Disciplined intelligence: How Byte by Yum!™ is scaling AI at global speed

Rockford Yost, Senior Director, Data Science and AI, Yum! Brands, presenting at NVIDIA's GTC Conference on conversational AI.

The conversation coming out of ICR Conference 2026, a meeting that "Mad Money" host Jim Cramer has called "the Super Bowl of consumer conferences," was that the future of global restaurant leadership will be defined by companies that own, operate and scale intelligence as a core capability - and that winning restaurant brands will be those that shift from buying AI tools to building AI capabilities.

Ahead of the industry, Yum! Brands, the world's largest restaurant company, built Byte by Yum!, a proprietary AI-driven restaurant technology platform that allows the company to own its technology and provide its franchisees with better, faster, cheaper and safer products.

Last year, Yum! and software company NVIDIA partnered to accelerate the development of innovative AI technologies for Yum! restaurants and to power and scale the existing Byte by Yum! platform. With the integration of NVIDIA AI software and infrastructure capabilities directly into Byte by Yum!, the partnership enabled intelligence development and scaling across restaurants.

After building, integrating and proving the Byte by Yum! platform largely in its United States-based restaurants, Yum! is now focusing on product excellence and accelerating global adoption. To assist in meeting its goals, NVIDIA and Yum!'s partnership is helping to solve for costly data and to grow globally.

Solving costly data

According to Cameron Davies, Yum! Brands chief data officer, Yum!'s system gets smarter with every order. Each one generates more data, which leads to better AI and improved user experience. However, a central obstacle to scaling AI is what he calls "costly data." Because data is created across many restaurants and systems, franchise systems are inherently complex and create challenges to data gathering and AI training.

Yum! Brands Chief Data Officer Cameron Davies speaking with NVIDIASenior Directors Andrew Sun and Joey Conway about scaling AI agents globally across brands and restaurants.

"We've taken a very deliberate approach to our AI build-versus-buy strategy, ensuring we own the most strategic elements, especially our data, while maintaining flexibility and avoiding over-reliance on any single vendor," said Davies. "NVIDIA's platform lets us generate the training data we need at scale, so our Byte by Yum! team can now build and deploy AI agents in weeks or months, not quarters or years."

Instead of starting from scratch, Yum! fine-tuned NVIDIA Nemotron open-source small language models (SLMs). SLMs are compact language models designed to deliver strong performance on targeted tasks with lower latency and compute cost than larger models. Switching to an open-source SLM gives Yum! the ability to retain its data, own its IP and AI models and improve cost efficiency and scalability of those AI models.

To overcome data collection challenges, NVIDIA helped Yum! rethink how its data is created, refined and deployed. After adopting open-source SLMs, Yum! utilized NVIDIA NeMo Data Designer to synthetically generate 20,000-plus training records, meaning that Yum! is creating data through algorithms that mimic customer orders and queries. From there, the company used NVIDIA NeMo Customizer to train SLMs to appropriately respond to customers based off the synthetic data. Lastly, Yum! employed NVIDIA NeMo Evaluator to score how successfully its AI modes are at making decisions. For example, the Digital & Technology team taught an SLM how to respond to customers ordering Crunchy Tacos and then evaluated the AI's reply to those orders, revealing a 3X performance improvement.

The AI models created from this process were used for automated ordering and customer service queries. Through this process, the SLM has tripled its accuracy in selecting the correct tool or function with the right parameters during complex, multi-step ordering processes, meaning that when a customer orders a taco, pizza, burger or chicken sandwich, they'll receive the correct item. The integration of NVIDIA showed a drastic improvement in accuracy in this technology.

This isn't incremental improvement. It's a change in performance and a validation that controlling the data layer unlocks scalable intelligence.

Global growth

To accelerate Byte by Yum!'s global implementation, the Digital & Tech team rapidly built and deployed AI agents across Yum!'s global portfolio. KFC India has enhanced its customer service by using AI agents to help customers access their order status, contact their local restaurant and delivery driver, review past orders, send feedback and more through a chatbot.

Pizza hut team member hands a customer her accurate order thanks to AI at a Plano, Texas, drive-thru.

To improve the restaurant team member experience, Pizza Hut UK, Middle East and Africa used AI agents for Byte Coach, helping team members access operational standards via a chat interface. This avoids the need for team members to search an intranet or run through restaurant operational manuals.

Worldwide, AI agents have been integrated in Yum! corporate pilots around financial intelligence and internal workflows.

This coordinated, portfolio-wide deployment is proof that AI can scale across brands, markets and functions when built on a strong data foundation.

Forward-looking partnership goals

While Yum! Brands focuses on accelerating Byte by Yum!'s global implementation and product excellence, the key to success lies in building a strong AI infrastructure from data that can be easily deployed globally.

"As the pace of technology accelerates, the ability to own data and key strategic components of restaurant technology stacks will provide a competitive advantage in the quick service restaurant industry," said Davies. "By combining Yum! Brands' scale, data and operational expertise with NVIDIA's AI platform leadership, both organizations are working to accelerate innovation and build a durable competitive advantage in an AI-driven world."