We spend a lot of time talking about what a great agentic customer experience looks like. With the ACX Awards, we get to show it in practice through our incredible customers. This year's winners handle millions of customer conversations a year and have certainly raised the bar. What earned them the awards is how deliberately they built AI customer service programs to scale, without compromising quality.

Five awards, each highlighting something different: the quality of the experience, whether conversations create value, the volume and channels supported, how insight from conversations feeds a continuous improvement loop, and how a team achieves outsized results.

Without further ado, meet this year's winners.

The ACX Architect: Luno

What the award recognizes. This award goes to the team that has most intentionally built and operationalized a complete agentic customer experience, with industry-leading results and deep technical expertise behind it. We were looking for a program built on purpose and end to end, rather than one strong use case.

Luno is a cryptocurrency exchange where customers buy, sell and store crypto. Support volume moves with the market. Every account holds real money, which puts those conversations inside KYC requirements. Their agent, Toshi, runs on Ada's Conversations API over their existing Zendesk stack, so it answers with real account data rather than generic help content. What set Luno apart was the team behind it. They built a team of specialists around Toshi, a group they proudly call the Lunauts, covering knowledge, data, integrations and content. To work out what to improve, the team ran a hackathon against thousands of real conversations. They also pipe their own data into Databricks and query it through MCP. Luno now runs 13 points ahead of the fintech and insurance benchmark on automated resolution, 18 ahead on CSAT and 7 ahead on containment.

The Lifetime Value Driver: Guthy Renker

What the award recognizes. This award is for turning support conversations from a cost to be managed into a driver of customer lifetime value. We were looking for a named revenue use case, not a cost saving reframed as growth.

Guthy Renker sells consumer beauty and wellness brands direct to customers on subscription, which leaves real value on the table when a customer asks to cancel. On their Smileactives brand, they first let the AI agent handle cancellations end to end. That alone took automated resolution on those conversations from 26% to 66% and lifted a large queue off the support team. The save flow came next: when a customer starts to cancel, the agent offers a loyalty discount or the option to skip or postpone a shipment instead. That is what changed the outcomes. In five months it kept more than a thousand subscriptions alive that would otherwise have ended. Because these are subscriptions rather than one-off orders, the value compounds into millions in retained revenue each year.

The Agentic Experience Optimizer: Dott

What the award recognizes. This award goes to the team that reads what conversations reveal, fixes causes rather than symptoms, and makes the experience measurably better over time. We were looking for a working improvement loop with a demonstrable before and after, not a one-off fix.

Dott operates shared e-bikes and scooters across more than 400 cities, handling roughly two million customer contacts a year through their agent, Dottie. Dottie runs 25+ API-powered automations so it can act rather than explain. With MCP, the team analyzes a thousand conversations at a time, so patterns that used to be hunches now come with evidence behind them. That precision let the team improve the agent quickly, and it also made support a partner to engineering. When something is not performing, engineering now receives a detailed diagnosis instead of a symptom, and a well-scoped request ships in a fraction of the time. Automated resolution has climbed from 32% to 77%, and containment now sits at 91%.

Read Dott’s customer story →

The Leverage Multiplier: NinjaTrader

What the award recognizes. This award is for outsized results built on smart system design, where the size of the team stops being the ceiling. We were looking for evidence that the results came from how the system was built, not from how many people were assigned to it.

NinjaTrader is a futures trading platform serving traders across North America and Europe. Questions tend to arrive while the market is moving, so waiting is expensive. Their agents handle 800,000 conversations a year across two brands and multiple channels at 70% automated resolution, and the team running all of it fits around one desk. In most improvement loops the bottleneck is diagnosis: working out why something went wrong before you can fix it. NinjaTrader automated that step, so the team arrives at a cause instead of hunting for one. Their hours go to strategy and to extending what the agent can do, rather than to triage.

The Voice Pioneer: Branch

What the award recognizes. This award is for taking voice past call deflection toward an experience that is voice-native and AI-directed. We were looking for a standout voice AI agent that’s handling real work for the company.

Branch provides the financial infrastructure that gets hourly and gig workers paid faster: cashless tips, payouts right after a shift, earned wage access and paycards. When someone calls, it is usually because their money has not arrived, which makes this one of the least forgiving support environments there is. Branch runs 50,000 of those calls a week through voice. They have also retired containment as their headline metric in favor of automated resolution. The logic is straightforward: a caller who never reached a human is not the same as a caller whose problem was solved. The agent greets cardholders by name and pulls live account data to answer them. When analysis showed that account closure calls all followed the same path, one new API endpoint let a Playbook resolve 60% of them on day one. Voice CSAT sits at 89%.

Crypto, beauty subscriptions, scooters, futures trading and payroll. Five industries with little in common, and five teams who arrived at the same habit: they treat the agent as an operation to be staffed, measured and improved, rather than a product to install and leave alone.