How Cornerstone is transforming XR training to build essential skills for the modern workforce

Cornerstone, a global leader in workforce agility platforms, helps companies provide multimodal learning experiences through advanced technologies like augmented and virtual reality. Their solutions have helped to engage top talent and bridge skills gaps at more than 7,000 organizations, accounting for approximately 140 million users. Available in 186 countries, Cornerstone’s Galaxy platform leverages AI to build future-ready organizations and foster professional development.

Blending immersive learning and AI to present scenarios that people might encounter in the physical world, Cornerstone is transforming how employees learn. Using Llama, they’re taking that experience to the next level through extended reality (XR) training scenarios, resulting in improved learner and admin experiences across their suite of products.

Cornerstone currently uses the Llama 3.1 8B model as a foundational component of its platform to power tools like Content Studio and Learning Object Creation for Cornerstone Learning Products, as well as various other use cases involving retrieval-augmented generation (RAG) to improve the usability of their products for their global customer base.

“Cornerstone has more than 25 years of experience and a long history of investments in AI,” says Karthik Suri, Chief Product Officer at Cornerstone. “When we started our effort on building an LLM-agnostic, open platform that will power our future generative AI use cases, Llama and its technology ecosystem align perfectly with our ethos of providing ethical, responsible, and transparent AI.”

Leaning into the growing trend of delivering content via XR to drive accelerated learning outcomes, Cornerstone’s new Immerse Companion leverages Llama 3.1 8B as a foundational model to create live training simulations populated with virtual human companions, including coaches, mentors, and role-playing characters. These simulations provide highly contextualized, scenario-based, and flexible training that immerses employees in real-world situations, helping them develop essential skills in areas like sales, leadership, customer service, compliance, and HR.

Despite the promise of XR training, Cornerstone faced challenges with traditional approaches to creating these simulations for Immerse Companion.

“We wanted to develop an AI-native solution that would simplify production while providing a more diverse and realistic range of character behavior,” says Kyle Jackson, VP of Product for AI and Spatial Computing.

Generating realistic and accurate LLM-based outputs requires additional engineering and advanced workflows to manage complex data inputs while also giving learning and development professionals the controls they need to ensure they can deliver results that the business needs. Traditionally, production timelines for XR simulations could stretch up to five months, making it difficult for organizations to keep training content timely and relevant. Additionally, virtual characters often lack the contextual understanding and natural conversational abilities to engage learners effectively.

Cornerstone’s use of Llama’s multilingual capabilities became an important innovation. These capabilities allow training simulations to be seamlessly translated into over 40 languages, meeting the needs of Cornerstone’s global clients.

Cornerstone fine-tuned Llama 3.1 using its vast learning and skills data repository to power experiences across their various products. With Immerse Companion increasing accuracy and enabling the model to control downstream tasks like animation, text-to-speech, and reflections took careful prompt engineering and developing repeatable syntax for the outputs. Additionally, RAG workflows will soon provide real-time context by injecting information from organizational strategies, product documentation, or hiring data.

“Managing our overall wide range of use cases means that our software stack is complex, and our customers require extremely tight security controls for their content, yet expect powerful RAG and other capabilities, which Llama helps us to deliver,” says Hitesh Dholakia, GVP of Engineering at Cornerstone.

With the Llama-backed Immerse Creator and Companion, it now takes one person as little as 10 minutes to create a 30-minute training simulation complete with highly intelligent expert virtual characters with the capability to speak naturally. From immersive role-playing to AI advisors, these capabilities are reshaping employee training by offering ultra-personalized experiences at scale.

For Cornerstone, using an open source model like Llama ensures cost efficiency, transparency, and the flexibility to adapt its solutions to customer needs. “When we tested against other foundational models, we found Llama to be just as accurate, but it gave us the controls we needed to implement for a wide range of customer requirements,” noted Cornerstone’s Chief Technology Officer, Srini Ogireddy.

Open source plays an important role in advancing AI and its broader ecosystem. It provides Cornerstone with an essential counterbalance to the dominance of commercial interests. Their customers expect them to be a trusted and agnostic source of innovation. Open source systems give them the flexibility they need to deliver on that promise. An open and transparent ecosystem enables developers and organizations to build and innovate collaboratively, avoiding the risk of skewed priorities that can arise from proprietary systems.

“Open source models like Llama are supported by a large community of developers, which fosters innovation and ensures greater transparency,” adds Ogireddy. “The gap between premium and commercial models has narrowed significantly, especially when addressing the specific use cases we’re solving for in our products.”

The company believes the open source model is imperative to democratizing access to AI and its ecosystem. By choosing open source, organizations like Cornerstone benefit from cost efficiency, adaptability, and a shared commitment to ethical and responsible AI development.

As Cornerstone continues to innovate, the team will focus on expanding Immerse Companion with more multimodal capabilities. By incorporating vision-based AI into simulations, the company aims to unlock new use cases and further enhance the realism of XR training. The team is also experimenting with the Llama 3.2 1B and 3B models, with plans to evaluate the 11B model in the near future.

“The Llama 3B model has stood out for its lightweight yet powerful capabilities and shows much promise for the future of the immersive learning space,” says Jackson.

Learn more about Cornerstone.