Des Traynor (Intercom): How we rebuilt the company for the AI era

Intercom is one of the most iconic SaaS companies of the past decade – a platform that reshaped how businesses communicate with their customers and set the standard for modern, scalable customer support. At the center of that evolution is Des Traynor, Intercom’s co-founder and the product strategist who helped steer the company through every major inflection point, including its recent shift to AI.

Des is also the co-founder of Fin AI, the rapidly growing customer support AI agent built on large language models. With more than 7,000 customers and over a million resolutions per week, Fin has become one of the clearest examples of AI delivering real business outcomes at scale.

For leaders wrestling with how to integrate AI into their products, teams, or operating models, Des’s perspective is uniquely valuable. He’s lived the before and after, from the moment generative AI upended Intercom’s roadmap to the hard organizational rewiring required to survive the new pace of innovation. His experience offers something rare in the current AI noise: clarity, candor, and a blueprint for how companies can move from experimentation to transformation.

A pivot born in a single weekend

There are only a handful of technological breakthroughs in our lives that become personal timestamps – moments so transformative that we remember exactly where we were when we first used them. For Des Traynor, that moment was the launch of ChatGPT.

The day it went live, Des received a Slack message urging him to try it. His first prompts were no different from anyone else’s: quick, simple questions that felt like slightly more elegant Google searches. But everything changed when he asked ChatGPT how to install Intercom in an iOS app. The model returned an answer in seconds. And, in that instant, he knew that the future of customer support had arrived.

Within 72 hours, Intercom scrapped roadmaps, reorganized teams, and made a company-wide decision to go all in on AI – a pivot that would ultimately reshape its product, its culture, and its trajectory.

The hard part no one warns you about

Intercom’s hard-turn into AI was fast, but it wasn’t frictionless. Years of well-honed product instincts suddenly collided with a new technical reality: uncertainty. Before AI, Intercom’s teams could trust their craft. If they wanted to build a dashboard, design a workflow, or ship a new feature, they generally knew it was possible. With AI, that confidence dissolved overnight.

Des describes the learning curve bluntly.

Even seemingly simple ideas like auto-generated replies for support reps broke down quickly when hallucinations, edge cases, and verification costs crept in. If a human had to spend the same amount of effort double-checking the AI’s output, the value disappeared.

The challenge was also a cultural one. Thirteen years of reliable, linear product development had to be dismantled and rebuilt. “We had to entirely redesign how we build software from the bottom up because everything is different now,” Des explains. Old processes had to be thrown out. New ones had to be invented. And not everyone loved the pace of change.

Fin had demand, but Intercom had a distribution problem

Looking back, Des is clear. Even for a company moving fast, Intercom didn’t move fast enough in the right direction. One of the earliest missteps was assuming Fin’s value would be so compelling that customers would switch their entire help desk to use it. The team expected the product to pull the market toward Intercom. Instead, they discovered exactly how hard platform migrations are in practice.

Customers wanted Fin. They just didn’t want to uproot their existing systems to get it.

If Intercom had been building Fin as an AI-native startup rather than a product within a mature platform, Des says the first move would have been obvious: meet customers where they are.

Once Fin became available as a layer on top of platforms like Zendesk, adoption exploded. And when Fin started resolving 60% to 70% of inquiries on those legacy systems, migrations to Intercom suddenly became easier, not harder. Beyond enhancing support workflows, AI actually changed the economics of switching.

Decreasing the cost of exploration

As Des explains it, one of the most overlooked advantages of AI isn’t speed, but the ability to see around corners. Before generative models, exploring multiple product ideas, pricing models, or design directions required time, coordination, and a lot of people. Teams had to pick a lane early because exploring five different options simply wasn’t practical.

AI changed that overnight.

“AI massively decreases the cost of exploration,” Des explains. At Intercom, that meant roadmaps could become something almost interactive. Instead of debating which direction to pursue, teams could “vibe code” an entire set of ideas and evaluate them side by side before committing. It marked a shift in the company’s product process from guessing what was to come to truly looking ahead.

