It is 2:47 a.m. and a woman in Tulsa is trying to refinance her mortgage. She does not know it, but the voice on the other end of the phone is a Sierra agent. It does not put her on hold. It does not ask for her account number twice. It just finishes the job.
The largest AI company most customers will never meet
Sierra is, by design, invisible. The company does not run a consumer app. It does not sell a chatbot in an app store. What it sells is the conversation itself - the one that used to live inside a call center in Manila or a chatbot widget that never quite worked. Sierra ships AI agents that handle phone calls, returns, insurance claims, and account changes for some of the largest companies on the planet.
The roster is the giveaway. Prudential. Cigna. Blue Cross Blue Shield. Rocket Mortgage. CLEAR. Casper. Minted. One in three of the world's largest banks. More than 40% of the Fortune 50. If you have called a customer service line in the last six months and had a surprisingly good time, there is a real chance Sierra was on the line.
The call center never worked
For 40 years, customer service has been the unloved cost center of every public company. It is the line item that grows when the business grows and the department that gets cut first when the business doesn't. Outsourced. Scripted. Chatbotted. Rebranded as "customer success." Nothing fixed it. Customers still hated calling. Companies still hated paying.
Sierra's founders saw the same chart everyone else saw - call volume up, satisfaction down, headcount unsustainable - and asked a different question. Not "how do we make the call shorter?" but "what if a customer could actually finish the thing they called about?" Most companies had quietly given up on that being possible.
A résumé that buys the room
Bret Taylor co-created Google Maps. He was Facebook's CTO. He co-CEO'd Salesforce. He chairs OpenAI's board. The standing joke in Silicon Valley is that Bret has already worked at the four companies your kids are going to interview at. Clay Bavor spent 18 years at Google running Labs, AR/VR, and Workspace. When the two left in 2023 to start something together, the question was not whether they would raise money. The question was what they were going to do with the room.
The bet was specific and a little contrarian. Most generative AI startups in 2023 were chasing the consumer - new chat interfaces, new search engines, new toys. Sierra walked in the opposite direction, toward the part of the enterprise software stack everyone else found boring. The thesis: customer service was the first place where AI would be allowed to do real work, because the alternative - an unanswered phone or a 47-minute hold - was already worse than anything a model could plausibly do.
The customer wall
- CLEAR
- Casper
- Minted
- Prudential
- Cigna
- Blue Cross Blue Shield
- Rocket Mortgage
- 1 in 3 of the world's largest banks
An agent platform, not a magic box
Sierra ships software, not promises. The platform has grown into a small suite, and each piece is aimed at a specific person inside a customer's org.
Agent Studio
A visual builder for the people who used to write IVR call trees, except this time the tree actually answers back.
Agent SDK
For engineers who want to wire agents into custom tools and internal APIs without filing a ticket with procurement.
Voice
Real-time voice agents. Sub-second latency. Will not say "I'm sorry, I didn't catch that" for the 12th time.
Live Assist
A co-pilot for human reps. Quietly drafts the next sentence so the human can think about the customer instead.
Insights
The analytics layer. What customers actually asked. Where they gave up. Who solved it. Useful to managers, occasionally to legal.
Ghostwriter
Launched April 2026. You describe an agent in English. It builds the agent. Yes, an agent that ships agents.
The two-year sprint
The numbers stopped being arguable
For most of 2024, the polite skepticism around enterprise AI was that the demos were impressive and the deployments were not. Sierra's response was a revenue chart. ARR went from a rounding error to $150M in roughly 18 months. Eight months between a $10B round and a $15.8B round. The growth was not the kind a company manufactures for a pitch deck - it was the kind that shows up because customers keep signing more seats.
Sierra ARR · the line that closed the round
↳ The kind of slope investors quote without rounding.
Making AI sound a little more like a person who cares
Sierra writes its mission almost embarrassingly plainly: help businesses build better, more human customer experiences with AI. The wording is unfashionable. Most AI companies in 2026 are writing manifestos. Sierra is writing service-level agreements.
The interesting part is the second word: human. The pitch is not that AI will replace the agent. The pitch is that AI will replace the part of the agent's job that everyone hates - the password reset, the address change, the third call about the same package. Humans get the rest. It is a stance that is almost impossible to disagree with, which may be why the Fortune 50 keeps signing up for it.
The category Sierra is quietly inventing
If Sierra is right, "AI agent" is not a feature. It is a tier of the enterprise software stack, the way CRM became a tier in the 2000s and data warehouses became one in the 2010s. Salesforce - where Bret Taylor was co-CEO - did not invent customer relationships. It just gave companies a single place to put them. Sierra is making the same bet on conversations.
The competitive picture is getting crowded. Every model lab is shipping agents. Every CRM vendor is bolting on agent features. What Sierra has that the rest do not yet, by every public metric, is the customer list and the revenue to back it up. The Ghostwriter launch in April hinted at the next move: a platform where the agent's job is to build the next agent. If that works at scale, Sierra stops being a product company and becomes infrastructure.
It is 2:47 a.m. and the woman in Tulsa has finished refinancing her mortgage. She is back in bed in eleven minutes. She will not write a review. She will not tell her friends. She will just notice, the next time she calls a bank, that this one was different. That is the metric Sierra is being valued on. ↳ quiet win