Recommended Free Tools
The headline refers to Sierra, an enterprise AI-agent company co-founded by Bret Taylor and former Google executive Clay Bavor. Sierra launched on February 13, 2024, while Taylor was chair of OpenAI’s board. Its pitch was to build branded agents that do more than answer customer questions: they can use a company’s systems to try to complete tasks such as tracking orders, managing subscriptions, and processing service requests.
That launch is now part of a larger story. By August 2026, Sierra was marketing agents for voice, chat, email, and WhatsApp, as well as tools for building agents and workflows that can pursue business outcomes over longer periods. Those capabilities and performance figures below are Sierra’s claims unless otherwise attributed.
What Sierra is—and what an AI agent does
Sierra is a platform for companies to build customer-facing AI agents, not a general-purpose chatbot for consumers. Its central distinction is action: an agent is meant to interpret a customer’s goal, use approved tools or business systems, and attempt to complete the work. A conventional chatbot may answer where an order is; an agent could, in principle, look up the order, determine whether it qualifies for an exchange, initiate that exchange, and explain what happens next.
That example describes the intended workflow, not a guarantee that every agent or deployment handles every case correctly. Sierra’s 2024 launch announcement said its agents could answer nuanced questions, make recommendations, manage subscriptions, track packages, and process exchanges while connecting to systems such as customer relationship management and order-management software. The company also said agents could follow business policies and guardrails. Sierra’s launch announcement sets out those original capabilities.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitches#1 Best Overall
How the intended workflow works
- A customer states a goal in ordinary language.
- The agent interprets the request and retrieves relevant customer, transaction, and policy information.
- It uses tools and connected systems that the company has permitted it to access.
- It completes an eligible action, proposes one for approval, or routes the case to a person.
- The company monitors the interaction and reviews whether the result was appropriate.
The exact steps and level of automation depend on each deployment. The distinction between agent and platform matters: an agent handles a task, while an agent platform provides integrations, permissions, testing, monitoring, analytics, and deployment controls for operating those agents at scale.
What Sierra emphasized at launch
Sierra described its agents as sophisticated, authentic, and trustworthy. In the company’s framing, “sophisticated” meant handling requests beyond fixed FAQ flows; “authentic” meant adapting to a company’s voice and customer context; and “trustworthy” meant adding auditing, quality-assurance, data-governance, and access-control features. Sierra also said interactions with systems of record could be deterministic. These were product-design claims from the company, not independent findings that errors or unsafe actions could not occur.
The founders’ broader argument was that conversational agents could become a major customer-interaction layer, much as websites and mobile apps had become important ways for customers to interact with businesses. The test of that argument is whether agents reliably complete useful tasks—not simply whether they can hold a natural-sounding conversation. Accuracy, escalation, privacy, integration effort, and customers’ willingness to use automation all affect whether the idea works in practice.
Rank #2
Who Bret Taylor is, and why his role matters
Taylor’s background spans consumer technology and enterprise software. Sierra’s biography credits him with co-creating Google Maps, serving as Facebook’s chief technology officer, founding Quip, and co-leading Salesforce as co-CEO. He co-founded Sierra with Bavor; Taylor was not the company’s sole founder. Sierra’s company biography describes the founders and their experience.
OpenAI announced Taylor as chair of its board in November 2023, before Sierra’s launch. He remains listed as chair of OpenAI’s Foundation board in the organization’s current structure information. The accurate framing is therefore that Taylor launched Sierra in 2024 while chairing OpenAI’s board—not that he was an OpenAI executive or that Sierra was an OpenAI product. See OpenAI’s board announcement and its current structure page.
Early customers and the limits of the launch figures
Sierra named WeightWatchers, SiriusXM, Sonos, and OluKai among its early customers. In its launch announcement, the company said the WeightWatchers agent handled nearly 70% of customer sessions and received a 4.6-out-of-5 customer-satisfaction score; it said OluKai’s agent handled more than half of customer cases during the holiday surge. Those are company-reported figures, not independently audited results. The announcement does not establish that the same outcomes would apply across other customers or deployments.
