Expedia AI Principles Focus on Business Outcomes, Not Just Models
Expedia is shifting its AI development focus towards scalable, reliable systems rather than purely predictive models. The company has established principles to ensure AI systems create value and operate safely.
Expedia is formalizing its approach to artificial intelligence (AI) and machine learning (ML) development by publishing a set of guiding principles. This initiative marks a strategic shift towards building scalable and dependable AI systems, particularly as agentic AI becomes more prevalent in decision-making roles.
The core objective of these principles is to ensure that AI solutions deliver tangible business value, can scale across different teams and use cases, and achieve continuous improvement. A significant emphasis is placed on safety and accountability, especially as AI increasingly makes decisions on behalf of travelers.
To facilitate this, Expedia has introduced 'Agentic Release' tollgates. These are a set of recommended and sometimes required checks that must be completed before launching agentic AI features, translating broad principles like clear ownership and risk governance into specific team expectations.
Key operational guidelines include measuring AI impact against business outcomes rather than solely technical metrics, justifying complexity against strong baseline solutions, and mandating both offline and online evaluations before widespread deployment. The company prioritizes building on shared foundational systems and treating data as a product.
Expedia aims to automate these requirements and embed them within the software development lifecycle. The ultimate goal is for these expectations regarding the design, evaluation, approval, launch, and monitoring of AI systems to become an integrated part of the standard development process.