Jesse Zhang grew up in Boulder, Colorado.
He studied computer science at Harvard. He interned at Google, Citadel, and Intel. He founded a gaming clip-sharing app called Lowkey, built it quietly, and sold it to Niantic, the company behind Pokémon Go, in 2021.
By 2023 he was 27 years old, had one exit behind him, and was looking at a problem that every large company in the world had in common.
Customer service was broken.
Not broken in the way that produces a bad review on a slow Tuesday. Broken structurally. The ticket volumes were growing faster than the teams hired to handle them. The response times were lengthening while customer expectations were moving in the opposite direction. Companies were spending extraordinary sums outsourcing to large business process outsourcing firms, organizations employing thousands of people in call centers across multiple time zones, and still ending up with customer experiences that nobody was proud of and nobody could easily fix.
Zhang had watched this problem up close during his first company. He had seen what happened to support teams when user numbers grew faster than headcount could follow.
He decided to solve it.
In August 2023, he co-founded Decagon with Ashwin Sreenivas, a fellow second-time founder whose previous AI startup had also been acquired.
They did not announce themselves.
They went to work.
THE PROBLEM WITH CHATBOTS
To understand what Decagon built, you first have to understand what everyone else had already tried and why it had not worked.
The AI customer service space was not empty when Zhang arrived. It was crowded with chatbots, decision trees, scripted response platforms, and automated ticketing systems that had been sold to enterprises on the promise of reducing support costs and improving response times.
Most of them had produced something considerably less appealing than the promise suggested.
The chatbots were rigid. They handled the questions they had been programmed to handle and failed visibly on everything else. Customers who encountered them learned quickly that the path to an actual resolution ran through the bot and toward a human, which meant the bot was not reducing support load so much as delaying it. Enterprise customers who evaluated these tools in head-to-head comparisons against their existing human teams found the gap in quality significant enough that deploying the bot at scale felt like a reputational risk rather than an operational improvement.
The fundamental problem was not the technology of the response.
It was the architecture of the agent.
Existing tools were built to answer questions. Decagon was built to resolve problems. The distinction sounds semantic. It is not. Answering a question requires understanding the question and producing a relevant response. Resolving a problem requires understanding the customer's actual situation, accessing the company's systems and data, making judgment calls about what the right outcome is, executing actions across multiple platforms, and doing all of this in a way that leaves the customer feeling that the interaction was with a company that understood them rather than a system that processed them.
Zhang's core belief was that customer service should feel human even when it runs on AI.
That belief shaped every architectural decision the company made from day one.
THE FIRST EIGHTEEN MONTHS
What Zhang did in the first year and a half of building Decagon is worth examining closely, because it is the part of the valuation headline that tends to obscure.
He focused entirely on execution and customer deployment.
Not the long-term product vision. Not the competitive moat. Not the platform strategy or the partnership ecosystem or the enterprise sales motion that would eventually need to exist at scale.
Zhang focused Decagon purely on short-term execution and customer deployment for the first 18 months, avoiding distractions about long-term product vision. This speed-first approach helped them scale from stealth to unicorn status in a year.
He hired senior generalists from his personal network. People he had already worked with, whose judgment he trusted, whose capabilities he did not need to evaluate from a cold start. The early team was small deliberately. Not because the company could not raise money for headcount, but because Zhang understood that the quality of the first hires sets the trajectory for every hire that follows.
The first hires set the trajectory for all future hiring. Jesse prioritized senior generalists from his network in the early days, which created a strong foundation that attracted subsequent talent as they grew from 12 to 200 in 18 months.
Twelve people to two hundred in eighteen months.
That is not gradual scaling.
That is a company that built the architecture correctly the first time and then expanded it rapidly because the foundation held.
THE BAKE-OFFS
Here is the part of the story that the Silicon Valley narrative tends to skip, because it is unglamorous and it is where most companies that promise to disrupt enterprise software quietly lose.
Decagon had to win sales.
Not in the comfortable sense of a compelling demo and a sympathetic early adopter. In the genuinely competitive sense of being evaluated against entrenched competitors with larger teams, longer track records, established relationships inside the enterprise procurement chain, and the considerable gravitational pull of brand names that had been in the customer service software business before Zhang had graduated from high school.
Almost all of Decagon's clients conducted what's called a bake-off between its software and customer support tools from competitors like Salesforce, pitting one chatbot against another.
Decagon won the bake-offs.
Repeatedly, consistently, against companies ten times its size and with a fraction of its resources.
The reason it won is the reason Zhang was not intimidated by standing in the shadow of the Salesforce Tower during his Forbes interview.
