Product

Five Years, 250 Chats a Year: Building Live Engagement with The Washington Post

Over the past five years, we’ve had the privilege of powering Community Chats for The Washington Post , delivering the live technology behind approximately 250 chats each year — or about one per business day.

When we first began working together, we were a small, ambitious SaaS company with a clear goal: build a Live Q&A solution that could meet the demands of a world-class newsroom. Entering into a partnership with a global media organisation — anticipating the scale, standards, and pace — was both exciting and daunting.

This post is a brief timeline of that journey.

Starting as a New Supplier

In the early days, our focus was simple: listen carefully and execute reliably. We worked closely with editorial, product, and technical teams to understand exactly how their Community Chats needed to function — from moderation workflows and audience engagement features to UI compatibility and performance under high traffic.

As a small team, agility was our advantage. We iterated quickly, refined features based on real feedback, and adapted our product to align seamlessly with The Washington Post’s brand and user experience. The goal wasn’t just to provide software — it was to make the chats feel native to their platform.

Growing Through Collaboration

Over time, our role evolved. What began as a supplier relationship became a collaborative partnership. We anticipated needs instead of simply responding to them. We invested in stability, scalability, and user experience improvements to support a cadence of one or more live events per business day.

Reliability became non-negotiable. With hundreds of high-profile journalists and thousands of readers participating annually, every chat had to work flawlessly. That consistency built trust — the most valuable currency in any long-term partnership.

Becoming a Trusted Partner

Five years on, we’re proud of what the relationship represents. As a small company, we’ve proven that size doesn’t determine impact. Focus, responsiveness, and commitment do.

Today, we’re not just the team behind the technology. We’re a trusted partner supporting one of the world’s leading news organisations in engaging directly with its audience.

And while the platform continues to evolve, one thing remains constant: our belief that strong partnerships are built the same way great products are — through listening, collaboration, and delivering consistently over time.

Technology

From Dozens to 100,000 in Real Time: Five Years of Scaling Engagement

Over the past five years, the definition of “real-time” has changed dramatically for us.

What once meant supporting a few hundred people in a live Q&A now means architecting systems that can handle 100,000 concurrent connections without hesitation. The journey from small, controlled live sessions to global-scale interaction wasn’t a single breakthrough moment. It was a steady progression shaped by client demands, live pressure, and a growing understanding that engagement is not a feature — it’s infrastructure.

In the early days, our focus was simple: make it work, make it stable, make it clean. If a few hundred participants joined a live chat, we needed low latency and intuitive moderation. That was success. But as our clients grew — and as global events moved increasingly online and hybrid — expectations changed. Audiences didn’t arrive gradually anymore. They flooded in. Traffic spiked in minutes. Messages surged in waves during key announcements or on-stage moments. The system had to hold, no matter what.

Scaling to 100,000 real-time users forced us to rethink everything. Infrastructure became distributed and elastic. We built for traffic bursts rather than steady flow. Redundancy and failover planning became standard, not optional. Performance testing stopped being theoretical and started reflecting real-world stress conditions. At that level, even milliseconds matter.

But scale wasn’t the only shift.

Five years ago, moderation was largely human-driven. Teams reviewed questions manually, filtered inappropriate content, tagged themes, and tried to keep pace with fast-moving conversations. That model worked — until it didn’t. As volume increased, the velocity of messages outpaced what any team could reliably process in real time.

That’s when AI stopped being a future concept and became operational reality.

Today, AI assists in filtering toxicity, detecting spam, clustering similar questions, identifying sentiment trends, and surfacing priority topics before they overwhelm moderators. The role of human moderators didn’t disappear; it evolved. Instead of firefighting chaos, they now oversee structured streams of pre-processed signal. The difference is profound. What used to feel like managing noise now feels like governing flow.

This evolution became especially visible as we moved from small trials to long-term strategic partnerships. Early on, conversations were often framed around pilots — a single event, a single newsroom initiative, a proof of concept. Over time, as reliability and intelligence improved, those conversations changed. Engagement wasn’t just an add-on anymore. It became part of how organizations thought about live communication itself.

Working alongside global enterprises like Microsoft marked an important milestone in that shift. The discussion moved beyond “how do we enable chat?” to “how should live audience input shape the event experience?” Engagement became part of event design, moderation became part of governance, and audience signal became part of enterprise intelligence. When operating at global scale, experimentation gives way to expectation. Systems must be secure, compliant, measurable, and resilient — every time.

Looking back, the most meaningful transformation isn’t the jump from hundreds to 100,000 users. It’s the conceptual shift from interaction to structured signal. Real-time engagement is no longer about collecting as many messages as possible. It’s about extracting clarity from thousands of simultaneous inputs and enabling organizations to respond intelligently while the moment is still unfolding.

The next chapter will likely move even further in that direction. Predictive models, deeper AI-assisted governance, cross-event intelligence, and tighter integration between audience insight and executive decision-making are already emerging. The velocity of expectation isn’t slowing down.

Five years ago, we were focused on enabling conversation. Today, we are focused on enabling understanding at scale. The journey from small real-time rooms to global concurrent audiences taught us that growth isn’t just about numbers. It’s about responsibility — to infrastructure, to safety, to clarity, and to the organizations that trust us when their audience shows up all at once.

And if the last five years were about proving we could scale, the next five will be about proving we can make that scale intelligent.