King’s Paula Ingvar defends 21,000 handcrafted Candy Crush levels, not agentic AI

Candy Crush is still printing about $1 billion a year, and King’s general manager Paula Ingvar is making a point as bold as a spoiler: she says the game’s 21,000-plus levels are handcrafted, and she does not want agentic AI placed between creators and players.

Ingvar, who has been with King for 11 years, frames the fight over AI as a fight over experience. Her argument is direct. “We’re not budging on the fact that level design and game design is a craft,” she says, adding that she “really struggle[s] to see right now that AI would add any player value if we put it between the people who make the game and the players.” In her view, the tension that makes Candy Crush fun is human. Game makers throw challenges at players, players beat them, and that constant battle is part of the enjoyment.

This is not just a creative philosophy. It is a business strategy built on retention, which, for mobile games, is everything. Ingvar emphasizes that the real scarce commodity is attention. “So, for us, it’s about two vectors,” she says. “One, the experience needs to be fun. Retention starts with joy.” Her team’s goal is that after you play “a round or two,” you are in a better mood, and Candy fits into your life rather than demanding too much. The sustainable approach, she says, is “to not pressure too much, because then we’re just going to exhaust people.”

That retention mindset explains why Candy Crush has stayed relevant in an industry that is increasingly tempted by automation. The global gaming market is valued at $386 billion this year, and the sector employs as many as 350,000 people directly, according to market research firm Gitnux. Meanwhile, the AI conversation keeps escalating across industries because it is still easier than shipping: agentic AI is now able to code new games at speeds that human game creators cannot match. So executives everywhere are asking a money question dressed as a talent question: can we automate the work and still keep the magic?

Ingvar’s answer is essentially: you can automate some parts, but you should not replace the craft that creates “human thought” players can feel. King does use bots internally. She concedes, “We have systems of bots that help us play through the progression [new game features] in 20 or 30 minutes.” The point is speed for internal testing or progression evaluation, not shipping levels that feel algorithmic. She contrasts that with a darker alternative: “The alternative would be to release straight to players, and that is what we did in the past, sometimes at the expense of the player experience.” In other words, King is using AI to reduce friction for operations, while protecting the part of the product that directly drives joy.

Her insistence that levels are “handmade and crafted by level designers” also looks like a defensive play against an emerging stigma. Ingvar says teams have been “met with skepticism or even cynicism,” including assumptions that Candy Crush levels are not built by humans. Her response is to wager their longevity and attention to detail: she says it is “quickly going to disprove that we are AI slop.” The bet is that players notice minutiae, and that the result of careful design is a sense that “someone has been paying attention” when you play.

That matters more than it sounds, because mobile gaming is a brutally competitive retail environment. Ingvar mentions users have “a hundred different options laid out before them every day.” When attention is scarce, small quality differences become survival differences. King also designs for access. Ingvar says the company has consciously maintained compatibility with older mobile devices and built for offline mode, calling it “one of the few games today you can play completely offline.” She ties these choices to inclusion and to the costs they sometimes bring: “Every day, this is a conscious choice, because sometimes there is a cost to these choices.”

There is also an international, infrastructure angle embedded in her comments. She references Ralph Mupita, chief executive of MTN, at Mobile World Congress in March, and the point is that Western tech leaders can forget many millions of people still operate on 4G and 5G devices, if they have experienced the digital world at all. For executives, that means product decisions are not purely aesthetic. They are constrained by networks, devices, and real user environments, which changes how you evaluate the payoff of automation.

Finally, zoom out to the corporate gravity around King. Candy Crush Saga was launched in 2012 with 65 levels, and now the number exceeds 20,000. King, its developer, was acquired for $5.9 billion by Activision in 2016. With King’s success, Activision’s price looks like a bargain by the standards of most mega-deals. The spinoffs include Soda Saga and Jelly Saga. And King now sits inside a broader gaming empire structure shaped by Microsoft’s $69 billion Activision Blizzard acquisition in 2023, alongside competitors like Tencent and Sony. In that ecosystem, a billion-dollar product is both an asset and a benchmark. The executives on every side of the AI debate now have a case study: when you are already winning, the question is not “can AI build games,” it is “can AI protect what makes players stay.”

Ingvar is effectively telling the industry to trust player taste over automation hype. “We just have to trust the taste of our players in the end,” she says. For peers, that raises the strategic stakes. If you replace human craft with agentic AI, you might save money, but you also risk eroding joy, weakening retention, and losing the tension that turns gameplay into a daily habit. In a market where attention is scarce and $1B revenues do not happen by accident, the cost of a bad experience is immediate.