"Internet business should feel personal."
The vertical wordmark, photographed mid-pivot - a logo that outlived three product eras and one identity crisis.
It is a Tuesday at any one of eight thousand companies, and a customer types a question into a chat box. A few seconds later they have an answer. Not a ticket number. Not a "we'll get back to you within 24-48 hours." An answer. The thing that handled it was Fin, Intercom's AI agent, and on a busy week it does this more than a million times.
That is Intercom in 2026: a customer service company whose flagship product is a piece of software that talks to your customers so your humans don't have to. In May 2026 it went a step further and renamed the entire 15-year-old company after that product. The brand everyone knew became the footnote; the AI became the headline.
For two decades, "customer support" meant a help desk, a queue, and the polite fiction that someone cared. Companies measured it by how cheaply they could make it disappear. Customers measured it by how long they sat on hold. Everybody lost, slowly, in a way nobody had the energy to fix.
Intercom's founders had a less fashionable opinion: that talking to your customers was not overhead, but the point. They started in 2011 with a simple, slightly stubborn belief - that internet business should feel personal - and built a messenger to prove it. The trouble with personal, of course, is that it does not scale. A human can hold maybe a few good conversations at once. A growing company has millions.
That contradiction sat unresolved for over a decade. The messenger was lovely. It still didn't make the math work. Then the math changed.
Four Irish engineers - Eoghan McCabe, Des Traynor, Ciaran Lee and David Barrett - founded Intercom in San Francisco after selling their bug-tracking tool, Exceptional, to Rackspace. They had run a design consultancy called Contrast and carried the design-led, opinion-heavy culture into the new company.
The plot then took a turn worthy of a screenwriter who'd been told to dial it back. McCabe stepped down as CEO in 2020 for health reasons. He was reappointed in October 2022 and found growth had stalled. One month later, ChatGPT launched. Within six weeks, Intercom had a working prototype of what became Fin.
The bet was not subtle. Rather than treating large language models as a feature to tack on, Intercom pointed the whole company at them. That is a risky thing to do to a profitable business. It is a much riskier thing to do to a brand. They did both.
What made the bet credible was that Intercom had been arguing for years that support should be conversational rather than transactional. The team had the data, the help-center content and the conversation history to feed a model - and, just as importantly, the conviction that this was where the company belonged. When the technology finally caught up to the opinion, they didn't hesitate. To fund it, they took on $250M in debt financing in March 2026 and planned to hire hundreds more people, which is a strange thing to do while the rest of the industry was cutting headcount around AI.
Four Irish founders launch Intercom in San Francisco, funded by the sale of Exceptional to Rackspace.
Series B and C rounds (Bessemer, Index) fuel the rise of the in-app business messenger category.
$125M Series D at a $1.275B valuation, led by Kleiner Perkins and Google Ventures.
McCabe steps down as CEO for health reasons; Karen Peacock takes over.
McCabe reappointed CEO - one month before ChatGPT launches.
Fin, the AI agent, ships - built within six weeks of ChatGPT's debut.
$250M debt financing led by Hercules Capital to fund AI and global hiring.
Intercom renames its parent company to Fin. The AI agent is now the business.
Fin is an AI agent for customer service. It reads a company's help center, past conversations and documentation, then answers customer questions across chat and email - and, crucially, it is priced per resolution rather than per seat. Companies pay when it actually solves something, which is a refreshingly honest way to sell software.
Underneath sits the Fin AI Engine, a patented architecture tuned for the messy, high-stakes reality of support: the answers have to be accurate, fast and not made up. Around it, Intercom kept its helpdesk - shared inbox, tickets, Messenger, help center - now repositioned as the desk built for the AI agent era.
Then came the part that sounds like a joke until you run a support org at scale: Fin Operator, an AI whose only job is to configure, monitor and improve Fin. An AI agent managing another AI agent. Someone, somewhere, has to keep the robots honest, and Intercom's answer was to hand that job to a robot too. Cynics called it turtles all the way down. The teams drowning in configuration work called it a relief.
The deeper move is the pricing. By charging per resolution rather than per seat, Intercom tied its own revenue to whether the product actually works. If Fin fumbles, the customer doesn't pay - which is an unusually direct incentive to make the engine good, and a quiet rebuke to every support tool that charged you whether or not it ever solved anything.
The AI agent that resolves customer conversations automatically, billed per resolution.
Shared inbox, tickets, Messenger and help center - the human layer behind Fin.
An AI that tunes and watches Fin so support teams don't have to babysit it.
The chat widget that started it all - real-time conversations on web and in-app.
Four products, one obsession: getting a customer an answer before they've finished being annoyed.
The skeptic's question is fair: does any of this work, or is it a very expensive way to frustrate people faster? The customer roster is the first answer. Anthropic - a company that could obviously build its own - chose to buy Fin instead, and used it to resolve tens of thousands of queries at a reported 50.8% resolution rate. DoorDash and Mercury run on it too.
Synthesia's support team saw Fin resolve over 6,000 conversations and save more than 1,300 hours within six months, with self-serve rates as high as 87%. Lightspeed has reported resolution rates up to 65% while keeping satisfaction stable. The pattern holds: fewer humans answering the same questions, customers waiting less.
Stat strip, lightly editorialized: every one of these numbers was a rounding error three years ago.
Remember the arithmetic - personal or affordable, pick one. Intercom's mission is to refuse the choice. The stated goal is a Customer Agent that can handle the entire customer experience: instant, accurate, available at 3am, and cheap enough that no company has to ration it.
It is worth being clear-eyed here. AI agents get things wrong, resolution rates are not 100%, and "the robot handled it" is not always a compliment from the person on the other end. Intercom's whole pitch rests on the engine being accurate enough that the trade is worth it. The numbers suggest it often is. They do not suggest it always is.
If software companies start training their own models for the jobs that matter most to them - and Intercom argues they must - then customer service is the first big proving ground. It is high-volume, high-stakes, and measurable to the decimal. Get it right and you have a template. Get it wrong and you have a very public mess.
Renaming a globally recognized 15-year-old brand after one product is the kind of decision that looks either visionary or reckless, with very little room in between. Intercom has made its choice. The rest of the industry is watching to see which it turns out to be.
So return to that Tuesday. A customer types a question. A few seconds later, an answer - personal, instant, and handled before anyone had to be put on hold. For fifteen years that was a promise the company couldn't quite keep at scale. Now it keeps it a million times a week, and it has renamed itself after the thing that finally made it true.