Try Inworld
Request a Demo
Try Inworld

How AI NPCs helped MageCosmos increase playtime by 5%

AI NPCs are the future of social sims

Metrics
⬆️ playtime
Genre
Open world/Social
Integration
Existing game
Engine
Unity
Platforms
PC, Meta Quest

Overview

Ryan Yang, the founder and CEO of MageCosmos, is no stranger to AI. Before starting MageCosmos to build an engaging PC and VR open world sim game, Ryan was a Managing Director at a company that paired virtual and augmented reality with AI to create Smart Tutors for language learners. 

In that role, he witnessed the transformative potential of AI for interactive experiences. Four years after starting his own company, Ryan came across Inworld and was excited by how AI NPCs could also transform his game world. 

“Inworld and MageCosmos have a shared vision,” he explained. “We both want a future where AI elevates our digital experiences to realms previously uncharted.”

MageCosmos’ goal in adding AI NPCs, according to Ryan, was to create a transformative experience for players and, by doing so, increase their engagement and other key game metrics. 

Context

MageCosmos is a Unity-based sim game whose core gameplay includes farm management activities, island visits, mini-games, and social interaction. Set on a floating island in an enormous open world, players can visit islands belonging to other players or the game’s NPCs. 

Like other social sims, a core game loop in MageCosmos is social interaction – often with other players. This is a strategic choice by social sim developers since research shows that social-centric games are stickier. Casual mobile games, for example, often lose 90% of new players in just one week. But games with social features and multiplayer options are able to better retain players for months or even years. 

Younger players are especially motivated by social gameplay. A 2021 study by the Entertainment Software Association found that 89% of 18 to 34 year olds prefer games they can play with friends. For 35 to 44 year olds, that number falls to 81%. 

But building a social virtual world like MageCosmos can be challenging. Many users have limited weekly VR play time and they don’t want to encounter ‘empty’ worlds when they log on. They want to meet and talk to people every play session. 

In early gameplay, MageCosmos users live on scattered islands with low-density, so it takes time to find a neighbor. “People get bored as they explore blindly and that reduces play time,” said Ryan. “I realized that one way to fix that was to add AI NPCs to the world to make it come alive.”

Challenge

While Ryan wanted to add AI NPCs to MageCosmos, he wanted to make sure their integration into the world added to it. The AI NPCs needed to have different backgrounds, experiences, and knowledge so that they came across as distinctive and interesting characters. But he also wanted to make sure they were deeply embedded in the context and lore of MageCosmos’ world. 

“When a user talks to any AI NPC,” explained Ryan, “it’s important that the interaction is always interesting but it also needs to remain relevant.” 

What Ryan loved about Inworld’s Character Engine was how it allowed him to create AI NPCs with infinite possibilities but also to add the contextual awareness to characters that his game world needed. 

“AI NPCs live around the user,” he explained. “Their existence needs to have a reason and impact within the MageCosmos world. They should be able to develop memories and relationships with the players they speak to.” 

When he discovered Inworld with its Contextual Awareness layer and Long-Term Memory feature, Ryan said, “I knew I’d found exactly the tool I was looking for.” 

Why Inworld? 

Aside from the gaming native features that the Inworld Character Engine offered, there were a number of other key reasons Ryan felt Inworld was the right choice for MageCosmos. 

Easy to learn and implement

Had MageCosmos implemented AI NPCs directly from a large language model (LLM) API, the team would have needed to hire a machine learning engineer, a prompt engineer, a conversation designer, and other experts. However, with Inworld, the implementation was initially done by an intern. 

“We just hired an undergrad with a gaming and technology background and tasked him with learning Inworld and then implementing AI NPCs in our game,” explained Ryan. “It only took him a week to master the tool. The first group of AI neighbors were sent to MageCosmos world very soon after that.”

That ease-of-use was a key selling point for Ryan. “For small studios, understandable, affordable, and ready-to-use products mean everything,” he said. “It didn’t take long to decide to integrate Inworld and it turned out to be the right choice.”

Scalability and flexibility 

From his past experience working in conversational AI, Ryan knew first-hand how time-consuming it can be to train and tweak new AI agents. He was, therefore, pleasantly surprised that Inworld made both so simple. 

That’s allowed MageCosmos to scale their population of AI NPCs in a short time. “We’ve added AI NPCs in the public areas to guide new users and others that live on islands users can freely visit,” he explained. 

Given how simple creating new characters is, the team also had time to create a housekeeper named Naoko that lives on each players’ personal island. 

Ryan also likes how easy it is to analyze and tweak characters based on how players are responding to them. “I appreciate how flexible Inworld’s Character Engine is. You can monitor, make changes, or modify any AI," he said.

Deep relationships

What excites Ryan most, however, is how Inworld allows players to create a deeper relationship with AI NPCs over time via Inworld’s Relationship and Long-term Memory features. 

“We’re excited about how we can help develop a deeper relationship between users and AI,” he explained. “To achieve this, the AI character needs to have short term and long term memory like Inworld’s characters do so that conversations can be deeper and more personalized.” 

MageCosmos plans to add new features and functionalities to their game world that build on these capabilities. “We ultimately want users to be able to effortlessly create and train bespoke AI characters,” he explained. “Imagine a world where virtual reality is enhanced by intelligent characters you’ve personally tailored.” 

Impact

MageCosmos is in the early stages of implementing Inworld but they’re already seeing promising results. 

“In the recent game test, many testers showed a strong interest in the AI NPCs,” explained Ryan. “The average game time increased by 5% due to the interaction with AI.” 

Ryan expects MageCosmos’ average game time to increase even more once they optimize the characters and experience. “We’re excited about Inworld,” he said. “We never originally planned to add AI NPCs but Inworld has made our game more alive than ever and our players like it.”

For Ryan, this is just the beginning of what he believes will be an industry-wide shift. “I think there is going to be a revolution in companionship in game worlds,” he added.

Try out AI NPCs today

Want to add AI NPCs to your game? You can be up and running in minutes using Inworld Studio. For large or custom implementations, reach out to Inworld’s Partnership Team.

Ready to get started?

Try for free

Read documentation