How Fin (formerly Intercom) drove 80% more efficiency in realtime delivery for AI chat

Fin (formerly Intercom) powers realtime conversations between businesses and their customers at global scale. Those conversations now increasingly include AI, with customers expecting fast, seamless responses that feel as reliable as the rest of the product.

For years, Fin used an in-house realtime system called Nexus to support experiences like chat updates, presence, and typing indicators. Nexus served the company well for that era of the product. But as Fin's needs evolved, particularly with the growth of Fin AI Agent, its AI customer service agent, the team needed a different model for realtime delivery.

Rather than continue investing in infrastructure that wasn't core to Fin's differentiation, the team chose to partner with Ably. The result was a more dependable foundation for AI-powered conversations, a simpler path for delivering realtime message data, and more engineering focus on the product experiences that matter most to customers.

The challenge: AI changed what Fin needed from realtime

At Fin, realtime has always been central to the customer experience. Conversations need to move quickly and feel responsive across devices, regions, and channels.

For a long time, Nexus was a good fit for that job. It was designed for a world where lightweight signaling was enough to keep clients in sync and support the messaging experiences Fin needed to deliver.

That changed as the product evolved.

Colin Kennedy, Principal Product Engineer at Fin, saw this shift play out directly. "With AI, interactions are actually in realtime, as opposed to delayed responses from humans. Therefore realtime connections matter even more. A dropped connection can mean that users lose entire AI responses with detailed information. That's unacceptable.”

The rise of AI made realtime delivery more critical. With Fin AI Agent, Fin wasn't just updating interfaces or signaling that a conversation had changed. The company was building experiences where users expected answers to arrive fluidly and reliably, with no gaps, stalls, or ambiguity.

That raised the bar for what the underlying infrastructure needed to do. Fin wanted stronger delivery guarantees, more confidence in how message data reached clients, and a foundation that could support the next generation of AI-driven customer conversations.

Just as importantly, the team recognized that building and maintaining that foundation themselves was no longer the best use of engineering time.

Fin considered what it would take to keep evolving its in-house realtime platform. But the question wasn't simply whether the team could do it. It was whether they should.

As the company pushed further into AI, the priority was clear: spend more time building differentiated customer experiences, and less time maintaining infrastructure that specialist platforms were designed to handle.

For Colin, the answer came down to a core engineering principle. “We have a philosophy at Fin: run less software. Realtime infrastructure isn't our core competency. We'd rather invest engineering time in building better AI agents and customer experiences than maintaining websocket infrastructure.”

That philosophy shaped the decision. Nexus had served the business well, but realtime infrastructure was not the area where Fin wanted to keep concentrating operational and engineering effort. The team wanted a partner that could provide dependable, production-grade realtime delivery while allowing Intercom to stay focused on its core product.

That made Ably a strong fit.

Why Ably

Fin needed a realtime platform it could confidently put on a critical path for customer conversations and AI-powered messaging.

Ably stood out because it offered the delivery guarantees and operational maturity the team was looking for, without requiring Fin to rebuild or keep extending its own infrastructure. It also aligned with a broader architectural goal: move away from indirect, workaround-heavy patterns and toward a simpler model for delivering realtime message data.

The team ran a proof of concept and saw quickly that Ably could integrate cleanly into Fin's existing architecture. That gave them confidence not just in the technology itself, but in the practicality of adopting it without creating unnecessary disruption.

For a team trying to make a disciplined build-versus-buy decision, that mattered. Ably gave Fin a way to modernize a critical part of its realtime stack without turning that modernization into a major internal platform project.

The implementation: starting where reliability mattered most

Fin took an incremental approach.

Rather than trying to replace everything at once, the team started with the use case where reliability mattered most and where the value of a stronger realtime foundation would be easiest to see: Fin conversations.

That made the rollout strategically important. Fin AI Agent was already becoming a central part of the customer experience, and it highlighted why Fin's realtime architecture needed to evolve. By introducing Ably first in this part of the system, the team could validate the new approach in a high-impact environment while keeping the migration focused and manageable.

This phased rollout also made it easier to build internal confidence. Instead of arguing for a broad architectural shift in the abstract, the team could show how a managed realtime platform improved an important product experience and reduced the need for workarounds in production.

Working with Ably gave Fin more confidence in how realtime messages are delivered, especially in the parts of the product where reliability has the greatest impact on user experience.

That confidence created room for architectural simplification. Instead of designing around the limitations of an older realtime model, the team could move toward a cleaner approach to delivering message data and supporting richer, more dependable experiences.

It also reinforced the company's broader engineering philosophy. By relying on Ably for a specialized infrastructure capability, Fin could spend less effort maintaining realtime plumbing and more effort improving the product itself.

“The biggest win isn't just that things work,” Colin reflects. “It's that we trust the system. We're not designing around gaps anymore. We're building what we actually want to build.”

For a company investing heavily in AI, that trade-off matters. The goal wasn't just to modernize one layer of the stack. It was to create a foundation better suited to the kinds of product experiences Fin wants to build next.

What's next

With Ably established as part of Fin's realtime architecture, the team is looking at where else that foundation can support the company's AI roadmap.

A natural next step is expanding Ably's role in Fin's AI Agent, including more of the realtime experiences that power AI-driven conversations. Beyond that, Fin sees broader potential for Ably across additional realtime use cases where dependable delivery and a simpler architecture can unlock better product experiences.

In that sense, the move to Ably is not just a platform migration. It's part of a larger shift in how Fin thinks about building for an AI-first future: use managed infrastructure where it makes sense, reduce operational complexity, and keep engineering focus on what differentiates the product.

Advice for other teams

For companies still operating homegrown realtime systems, Colin's question is a simple one: “Is maintaining this system helping you build a better product? If the answer is no – if you're spending time keeping the lights on instead of innovating – it's time to evaluate alternatives.”

Learn more about Ably AI Transport

Ably AI Transport gives AI products a real-time delivery layer they can rely on — so your team spends less time on infrastructure and more time building the experiences that matter. Learn more here.