Sutter Health, Redfin, CarMax, SoFi and CLEAR.

Compliance carries much of the pitch: SOC 2, HIPAA, GDPR, PCI DSS Level 1 Service Provider, FedRAMP High, CCPA, CSA STAR, ISO 27001 and ISO 42001. Commercially, Sierra sells through a single form and prices by outcome, so a client pays only when the software achieves a defined, valuable result. No price is published anywhere on the site.

What it does

  • Build a production-ready AI agent by describing a goal in plain English or uploading SOPs, transcripts, whiteboard photos and audio recordings
  • Deploy a single agent across voice, web and in-app chat, SMS, WhatsApp, Apple Business Chat, email and ChatGPT
  • Connect the agent to CRM and order management systems so it completes transactions instead of only answering questions
  • Validate behaviour with simulations across edge-case scenarios and catch regressions before release
  • Analyse conversations with Deep Research-style exploration, proactive monitors and multivariate experiments
  • Assist human care representatives in real time with grounded, ready-to-send responses through Live Assist
  • Collect card and ACH payments over the phone through PCI DSS Level 1 infrastructure, with no IVR handoff
Audience

When to use Sierra / When not to

A quick filter to help you decide if Sierra is the right fit.

When to use Sierra

  • Enterprise customer experience leaders running high-volume support across voice, chat and messaging
  • Engineering teams that want to build and version customer journeys as code through the Agent SDK
  • Regulated organisations needing SOC 2, HIPAA, PCI DSS Level 1, FedRAMP High and ISO 27001 coverage
  • Financial services, insurance and healthcare firms automating transactional requests such as claims, loans and appointments
  • International brands that must serve customers in dozens of languages around the clock

When not to use Sierra

  • Individuals and small teams: there is no self-service sign-up and no published price
  • Buyers who need to compare costs before talking to sales, since Sierra publishes no pricing page
  • Anyone looking for a free plan or a free trial, neither of which the site announces
  • Developers who want to read the documentation before committing: docs.sierra.ai is credential-gated and declared confidential
  • Organisations that require a contractually named hosting country, which Sierra does not publish
Get started

How to use Sierra

A typical end-to-end flow, from setup to results.

  1. Submit the form at the Learn more page with your name, job title, company email, company name and industry; there is no self-service sign-up
  2. Work with Sierra's team to scope use cases and define which outcomes will be measured and billed
  3. Gather source material: standard operating procedures, call transcripts, knowledge base articles, brand guidelines and policies
  4. Build a first agent in Ghostwriter by describing the goal in plain English or uploading that material
  5. Refine it in Agent Studio if your team works without code, or in the Agent SDK if you prefer journeys written as versioned code
  6. Connect the agent to your systems of record — CRM, order management, payment processor — so it can act rather than only answer
  7. Set goals, guardrails and the degree of determinism you want for each workflow
  8. Validate with simulations across edge-case scenarios and check for regressions before release
  9. Deploy across the channels you need, and publish to ChatGPT in one click or through CI/CD
  10. Monitor in production with Explorer, monitors and experiments, then review and ship improvements as the agent evolves
Quick read

Pros & Cons

Pros

  • One agent deployed consistently across every channel, including publication to ChatGPT via Model Context Protocol
  • Two genuine build paths: no-code for business teams, and a versionable SDK for engineers
  • Unusually broad compliance coverage, including PCI DSS Level 1, FedRAMP High, HIPAA, ISO 27001 and ISO 42001
  • Explicit, unconditional commitment that customer data is never used to train models
  • Public trust centre with a maintained and updated subprocessor list
  • Simulation and regression testing before release, with fine-grained observability afterwards
  • Deep, named customer references in regulated sectors rather than anonymous logos

Cons

  • No public pricing whatsoever: the pricing URL returns a 404 and the sitemap holds no pricing page
  • No free plan and no free trial are announced anywhere on the site
  • Technical documentation sits behind credentials and is declared confidential and evaluation-only
  • The supported language count contradicts itself across the site — 59, 58 and over 55 — and is never listed by name
  • No hosting country or region is guaranteed; the policy permits storage anywhere in the world
  • No Article 27 EU representative and no named data protection officer; privacy@sierra.ai is the only published address
  • No support email and no self-service onboarding: everything begins with a sales conversation
Pricing

Pricing & Plans

Sierra publishes no prices. There is no free plan and no free trial announced on the site, and the pricing URL returns a 404. The company charges on an outcome basis, meaning the client pays only when the software achieves a specific, valuable result; what counts as a billable outcome is set contractually and is not disclosed publicly. A quotation is obtained through the sales form on the Learn more page. The only monetary figure anywhere on the site is a one-hundred-dollar liability cap in the website terms, which has no bearing on commercial pricing.

