The AI-Native Approach with Des Traynor

San Francisco-headquartered Intercom is a privately owned consumer interaction platform. It uses a subscription-based model to deliver a range of customer service solutions to companies. Despite the shakeups caused in many sectors by the advent of artificial intelligence (AI), customer service included, Intercom’s overall business mission remains the same, says Co-founder and Chief Strategy Officer Des Traynor. 

“We set out to build a product that would let internet businesses talk to their customers.”

Before becoming Chief Strategy Officer — “it’s a fluffy title for sure,” he admits —Traynor managed the product, marketing, customer support and content teams within Intercom. He also has also leveraged his experience as a researcher in computer science education to become a leading voice on startup scaling and product strategy, and has backed a number of startups, including Stripe and Notion. 

In the latest edition of OPTO Sessions, Traynor joins us to discuss how to build an AI-first company, the changing calculus of pricing AI solutions, and why he thinks the idea of a competitive moat is “really overstated”.

Building AI-First

What makes a company AI-first? In Traynor’s opinion, many big tech incumbents have taken what he calls the “AI agnostic” approach: “They sprinkle on bits of AI here and there, maybe a little bit of a co-pilot and a few sparkly buttons … but they’re kind of leaning on the strengths of yesteryear and hoping that they can survive.”

In contrast, an AI-first company has “taken an entirely AI-native approach toward solving the problem in their problem domain.” It has looked at the current landscape and decided to “build exactly what a startup would build if they were formed here today in 2025.”

For Intercom, that has meant a complete rework of how they do business, Traynor observes. “So much about the product prior to AI was assuming that it was humans who sat in seats having conversations with customers to deliver customer service. And when AI happened, it became very clear that most customer service could be dealt with through these [large language models]. And because of that, we had to change what we were building, because we could have the shiniest, fanciest seats in the world, but no one’s going to sit in them if there’s an agent around the corner who will do all your work for you.”

And so the focus became the agent. In Intercom’s case, that’s Fin. “Imagine you were delivering a version of [ChatGPT] that knew everything about your business, talked to your customers on your behalf, and was guardrailed to stay totally on topic and only have conversations that represent the brand and tone of voice that you would want your company to have,” Traynor says. “Then imagine it was fully agentic, where it could go and carry out backend tasks.”

As an AI customer service agent, Fin already boasts impressive ticket resolution rates, at around 65% of all cases. Better still, it is able to learn and improve on its performance: “And if you’re willing to keep using it, it’ll actually trend upwards and upwards,” Traynor says. “We have customers with as high as 94% total support resolutions.”

Becoming an AI-first company has also meant readdressing pricing. In part, this was a change that affected many software-focused businesses. “When AI happened it was the first time that software engineers and product teams had to deal with the fact that code didn’t execute for free anymore,” Traynor explains. 

Rather than choosing to create Fin as an add-on, the company decided to focus its full attention on Fin as its flagship product — and to tie revenue to its performance. “Then there’s a little bit of calculus around making sure that it didn’t bankrupt us along the way,” Traynor says. By pricing based on the volume of outcomes Fin delivers on, “we’re heavily incentivized to make Fin better and better,” he says. “Our customers want that too.”

But what of the remaining 35% of tasks that Fin cannot resolve? “What’s left is often a complex gnarly ambiguous back office task,” Traynor explains. “It’s not that it’s so complex [that Fin] can’t get it. But it’s just arduous to give Fin all of that in a policy. And a lot of people just say, ‘we’ll just let the humans do it.’” 

Intercom has a solution for that, as well, Traynor says. “If they want to hire us as a pro-serve agency, we’ll step in and just do it for them.”

However, as Fin improves, the tasks left over for human agents to resolve will drop in number. “Fin is a cannibalistic product to the Intercom platform. The more Fin succeeds, the less interior seats people will need in the future, or certainly the less seats they’ll progressively buy into the future.”

The Myth of the Competitive Moat

When speaking about a company’s competitive edge, many investors reach for the concept of a competitive moat — a feature that puts significant distance between one company and its closest competitors. In Traynor’s opinion, however, in software-as-a-service, moats tend to be “really overstated … It’s getting a lot easier to produce a lot of code… and it’s becoming easier now than ever to transpose data from one application to another.”

That said, being the only player on the field can help, Traynor admits. “We’re the only people building a world-class agent on top of an actual help desk product.” 

Even among AI-first companies, the model for deploying AI outcomes can make a big difference. First, you have what Traynor calls the “Palantir-style [PLTR] deployment, where consultants come on-site with their clipboards and configure things.” Then, you have companies like Intercom that are using AI to build solutions that clients can deploy on their own. 

The market is also trending toward unified solutions, Traynor notes. “We believe that the market equilibrium here will not be that every business goes and buys 11 different agents. We actually believe they’re going to buy and train and adopt one agent, and that that will be better for users, [and] better for business.”

Even with AI developments seeming to change the playing field every few months, it’s important to recognize that, in some ways, business continues as usual, Traynor says. “Brand and community are still going to be the residual longer modes.”