He was nonchalant in the face of stiff competition from public giants ten times the size of his tiny upstart. "What's there to be intimidated about? We like competing," he said. "We enjoy winning."
This is not bravado.
It is the specific confidence of a founder who understands that enterprise customers do not buy the brand. They buy the outcome. And if your outcome is demonstrably better in a controlled evaluation, the size of your competitor's marketing budget is not actually the variable that determines who wins the deal.
THE NUMBERS
Founded just two years ago, Decagon's AI-powered customer service agents are used by more than 100 companies including Notion, Bilt, Duolingo, Substack and Rippling.
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 in 2025.
And by April 2026, customers like Avis Budget Group, Duolingo and Fanatics had helped it reach a $4.5 billion valuation.
From zero to $4.5 billion in valuation in less than three years.
With a team of 200 people.
To put that in context, the companies Decagon is displacing built their customer service operations with thousands of employees, decades of institutional knowledge, and procurement relationships that were supposed to be impossible for a two-year-old company to penetrate.
THE SIGNAL FOR YOUR INDUSTRY
The family office principal, the fund manager, the corporate leader, and the private aviation client reading this newsletter do not need to understand the technical architecture of an AI customer service agent.
They need to understand what Decagon represents as a signal.
Every large organization running a customer-facing operation at scale has a version of the problem Zhang identified in 2023. Growing ticket volumes. Rising costs. Quality that does not match the expectation of the customer relationship the company is trying to maintain.
The solution that most organizations have deployed is more people.
Decagon is evidence that the solution is no longer more people.
In just a year since coming out of stealth, Decagon became a $1.5 billion AI unicorn, transforming customer service for companies with AI agents that achieve 70% to 80% deflection rates and 3x customer satisfaction improvements.
A 70% to 80% deflection rate means that 70% to 80% of customer interactions that would previously have required a human agent were resolved entirely by the AI without escalation. And customer satisfaction improved threefold.
This is not automation that produces a cheaper but inferior experience.
This is automation that produces a cheaper and superior experience.
When that combination exists in any industry, the industry reorganizes around it. Not immediately. Not without resistance. But with an inevitability that makes the question not whether the shift will happen but how quickly and to whom the advantage will flow.
THE MESSY PART
Zhang is not operating in a vacuum.
At a recent event in San Francisco sponsored by McKinsey, Zhang predicted that in three years, thanks to systems like his, there won't be a need for companies to have human workers offering entry-level customer support.
That prediction is not a sales pitch.
It is a statement about what Decagon's technology is already capable of doing, extrapolated eighteen months into a future that the technology's own improvement curve makes reasonable.
It also means that the jobs Decagon is replacing are real jobs. Entry-level customer service positions that, in markets like Kenya, the Philippines, India, and South Africa, represent primary employment for hundreds of thousands of people who entered the BPO industry because it offered stable wages, structured hours, and a path into the formal economy.
The technology does not have an opinion about this.
The market does not have an opinion about this.
The question of what happens to those workers, and who bears responsibility for the transition, is one that Zhang's investors, his enterprise clients, and the governments of the countries where BPO employment is concentrated will need to answer.
Not in the abstract future.
Within the three-year window Zhang himself described.
THE BIBLICAL MIRROR
The book of Nehemiah tells the story of a man who looked at a broken wall and decided he was the one to rebuild it.
Not a general. Not a king. A cupbearer. A person whose job was to taste the king's wine and ensure it was not poisoned, which is a useful job but not the job description of someone who typically ends up rebuilding the walls of a city.
Nehemiah heard about the state of Jerusalem. The walls were broken down and the gates had been burned. The people living there were in disgrace. Everyone who had assessed the situation had concluded, reasonably, that rebuilding required resources, authority, and institutional support that simply were not available.
Nehemiah asked the king for permission and provisions.
Then he went and inspected the walls at night, alone, before telling anyone what he was planning.
Then he gathered the people and told them what needed to happen.
Then they built the wall in fifty-two days.
The same people who had lived beside a broken wall for years, who had normalized the disgrace of it, who had concluded that restoration was beyond what they could achieve, built the entire wall in fifty-two days once someone arrived with a clear enough vision and the organizational will to move.
Jesse Zhang looked at enterprise customer service and saw a broken wall.
Every large company had normalized the cost, the inefficiency, the gap between the experience they were delivering and the experience their customers deserved.
He did not ask whether the incumbents had the resources and the institutional support to fix it.
He built the tool that fixed it.
In two years.
The wall that everyone else had decided was too broken to rebuild is being rebuilt.
The only question remaining is who is doing the rebuilding in your industry.