Prices and plans listed above may evolve. Always check the official pricing page before subscribing.
Trust & Privacy

Data, GDPR & hosting

A consolidated view of how Sierra handles your data.

GDPR overview

GDPR compliance is claimed explicitly, alongside SOC 2, HIPAA, PCI, FedRAMP High, CCPA, CSA STAR, ISO 27001 and ISO 42001. Sierra is certified under the EU-U.S. Data Privacy Framework, its UK Extension and the Swiss-U.S. DPF, and cites EU Standard Contractual Clauses as a transfer safeguard. The policy sets out four lawful bases — contract, legitimate interest, legal obligation and consent — and the full set of data subject rights, including portability, erasure, objection, restriction and appeal of a refusal. Complaints go first to privacy@sierra.ai, then to JAMS, then to binding arbitration, under Federal Trade Commission oversight. Two gaps deserve attention: no Article 27 EU representative is designated, and no data protection officer is named. The policy was last updated on 17 June 2026.

Who owns the data?

Sierra draws a sharp line between two categories. Personal information collected through its own website, marketing and recruiting is governed by its privacy policy, under which Sierra acts as controller for EU, UK and Swiss law. Customer Data — everything processed for a client through deployed agents — falls outside that policy entirely. There Sierra acts as a processor, handling data only on the client's documented instructions under a Data Processing Agreement. Individuals seeking access, correction or deletion of Customer Data must approach the client, not Sierra, which merely assists. Sierra states that a client's data is only used as instructed and is never shared with other customers.

Reuse rights

Sierra commits that customer data is never used to train models, a statement repeated across its Agent SDK and product pages. Personally identifiable information shared with an agent is automatically encrypted and masked, and data is used only as the client instructs. Any reuse beyond delivering the service is therefore governed by the client contract and the Data Processing Agreement rather than by a unilateral licence, and the client remains the party that answers to its own end users. Data collected on Sierra's own website is treated quite differently: the company states that it “sells” or “shares” such information with advertising partners within the meaning of the CCPA, enriches it from third-party sources including ZoomInfo and Clay, honours the Global Privacy Control signal, and does not honour Do Not Track.

Data retention & training

Retention summary
Sierra publishes no retention periods. Its policy states that personal information is kept for as long as you use the services, or as long as needed to fulfil the purpose of collection, resolve disputes, mount legal defences, conduct audits, pursue legitimate business interests, enforce agreements and comply with the law. Duration is set case by case against stated criteria: legal requirements, the volume, nature and sensitivity of the data, risk factors, the processing purposes, and whether those purposes could be met another way. Retention of Customer Data handled through deployed agents is not covered here at all; it is governed by the client contract and the Data Processing Agreement. Personal information belonging to a minor is deleted on notification unless a legal obligation requires keeping it. The policy was last updated on 17 June 2026.
Trains on customer data
No
Subprocessors disclosed
Yes
DPA available
Yes
GDPR contact

Hosting summary

Sierra names no hosting country and no hosting region. Its privacy policy states that personal information may be transferred, processed and stored anywhere in the world, expressly including the United States. Where data moves to a country without an adequacy decision, Sierra cites EU Standard Contractual Clauses as a safeguard, and it is certified under the EU-U.S. Data Privacy Framework, the UK Extension and the Swiss-U.S. Data Privacy Framework. No cloud provider is named. The trust centre does disclose subprocessors, and recently added four AI vendors: Baseten, Fireworks AI, Together AI and Modal. The website itself sits behind Cloudflare and resolves to an anycast address in San Francisco, but that reveals nothing about where customer data lives. Organisations with data residency obligations should treat hosting location as a contractual matter to negotiate, not a published fact.

Hosting countries
🇺🇸 United States
Watch-outs

Things to keep in mind

Risks and trade-offs to weigh before adopting Sierra.

  • Outcome-based pricing sounds aligned, but the contract defines what counts as a billable outcome and that definition is not public; model your costs before signing
  • Handing front-line conversations to an agent can erode the institutional knowledge your human teams build by working through difficult cases themselves
  • Sierra states it sells or shares website visitor data with advertising partners under the CCPA; this concerns your marketing footprint rather than your customers' conversations, but it is worth knowing
  • Do Not Track signals are ignored, and only the Global Privacy Control is honoured
  • No hosting country is guaranteed, which can conflict with data residency duties you owe your own customers or regulators
  • The subprocessor list changes, with four AI vendors added recently; re-check it at contract time and arrange notification of future changes
  • Customers can lose any realistic route to a human if escalation is configured too tightly; Sierra supports handoff, but keeping that door open is your decision
Setup

Setup & Integrations

Technical difficulty

Moderate to high, and never self-service. Every deployment starts with a sales conversation and proceeds with a Sierra implementation team; the company quotes a few weeks to go live. Business teams can build agents in Ghostwriter or Agent Studio without writing code, which genuinely lowers the barrier. The real work lies elsewhere: connecting the agent to CRM, order management and payment systems, defining guardrails, and validating behaviour through simulations. Engineering involvement is expected for transactional use cases, though the Agent SDK makes that work version-controlled and testable rather than ad hoc.

