Company snapshot

Industry: Media and entertainment
Region: United States, Mexico and Latin America
Channels supported: Web chat, Facebook Messenger, Instagram, X (Twitter)
Volume: ~66,000 conversations in a typical month; nearly 3x that during the World Cup
Ada customer since: May 2022

In an ordinary month, about 66,000 conversations reach TelevisaUnivision's AI agent from viewers of ViX, its Spanish-language streaming service, with a question about a subscription, a login, a device, or a show. Then a major programming moment arrives. During the 2026 World Cup, about 237,000 conversations came in over 39 days, from customers around the world who needed assistance.

Richard Cruz is the Director of Customer Care at TelevisaUnivision, the media company behind ViX. He leads the customer service department, which supports viewers across the United States, Mexico, and Latin America. They have worked with Ada since 2022, starting with a scripted bot and moving to a generative AI agent in 2025.

Viewers in their own words

The scripted bot gave the team a good starting point, and it came with a standing task: for every question a viewer might have, someone on Rich’s team had to anticipate how it would be phrased and build a path for each variation. Generative AI changed that. "Generative AI lets viewers explain an issue in their own words," Rich says. "It also gives us more flexibility because we do not have to script every step of the conversation." Viewers can also raise more than one issue in a single conversation, instead of finishing one path and starting over with the next question. The team wanted the experience to feel less mechanical, answer more questions and make better use of the information already sitting in their systems.

Accuracy and brand voice mattered, especially in Spanish, the primary language for most ViX viewers, so the move was something the team had to weigh. What settled it was how much of the experience stayed in the team's hands.

After the move to an AI agent, automated resolution climbed from 43% to 51%. Satisfaction with the AI agent nearly doubled, from 20% to 37%, and overall CSAT rose from 40% to 62%. Rich points to several things working together. Generative AI got better at understanding how viewers describe a problem, and the team improved the knowledge behind it, built stronger Playbooks (the step-by-step flows the team designs for specific processes), and connected the agent to more of their systems. Underneath all of it sat a routine: review conversations, find where viewers get stuck, adjust the content and decide with Ada what to improve next. "The technology helped," he says, "but the results also came from regular coaching, good decisions by the Customer Care team and close work with our partners."

In-language support, the way viewers speak it

Rich's team supports an audience that writes the way it talks: both formal and informal Spanish, regional phrases, abbreviations and Spanglish. Getting Spanish right took more than translation. "Spanish support cannot just be English content translated word for word," Rich says. "Tone, regional language and the way something sounds to the viewer all matter." Automated translation could make content understandable, but the team wanted more control over the voice. For the messages that mattered most, the team created separate Spanish and English versions of their knowledge base. It meant more work whenever something changed, but helped them ensure that they got the Spanish voice and wording right. Ada's multilingual support has come a long way since the team started, and Rich credits Ada's team with working through the language questions alongside them the whole way.

Support where the conversation already is

The AI agent runs on web chat, Facebook Messenger and Instagram, because ViX viewers don't necessarily think of support as a separate place. "If they are already talking with ViX on Instagram or Facebook, they should be able to ask for help there too," Rich says. That's especially important during live programming, when people are talking about an event and asking questions in real time.

Social responses have to be short and fast, with the same accuracy and privacy as web chat. And when a viewer needs a person, the direct message becomes a support case that carries the conversation with it: what the viewer asked, what the AI agent already tried and why it passed the conversation over. The viewer doesn't need to explain everything again.

An answer that applies to you

A ViX viewer in Mexico may have different options than one in the United States. Someone who subscribed through a partner may need different steps than someone who subscribed directly. The AI agent needs to know who it's talking to before it can give the right answer. By connecting Ada to TelevisaUnivision's systems, the team lets the agent take into account where a viewer lives, what plan they have, how they subscribed and what's included in their account, instead of handing everyone the same instructions.

Take a viewer who can't get into their account. The goal is clear and the steps are well defined, so the AI agent can understand the problem, give the right instructions and use account information when it's available. Login and account recovery have become one of the agent's most successful use cases. So far this year it resolves about 80% of those conversations on its own, and password recovery flows earn strong CSAT from viewers. Simpler topics may post even higher automated resolution numbers, but this topic is the one Rich cares about, because getting back into an account means getting back to watching ViX.

Ready before the first match

All of that was in place before the biggest test of the year. The 2026 World Cup meant weeks of live matches, and the team knew the questions would come in far faster than usual. So the preparation started with addressing the issues viewers were most likely to bring: subscriptions, account access, devices, playback, programming and what content they could watch. The team reviewed the knowledge behind each of those, updated the Playbooks and tested the most common conversations. Then they got ready for the match days themselves, writing opening messages for known issues and live updates, and setting alerts so any change in the numbers would show up quickly. Throughout, Customer Care worked with internal teams and Ada to think through what viewers might need, and kept updating as they learned more.

Then the first major match arrived, and conversations rose fast. "That is when all the preparation becomes real," Rich says. The team watched volume, containment, new issues, customer feedback and how many conversations needed a person. The AI agent took the routine questions as they came in, which let the team focus on the issues that needed investigation or human judgment. Ada's Reporting helped them see what was working, spot problems and keep leadership informed about what viewers needed and how the operation was doing.

Over the tournament, about 237,000 conversations came through in 39 days, nearly three times the normal pace. Containment rose from 74.6% in May to 82.1% during the tournament. Handling that volume without the AI agent would have required a very different staffing plan, and even then, Rich says, responding immediately to thousands of viewers at once would have been hard.

Rich’s team also leaned on their Ada experts during the surge. Rich describes them as the people who help his team make changes, find opportunities and figure out where AI could make the biggest difference. "The focus for both teams was on keeping support running, helping viewers and working through the numbers together," he says. "That created a lot of trust."

AI ownership matters

That advice describes how the team works all year. TelevisaUnivision runs the AI agent as an ongoing Customer Care program. Rob Barranco handles much of the day-to-day work, alongside team members who build, test and improve Playbooks and content. Beyond Customer Care, the program partners with the teams responsible for systems, programming, product information and customer policies.

What the program does need is ownership. "The AI agent is not something we set up and walk away from," Rich says. "It needs people to own it, test it, review the results and keep making it better."

The insight flows outward too. Customer Care shares what viewers are asking, where they're having trouble and which issues are growing with other teams, input that can improve the product, the communications and the overall ViX experience.

What comes next

Rich's next ambition is a new channel. The team is exploring how to bring support inside the ViX video platform, so viewers can ask for help without leaving what they're doing. Beyond that, Rich wants support to become more personal, more connected to TelevisaUnivision's systems and better at catching issues early, and Ada's MCP server is part of how the team gets there. It connects their Ada data to their LLM, so they can ask questions about their results in plain language. They can pull up a single conversation, ask why CSAT is changing, or find what could improve the experience.

For the viewers who reach out in an ordinary month, and the many more who show up when a major programming moment starts, the experience is meant to be the same one: explain the problem in your own words and get an answer that applies to you. Rich's advice to other media companies weighing an AI agent follows from that.