DECAGON

Jesse Zhang and Ashwin Sreenivas were not first-time founders. Jesse had built a gaming startup called Lowkey and sold it to Niantic in 2021.

Ashwin had co-founded a computer vision company called Helia, sold it to Scale, and done a stint as a deployment strategist at Palantir. The two met at a retreat hosted by the venture firm Andreessen Horowitz.

They started Decagon in 2023, right as large language models got good enough to actually talk to customers. Their target was customer support, a giant cost center every company quietly hates.

They came out of stealth in June 2024 with $35 million already in the bank.

Companies pay Decagon to run their customer support with AI agents instead of huge human teams. A customer types a question.

The Decagon agent understands it, checks the company's systems, and actually resolves the issue, not just points at a help article. It handles refunds, account changes, and messy multi-step problems over chat, email, and voice.

Decagon charges enterprises based on usage and resolved tickets. In plain English, brands hand the flood of repetitive questions to AI and keep humans for the hard cases.

The pitch is faster answers for customers and lower costs for the company.

The core product is the Decagon AI agent for customer support. It plugs into a company's help desk, knowledge base, and internal systems, then handles customer conversations from start to finish.

It works across chat, email, and voice. Decagon also built tools that let support managers watch what the AI is doing, set rules, and step in when needed.

A newer focus is what the company calls agent operations, basically the control room for running a whole fleet of AI support agents. The idea is one system that resolves the boring bulk of tickets so humans handle the rest.

Decagon went straight for the enterprise and leaned hard on its investor network. The a16z and Accel connection opened doors to exactly the kind of fast-growing software companies that drown in support tickets.

Founders of Box, Airtable, Okta, and Klaviyo put in angel money, which doubled as a customer pipeline. The product had to be genuinely good, because enterprise buyers do not tolerate an AI that gives wrong answers.

So Decagon obsessed over accuracy and control. That reputation, plus relentless fundraising, let it triple its valuation to $4.5 billion by early 2026.

The AI customer support space is a bloodbath. Sierra, backed by Bret Taylor, is valued at $10 billion.

Intercom, Salesforce, and a dozen startups are all chasing the same budgets. Decagon does not build its own AI models.

It sits on top of others, so its edge is the product and the enterprise trust, not the core technology. That is a real risk if the model makers move downstream.

And the valuation is spicy. $4.5 billion is a huge number for a company only founded in 2023.

Decagon has to keep growing fast or that price looks silly in hindsight.

The investor list reads like an AI-era who's who. Andreessen Horowitz led the seed.

Accel led the Series A. Bain Capital Ventures came in for the Series B.

Then in early 2026 Coatue Management and Index Ventures led a $250 million round at a $4.5 billion valuation. On top of the funds, a heavy lineup of founder angels backed it, including the CEOs of Box, Airtable, and Lattice.

That kind of support is both fuel and pressure. It buys credibility, and it demands the growth to justify the price.