Deployment

Web appAPI

Integrations

ChatGPT WhatsApp Apple Business Chat Model Context Protocol

Supported languages

EnglishSpanishFrenchGermanJapanese
Company

Behind Sierra

Company name
Sierra Technologies, Inc.
Founded
21/03/2023
Country of origin
🇺🇸 United States
Headquarters
150 Sutter St. #690 San Francisco, CA 94104
UBO
INFORMATION_NOT_FOUND
UBO country
INFORMATION_NOT_FOUND
Domain registrar country
🇺🇸 United States
Legal contact

Fundraising

Series A: USD 110 million from Sequoia Capital and Benchmark, at a valuation approaching USD 1 billion (2024)
Series B: USD 175 million led by Greenoaks Capital, at a USD 4.5 billion valuation, October 2024
September 2025: USD 350 million led by Greenoaks Capital Partners, at a USD 10 billion valuation
May 2026: USD 950 million led by GV and Tiger Global, with Benchmark, Sequoia and Greenoaks participating, at a USD 15.8 billion post-money valuation
Sierra's own website publishes nothing about its funding; every figure above comes from press coverage and should be verified before reuse

Social

Official links

Resources

All the official URLs gathered for verification and reference.

FAQ

Frequently asked questions

Who publishes Sierra?
Sierra Technologies, Inc., a Delaware corporation headquartered at 150 Sutter St. #690, San Francisco, CA 94104. Sierra Technologies Ltd is its UK subsidiary.
Who founded the company, and when?
Bret Taylor and Clay Bavor. The site's structured data gives 2023 as the founding year; the Delaware corporate register records the incorporation on 21 March 2023.
Which channels can a Sierra agent work on?
Voice, web and in-app chat, SMS, WhatsApp, Apple Business Chat, email, and ChatGPT. A single agent is built once and deployed across all of them.
Is customer data used to train Sierra's models?
No. Sierra states plainly that your data is never used to train models, and that personally identifiable information shared with an agent is automatically encrypted and masked.
How many languages are supported?
The product page claims 59, the trust and reliability page 58, and the voice page over 55. Sierra never lists them by name, so treat the figure as approximate.
What does Sierra cost?
No price is published. Sierra bills on an outcome basis and quotes follow contact through its sales form. There is no free plan and no free trial.
Is there an API?
Yes, the Agent SDK. Its documentation requires credentials and is declared confidential and limited to evaluation, so it cannot be reviewed before engaging.
Is a data processing agreement available?
Yes. Customer Data is governed by client contracts that include a Data Processing Agreement, which the privacy policy references explicitly.
Where is data hosted?
Sierra names no country or region. Its policy permits storage anywhere in the world, expressly including the United States, with EU Standard Contractual Clauses and Data Privacy Framework certification as safeguards.
How do I report a security vulnerability?
Through Sierra's responsible disclosure programme on HackerOne. Reports are acknowledged within five business days, and critical issues are targeted for resolution within ten.
Conclusion

Should you pick Sierra?

Sierra occupies the enterprise end of the conversational AI market, and its credibility rests on execution rather than promise: named customers in mortgage lending, telecoms, healthcare and retail, a compliance sheet stretching from HIPAA to FedRAMP High, and a founding team with an unusual track record. The product is coherent. One agent is built once — by a business team in Ghostwriter or an engineer in the Agent SDK — then deployed across voice, chat, messaging, email and ChatGPT, wired into the systems of record that let it act rather than merely reply, tested by simulation before release and watched closely afterwards.

Two commitments stand out: customer data is never used to train models, stated without qualification, and billing follows outcomes, so cost tracks results rather than seats.

The reservations are equally clear, and they all concern opacity. Sierra publishes no price at all; the pricing URL returns an error and no plan is named. There is no free tier and no trial. The documentation is locked behind credentials and declared confidential, so a technical buyer cannot assess the platform before engaging. No hosting country is guaranteed. No Article 27 representative serves EU users, and a single address covers legal and privacy matters alike. Even the language count contradicts itself across three pages.

None of this makes Sierra a weak product; it makes it one you cannot evaluate from the outside. This is a platform bought through procurement, with a scoping phase, a contract defining what a billable outcome is, and a deployment measured in weeks. If your organisation has that process and that volume, Sierra is a serious candidate. If you hoped to try it this afternoon, it is not built for you.