Aside from product, this approach reshaped the way leadership made decisions, too. Des believes that most people readily accept that AI can make engineers more productive. What fewer leaders consider is how much it can improve their own judgment.

“Can AI make a VP better? Absolutely,” he says. “But many leaders still think AI is for the kids.”

The reality, he explains, is that AI gives executives a way to pressure-test decisions long before they land. Pricing strategies, hiring plans, go-to-market bets, all of them can be modeled, expanded, stress-tested, and refined in a fraction of the time. Not only are leaders choosing faster, they’re choosing with more insight.

In Des’s view, that’s the real shift: modern decisions are being informed by a richer understanding of how each option plays out. AI makes strategic planning a living process where teams can explore more paths, uncover more tradeoffs, and move forward with far greater confidence.

What AI unlocked behind the scenes at Intercom

While most of the attention around AI goes to the product itself, Des notes that some of the biggest gains appear behind the curtain – in the finance, pricing, and operations decisions that shape how a company moves. These teams have always been essential, but their work has traditionally been defined by manual analysis and long turnaround times. AI has changed that.

Pricing is one of the most notable examples. Intercom routinely evaluates new pricing and packaging ideas, especially as AI features introduce new forms of value. Historically, this meant handing a hypothesis to the business operations team and waiting weeks for models, projections, and revenue scenarios to come back. With AI, Des said, that dynamic flipped overnight.

In his words, you can sit down for a few hours with a model like Claude and explore an entire landscape of pricing possibilities, from token-based pricing to add-ons to outcome-based tiers. AI can read the relevant literature, reproduce known strategies, generate alternatives, build the models, and present interactive dashboards that let teams adjust assumptions in real time. What once took two weeks can now be done in an afternoon.

“This is truly the better, faster, cheaper type thing,” Des says.

Finding a solution in outcome-based pricing

When Intercom began thinking about how to price Fin, the team quickly realized they weren’t just shipping a new feature. They were introducing a product that would fundamentally reshape how customer support teams work. While it was true that Fin made support reps faster, in many cases, it replaced entire categories of work altogether.

That created a strange tension. Intercom’s business had always been tied to seats (the number of human agents using the platform). But Fin’s success meant those seats would naturally shrink over time.

“Fin is a very cannibalistic, hostile product to our core business,” Des says. “That’s just the reality.”

So, the team set forth on a journey to reimagine the economics of their product. If Fin resolved a customer’s question instantly and accurately, shouldn’t that outcome be the thing customers paid for – not the number of people on their support team?

“If Fin succeeds, there will be fewer CS people in the future. Seats will go down,” Des explains.

Yet, for customers, the value of those resolved conversations would only increase.

The answer, then, was to tie pricing directly to results. So Intercom shifted to outcome-based pricing – charging per successful resolution – because it aligned incentives with precision. The more Fin accomplished, the more value customers received. And the more accurate it became, the more the model improved itself.

Des is quick to point out that even this approach wasn’t simple. In reality, not all resolutions are created equal. Some involve long chains of reasoning. Others end in human handoffs that still consume tokens and model overhead. As he puts it, “Sometimes we burn a lot of tokens and still hand over to a human and get zero.”

Even with its imperfections, outcome-based pricing made one thing clear: alignment matters. Customers weren’t paying for the promise of AI. They were paying for what AI actually achieved.

The new frontier of enterprise software

For Des, the biggest opportunity (and the biggest threat) facing enterprise software isn’t a specific feature, model, or workflow. It’s the pace of change itself. With AI, the team can accelerate expectations, erasing the comfortable distance between idea, prototype, and production.

Des describes this as a mindset shift that many companies still underestimate.

“Reimagine your company in a way where the speed of progress is extremely, extremely fast,” he says.

The teams that thrive in this environment don’t cling to multi-year roadmaps or static operating structures. They build for movement: shorter cycles, more iteration, and a greater tolerance for ambiguity.

The challenge, Des explains, is that AI introduces far more uncertainty into a business than most leaders are used to. New models drop unexpectedly. Competitors take sharp turns into adjacent categories. Features that once required months of engineering can now be prototyped in a weekend.

