Founded just two years ago by second-time founders, Decagon’s AI agents are used by more than 100 companies to handle customer service tasks. But can the young, hungry team keep up with well-resourced incumbents?
By Rashi Shrivastava
Ona Friday afternoon in late October, Jesse Zhang, the 28-year-old cofounder and CEO of AI customer service startup Decagon, strides through the lush greenery of San Francisco’s Salesforce Park, a sprawling rooftop garden in the shadow of the 1,070-feet Salesforce Tower—named for one of his biggest rivals. But he’s nonchalant in the face of stiff competition from public giants 10 times the size of his tiny upstart. “What’s there to be intimidated about?…We like competing,” he said. “We enjoy winning.”
That’s evident. A few moments ago in his South of Market office, Zhang and cofounder Ashwin Sreenivas’ team of 200 erupted in hoots and cheers after a team lead rang a sales gong, touting a freshly signed deal. Blinds manufacturing giant Hunter Douglas had agreed to use Decagon’s AI software for customer support tasks like ordering replacements for damaged parts and answering questions about rebates and warranties.
That’s nothing compared to how the company commemorated a win last year. “I sat in a chair in the middle of the office and Ashwin and Jesse shaved my head,” recalls Accel partner and early investor Ivan Zhao, who promised to go bald if the company increased its revenue tenfold. Now after hitting yet another revenue milestone, Zhang is taking his employees on a trip to Hawaii in early 2026. “It’s going to be expensive,” he said.
Decagon has cause for celebration. Founded just two years ago, the startup’s AI-powered customer service agents are used by more than 100 companies including Notion, Bilt, Duolingo, Substack and Rippling to answer questions about how a product works, process refunds, order replacements and cancel subscriptions. Last valued at $1.5 billion in June, Decagon has picked up $255 million in funding from prominent VC firms like Andreessen Horowitz, Accel and Bain Capital Ventures. The company had $10 million in annualized revenue in 2024 and has crossed at least $30 million in annualized revenue this year. Forbes estimates Decagon’s 2024 revenue came in at $3 million, and it’s on track for $12 million by the end of 2025. (Decagon spokesperson Emilie Cole said that while Decagon does not disclose financial information, in Q3 2025 the business more than tripled both GAAP revenue and ARR year over year and continues to grow.)
Zhang, honored on this year’s 2026 Forbes 30 Under 30 list in the AI category, and Sreenivas, age 30, each own a roughly 25% stake in the company, per Forbes’ estimates, bringing each cofounders’ net worth to about $370 million. (They declined to comment.)
Customer service is one of the most obvious applications for artificial intelligence. The market is large ($12 billion in 2024, per research firm MarketsAndMarkets) and ripe for automation. The industry faces a massive retention issue, with most call centers seeing a roughly 40% employee churn on average. Humans, for the most part, hate answering mindless questions and dealing with angry customers. Plus, they’re expensive. Businesses are always on the look out for ways to integrate time-and cost-saving AI and find new ways to increase revenue. Decagon and its ilk promise much smarter AI that isn’t frustrating to talk to.
Decagon’s software is built on a mix of AI models from top-tier labs like OpenAI, Anthropic and ElevenLabs. It’s trained on enterprise data: frequently asked questions, help center blogs, manuals and past conversations. Its agents can have natural, human-like conversations with customers using powerful voice AI models and respond to queries and complaints via chat and email. The agents, which have handled 80 million conversations as of late November, also have access to internal databases and tools so that they can automatically pull up information like an order number or account details. To keep the agents on track, Decagon has designed what it calls an “agent operating procedure” — an instruction manual that guides the agent on where to pull information from and how to respond to certain questions, depending on who’s asking. Businesses are billed based on the number of conversations Decagon’s AI agents handle. More complex customer inquiries and voice features come with a higher price tag.
But Decagon is the new kid on a very crowded block. It’s up against Salesforce ($440 million in annualized revenue for its agentic AI segment in the latest quarter), Intercom (CEO Eoghan McCabe claims to have some 7,000 paying customers and five times the revenue of Decagon) and Zendesk (targeting $200 million in revenue this year from the 20,000 customers using its AI products). “The market as a whole is more competitive than it ever has been,” said Jason Maynard, the CTO of Zendesk. McCabe is more blunt: “Every performance bake-off we have with [Decagon], we win head to head,” he said, claiming that he’s also snagged some of Decagon’s customers. Decagon spokesperson Cole said “lightweight solutions” like Intercom can be a better fit for smaller and medium-sized companies. “We’re always happy if and when companies find good fits for their needs,” she said.
