Rovio spent over a year integrating AI tools to streamline its customer service ticketing. The results shine a light on the plusses (and minuses) of relying on AI chatbots.
Ah, customer service. Everyone's favorite job, and everyone's favorite experience. Something broke, and depending on the product, you either navigate a confusing phone tree to try and speak to an actual human being, or you file a ticket in an online portal and hope to get something other than a form response.
It's grueling, thankless work that can pit angry customers against underpaid humans already bound by a rigid script. So it should be the perfect job for an AI chatbot to step in...right?
Earlier this year we got pinged by Rovio, the developer and publisher of the Angry Birds series. They were keen to discuss how using Keywords' Helpshift AI tools were drastically improving their mobile retention and revenue. After chatting with player support lead Pascal Debroek, it was clear these AI tools were making a difference in their work.
According to Debroek, Rovio's integration of AI tools hasn't been so much about reducing expenses as it has been an exercise in improving player retention.
That's not an angle we've heard a lot about generative AI—and Rovio's experience shines a light on what developers can expect by integrating chatbot tools in their own customer service pipelines and beyond.
Rovio's process for integrating AI tools wasn't as simple as opening a new tool and just pushing "add chatbot." The company not only had to customize Keywords' tools to integrate its existing Zendesk-based tickets and customer data, but update its entire library of pre-written responses to work with short-form bot messages instead of long-form emails.
Debroek said it took about nine months to execute the switch between customer service platforms—and that was after a planning process that was drawn up about a year and a half before beginning the transition.
The results were immediate. To previously file a ticket, players could click a customer service button in one of the studio's many Angry Birds games, but they were taken out of the app to file the ticket through a webform, and would need to manually input information like Player ID and device specs. To follow up on tickets, they'd have to open up an email app and converse with an agent.
Rovio's mobile-first business model means that the company isn't just competing with other games, it's competing with other apps on a user's phone. Debroek noted that if a player runs into an error like not receiving currency for a payment, or not being able to log in during a major event, they're liable to just switch apps and zone out on social media.
Any friction like the kind described above could impact the company's business. The new AI chatbots make two improvements to that process. Because the bots can be loaded in-game, they're able to automatically ingest information like the user's player ID and device specs.
Additionally, because the player isn't exiting the app (and can be updated via the chatbot after the company addresses their issue), they're spending more time in the game, which will ultimately drive either in-game spending or viewing in-game advertisements, two crucial metrics for Rovio's business.
"By keeping that loop with player support very tight and within the game, it ensures that players are staying within the game for a longer time, they can still play most of the times when they're actually logging into support tickets," Debroek says. And while the engagement metrics noted above are important to the company's bottom line, there's something to be said about boosting how much time players are having fun rather than plodding through the support process.
Once players are talking to a bot (and Keywords vice president of customer success & growth, player engagement solutions Samantha Pang confirmed that players are informed they're speaking with a bot), they're led through Rovio's FAQs that will help resolve the majority of problems. Debroek explained that most of the automated responses were previously copy-and-pasted by human agents, so there isn't much of a shift in the customer's experience.
(That human customer service agents have been restricted to such form-driven responses...well, we'll get to that later).
The result, Debroek said, is that Rovio's human customer service agents receive "considerably [fewer]" tickets. "A lot of the issues get solved automatically," he said.
The bot-based approach means his team now spends more time studying analytics—both to identify what issues need to be escalated to the development team, and to spot which FAQ responses or lines of questioning from the bots aren't giving players the help they need. And of course, when the strangest and most complicated problems do surface, agents can jump into problem solving right away without a back-and-forth exchange collecting information.
Game developers are a savvy lot. If someone comes along saying something like "AI will automatically improve customer service pipelines," the first thing they'll ask is "okay, but what new problems will it cause?"
For instance, Debroek said Rovio has had to deal with humans metagaming the bots (which feels like a distant echo of "prompt engineering" that supposedly we'll all have to become experts in if we want to keep our jobs). Players quickly worked out that tickets for disruptive behavior or offensive names are given the highest priority and are acted on quickly. "What they started doing was they they were trying to abuse the bots.
They would report players, and once they actually got through to an agent, they would say 'oh yeah, I have this other problem.'"
The next challenge was that bots struggle with game-specific slang. For instance in the Angry Birds titles, players will deliberately try to lower their competitive ranking by attacking vastly higher-level opponents. This means they'll begin playing against lower-ranked players and can easily farm wins to earn rewards.
Angry Birds players call this "suicide." When they describe it to a bot, they escalate it to a trust and safety issue. You can probably guess why this was a rather unique challenge for Rovio to deal with.
Otherwise, Debroek said transitioning to AI chatbots has produced positive results. The tools have made it easier to scale up resources during the holiday season (when players are spending more money thanks to gift cards or special in-game events), and it's opened up new horizons to explore what "player support" means.
I quizzed Debroek on what this switch would mean for employing human customer service agents, but the question rang hollow even as I asked it. Digital customer service operations are not what I'd call a pleasant job. Many operations require workers to hit quotas with resolving tickets, and they're often tasked with facing customers at their angriest moments.
Bots can take the pressure off of humans but they feel like a band-aid on a bigger problem: the very retention and revenue metrics that Debroek said they help improve.
In order to efficiently scale user support for what may be thousands or millions of customers, corporations have turned the act of a knowledgeable human being helping someone out into a white collar-adjacent assembly line. The bots aren't taking the human element out of this customer service process because—well, that human element was excised a long time ago. The human instinct to help others solve their problems has been warped into a tool to extract every dollar possible.
That said, player support specialists in the world of game development can be in a lucky position. Their job, when well-structured, is about helping people have fun. Debroek expressed sincere relief that these improvements are helping people spend time playing an Angry Birds game instead of mucking about in support trees.
But the world of free-to-play game design has gotten intensely extractive—a problem players are beginning to respond to as the world of live service games takes blow after blow, erasing the reliable revenue they were meant to supply studios.
The solution to this capitalism-driven problem can't be sussed out in a few thousand words. It may in fact signal new challenges for Debroek and other player support leads down the line.
You May Also Like