“There’s no choice where ‘do nothing and succeed’ is actually on the table,” he says.

That reality forces companies to interrogate their assumptions. What would you build if you were starting today? Which functions still need to exist? Which processes slow you down? And which parts of the business are you maintaining out of habit, not necessity?

The organizations that succeed in this new frontier are the ones that are willing to rethink their systems entirely, even when it’s uncomfortable and breaks long-held structures. 

“Either you correct your business with AI, or the market will correct you,” Des says.

AI and the build-versus-buy trap

As more enterprise teams race to adopt AI, Des has noticed a familiar pattern. Companies dramatically underestimate what it takes to build AI tools that actually work. On the surface, it’s tempting. A quick demo, a hackathon prototype, or a simple model prompt can give the impression that an internal solution is only a few weekends away.

But, as Des puts it, most teams don’t grasp the depth of the problem they’re stepping into.

The difference between a demo and a dependable system is the difference between a hobby project and a product like Fin, which relies on dozens of subsystems, custom post-training, guardrails, and constant architectural refinement.

That complexity raises a harder question. Even if a company could build it, should they?

Des explains:

He notes a few exceptions, like companies with razor-thin margins or finished products that have shifted into pure efficiency mode. But, for the vast majority, Des believes that the opportunity cost is too high. Building internal AI systems often means spending months solving problems that specialized vendors have already spent years perfecting.

And, even then, teams risk backing themselves into brittle, hard-to-maintain systems that struggle to keep up as the AI landscape continues to evolve.

Des’s advice is pragmatic: buy where you can, build where it meaningfully advances your strategic edge, and stay honest about which category you’re in. For tech leaders, the end-goal shouldn’t be to own more software. Instead, Des advocates for making better decisions about where your talent creates the most value.

Designing for radical simplicity

Des believes the new bar for product design should be to ask the user for as little as possible, and let the system handle the rest. 

That philosophy pushed Intercom to rethink how customers configure Fin. Early versions relied on a single text box for guidance – beautifully simple, but not enough. People didn’t give the depth of instructions the model needed, so Intercom introduced just a bit more structure. 

You can see that approach in Fin’s onboarding. It can ingest a help center and produce a working demo in minutes, with no long setup or training sessions. That kind of immediacy not only improves the product experience, it also broadens who can use it.

For Des, this is where AI has its most transformative impact. It moves software from something you learn to something you ask. And when anyone can articulate what they want – without mastering a tool first – simplicity becomes a powerful growth engine.

What the next 3 to 5 years look like

Looking ahead, Des believes AI will change not only how teams work, but who does the work. The tidy boundaries between roles – product manager, designer, engineer – are already starting to blur as AI makes it easier for anyone to work across disciplines.

“Everyone will be broadly capable and dangerous in most roles,” he said.

He expects a convergence: product teams made up of builders who can design, prototype, write code, and shape strategy, with AI filling in the gaps. Specialists will still exist, but their advantage will be depth, not exclusivity. The same shift is coming to customer-facing teams. Instead of separate SDRs, AEs, and CSMs, Des sees a future where one AI-supported individual can manage far more of the customer journey.

It’s a return, in some ways, to the early days of tech, when generalists carried whole products on their backs. The difference now is that AI acts as the amplifier, giving each person more leverage. Des believes that the leaders and teams who embrace this shift, who see AI as a way to expand their range rather than protect their lane, will be the ones who move fastest.

Conclusion

Across the conversation, Des returns to a single idea. While it’s true that AI is changing how products work, it’s also enabling companies to rethink how they plan, make decisions, organize their teams, and define their competitive edge. The shift is a matter of embracing a new operating rhythm where exploration is cheaper, iteration cycles are shorter, and the distance between vision and execution narrows dramatically.

For organizations willing to forego their assumptions, AI becomes a force multiplier. It gives teams the ability to test ideas before committing, model outcomes before acting, and make choices with more clarity and confidence. But the companies that will thrive in this new era aren’t the ones who bolt AI onto old processes. They’re the ones who redesign those processes from the ground up.