But Zhang’s most worrisome threat: $10 billion-valued AI customer service startup Sierra, cofounded by OpenAI chairman Bret Taylor, who’s known for previous stints as the co-CEO of Salesforce, CTO of Facebook and board chair at Twitter. The company, backed by blue chips like Sequoia, Benchmark and Thrive Capital, crossed $100 million in annualized revenue in late November.
“Their competitors have famous people and many connections because of their previous great work that they have done in the industry,” said Aaref Hillay, a partner at Bain Capital Ventures and an investor. “Decagon is the David here in the sense that they don’t have that advantage.” But he thinks the company is playing to its strengths: a better product and more technical expertise.
Zhang acknowledges the disparity. “Compared to the big Sierras, Salesforces of the world, we have a bit of a younger team,” he said.
Still, investors are clamoring to give them money. Decagon is reportedly out raising again, aiming to double its valuation to at least $4 billion. Top VC outfits have showered Zhang and Sreenivas with gifts (most recently a framed art piece of a deconstructed iPhone) and taken them out to basketball games. “The same way that we compete for customers… they’re competing hard to invest in certain companies,” Zhang said.
Customers love them too. Car rental company Hertz’s Decagon chatbot resolves three out of four customer queries on its own without escalating to a human, handling tasks like cancelling or extending reservations. Decagon’s voice agents can have dozens of conversations at the same time, said Vikram Rajagopalan, a VP of customer experience at Hertz. That means no wait times for impatient customers.
Companies often compare the software’s performance against other customer service tools by pitting one chatbot against another. Danielle Doremus, a senior director of customer service at ClassPass said Decagon competed against 11 other vendors and had to respond to a list of 125 questions before it won the contract. Now, ClassPass’ Decagon chatbot can book fitness classes, waive fees for missed ones and respond to billing and refund questions. The chatbot has responded to 3.8 million inquiries to date and resolves 60% of support issues on its own.
The software also allows companies to simulate conversations to test and tweak the agent’s performance before it’s rolled out to the world. And it analyzes conversation transcripts to find patterns and insights that can help businesses understand where the AI agent might be struggling and how it can improve. Companies can build their own “watchtower” agents that scan conversations, and identify places to upsell and make sure the AI isn’t talking about medical emergencies or giving legal advice.
Despite his young age, this isn’t Zhang’s first company. In 2018, after realizing that the courses for his Harvard computer science degree weren’t very relevant to startup building, he finished the program a year early. Then came the brainstorming for company ideas. “I just felt like there was no time to waste,” Zhang said. That year, he founded Lowkey, an app for recording and sharing gaming clips which was acquired by game development company Niantic in 2021 for mid-eight figures.
In May 2023, Zhang was at an Andreessen Horowitz event in Utah when he met Sreenivas. The two bonded over their histories competing in math Olympiads as high schoolers (Zhang in Boulder, Colorado, Sreenivas in Kerala, India) and starting companies (Sreenivas sold his previous computer vision startup Helia to data labelling outfit Scale AI in 2020 for an undisclosed amount). They decided to team up and spent the first few weeks sending hundreds of LinkedIn messages and interviewing potential customers, asking them what AI product they would be willing to pay for. They soon realized that call centers are a major cost-suck for companies, especially those that are scaling quickly. “A lot of our job has been [making] sure that things are run very efficiently so that [our customer] doesn’t have to hire 500 people,” Zhang said.
Now as Decagon continues to grow, Zhang’s strategy to stay competitive is to move at breakneck speeds in an already fast-paced industry. Marc Manara, head of startups at OpenAI who has worked with the startup on accuracy and latency, said Decagon is “lightning fast” at building new features and evaluations, which assess how well an agent is performing particular tasks. “There’s a risk of falling behind and not being at the bleeding edge if you’re not really, really good at this,” Manara said.
Investors describe a “buzz” in the company’s offices with a fully in-person team and a good chunk of employees working six or seven days a week, sometimes late into the night. (Employees are also given a stipend to live close to the office.) One venture capitalist who’s not an investor in Decagon said the young, hungry founders hounded him for customer introductions. “They’re very aggressive in a good way.” They’ll have to be if they want to keep winning.