When evaluating such figures, buyers should ask about the measurement period, the types of cases included, how escalations were counted, and whether satisfaction scores included customers who never reached a human. Sierra’s current materials also identify deployments or customers including Clear, Casper, Minted, Santander, Rocket Mortgage, and Cigna; those examples are presented by Sierra. The company page and Horizon announcement provide current examples.
Funding and commercial model
VentureBeat reported that Sierra launched with $110 million in initial funding, including investment from Benchmark and Sequoia. That is a reported figure from contemporary coverage, not a financing amount established by Sierra’s launch announcement. VentureBeat’s launch coverage gives the funding context.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Sierra positions its pricing as outcome-based: the commercial idea is to charge in relation to results rather than only seats or tokens. The company does not publish a standard public price in the cited product materials, so buyers need to establish contract terms directly. Outcome-based billing can align payment with completed work, but it raises practical questions: what counts as a billable resolution, how partial completions or repeat contacts are treated, and who pays when a human must finish the case. It may also be harder to forecast than seat-based or usage-based pricing.
Rank #4
How Sierra’s product scope changed after 2024
Sierra’s evolution has been from customer-service agents toward a wider enterprise agent platform. Its updates describe phone-based agents in 2024, no-code and programmatic development under its Agent OS direction in 2025, and Ghostwriter in March 2026—a tool intended to create and optimize agents from natural-language instructions. These milestones are described in Sierra’s company updates.
On July 16, 2026, Sierra announced Horizon, a product direction for agents that can pursue goals over multiple interactions and longer periods. The company’s examples include closing a sale, upgrading a subscription, scheduling a test drive, originating a loan, and arranging healthcare referrals. Sierra’s current product page also describes work across voice, chat, email, and WhatsApp; support for 58 languages; and workflows such as insurance claims, product returns, and mortgage origination. These are current company product claims, not independently verified performance results. See the Horizon announcement and Sierra’s product overview.
Sierra also says it serves hundreds of companies and works with 40% of the Fortune 50. Those figures are company-reported, not independently audited measures of market share. Its positioning includes expert development and implementation support as well as software, which reflects the work often required to connect systems, define policies, test agents, and maintain oversight.
For further company milestones, Sierra reported more than $150 million in annual recurring revenue by its second year, and Taylor’s author page lists a $950 million financing round at a valuation above $15 billion, announced May 4, 2026. These are self-reported company figures, not independent assessments of profitability or long-term product performance. They appear in Sierra’s updates and Taylor’s author page.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What enterprise buyers should test before deployment
An agent that can change accounts or transactions needs more than a convincing demo. A buyer should evaluate the full path from customer request to completed action, including what happens when the system is uncertain or encounters an exception.
- Task completion: Measure whether the agent safely completes refunds, changes, bookings, claims, or other target work—not just whether its answers sound correct.
- Integration and permissions: Check connections to CRM, billing, order-management, identity, telephony, and knowledge systems. Test write permissions as carefully as read access, especially for refunds, cancellations, and account changes.
- Human handoff: Confirm that people can take over complex, emotional, or sensitive cases with the conversation context intact.
- Controls and auditability: Determine whether the company can inspect what the agent did, which data and tools it used, and why it took an action. Ask how tone, claims, offers, and policy decisions are constrained.
- Evaluation and maintenance: Ask about simulations, regression tests, transcript review, incident response, and how outdated knowledge is corrected.
- Data governance: Establish where customer data is stored, whether it is used to train models, how long it is retained, which subprocessors handle it, and what regional controls are available.
- Economics and portability: Define billable outcomes and treatment of escalations or repeat contacts. Ask which underlying models the platform can use and what switching costs may arise from proprietary workflows, memory, analytics, or integrations.
- Channel consistency: A shared agent across web, phone, email, and messaging can preserve context, but a faulty policy or integration can also affect more customer interactions at once.
Common failure modes include confident but incorrect policy answers, actions based on stale information, poor escalation, misunderstood names or numbers in voice interactions, billing disputes over what constitutes success, and metrics that reward short conversations instead of satisfactory outcomes. Customers may also lose trust if it is unclear they are speaking with a machine or if reaching a person is difficult. An “autonomous” workflow still requires permissions, oversight, and ongoing maintenance; the label does not remove those responsibilities.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
