Enterprise gen AI pipelines in 2025: Build or buy for scaling?

The core challenge: Enterprises in 2025 face complex decisions about scaling generative AI, moving beyond simple deployment to focus on operational transformation and cost optimization.

• Companies must balance the excitement of AI adoption with practical challenges of efficiency, cost management, and market competitiveness
• Success requires clear use cases, technological flexibility, and an adaptable workforce
• The focus has shifted from experimenting with AI to creating sustainable, enterprise-wide implementations

Leading examples: Wayfair and Expedia demonstrate successful hybrid approaches to large language model (LLM) adoption, combining external platforms with custom solutions.

• Wayfair utilizes Google’s Vertex AI for general applications while developing proprietary tools for specific needs like product tagging
• Expedia implements a multi-vendor LLM proxy layer for seamless integration of various models
• Both companies emphasize the importance of matching technology choices to specific business requirements

Operational implementation: Companies are finding success by targeting specific high-value applications rather than pursuing broad, unfocused AI adoption.

• Wayfair leverages AI to enhance product catalogs and analyze legacy database structures
• Expedia has integrated AI into customer service, achieving 30-second response times for 90% of travelers
• Both organizations focus on measurable impacts and clear business outcomes

Infrastructure considerations: Hardware and infrastructure choices play a crucial role in sustainable AI scaling.

• Cloud infrastructure currently dominates, with both companies primarily relying on major providers
• A proxy layer system helps balance performance with cost efficiency
• Companies are evaluating future needs for localized infrastructure to handle real-time applications

Organizational readiness: Technical deployment represents only part of the scaling challenge; cultural adaptation and governance are equally critical.

• Comprehensive training programs ensure employees can effectively work with AI tools
• Expedia’s Responsible AI Council oversees major AI decisions and ensures ethical implementation
• Traditional KPIs are being rethought to better measure AI effectiveness

Looking ahead to future challenges: The enterprise AI landscape continues to evolve, requiring ongoing adaptation and strategic planning.

• Inference costs and real-time processing capabilities remain key considerations
• Organizations must balance innovation with practical implementation
• Success requires continuous evaluation of use cases and investment priorities to maintain competitive advantage

Strategic implications: In 2025’s AI landscape, organizations that successfully scale their AI operations will be those that maintain flexibility while targeting specific, high-value use cases rather than pursuing broad, unfocused implementations. The experiences of Wayfair and Expedia suggest that a hybrid approach, combining custom solutions with external tools, offers the most promising path forward.