Build or buy? Expanding your gen AI pipeline in 2025

This article is part of a VentureBeat special, “AI at Scale: From Vision to Viability.” Read more about this story here.

Expanding the implementation of production equipment it has always been difficult to balance ambition and action, and in 2025, the stakes are higher than ever. Businesses rushing to adopt master languages ​​(LLMs) are facing a new reality: Scaling up isn’t just about using big brands or investing in advanced tools — it’s about incorporating AI into ways to transform operations, empower teams and manage costs. Success depends on more than technology; it calls for cultural and operational changes that align the power of AI with business objectives.

Greater importance: Why 2025 is different

As artificial AI transitions from experimentation to business, businesses are facing a challenge. The excitement of the initial launch has given way to the challenges of working efficiently, managing money and ensuring the markets are running smoothly. Augmenting AI in 2025 is about to answer some tough questions: How can businesses make artificial intelligence tools useful across departments? What infrastructure will support the growth of AI without silencing devices? And perhaps most importantly, how do teams adapt to AI-driven mobility?

Success depends on three key principles: identifying clear, usable values; maintaining technological flexibility; and motivating employees who are willing to change. Successful businesses don’t just adopt gen AI – they create strategies that align technology with business needs, and review the cost, performance and culture changes required to drive change. This approach is not limited to modern equipment; it’s about creating operational efficiencies and flexibility in an environment where technology and markets change rapidly.

Companies like Wayfair and Expedia includes these studies, showing how hybrid approaches based on LLM can improve performance. By combining external platforms with bespoke solutions, these businesses demonstrate the power of balancing agility and precision, setting an example for others.

Combining customization and flexibility

The decision to build or buy gen AI tools is often presented as a binary one, but Wayfair and Expedia are showing the advantages of a non-conventional approach. Fiona Tan, Wayfair’s CTO, emphasizes the importance of balancing flexibility and stability. Wayfair uses Google Vertex AI for general use while developing proprietary tools for niche requirements. Tan shared the company’s iterative strategies, sharing how the smaller, cheaper brands often win out over the bigger, more expensive ones when it comes to branding products such as fabric and furniture brands.

Similarly, Expedia uses a multi-level LLM system that allows for a seamless mix of different types. Rajesh Naidu, Executive Vice President of Expedia, describes their strategy as a way to improve revenue. “We always take a chance, we look for the best brands (models) that make sense, but we’re also willing to do it ourselves,” Naidu explains. This flexibility ensures that the team can adapt to business needs without being locked into a single vendor.

Such hybrid approaches recall the evolution of Enterprise Resource Planning (ERP) in the 1990s, when businesses had to choose between adopting rigid, external systems and adapting systems to fit their operations. Then, as now, successful companies realized the need to combine external resources with sustainable development to overcome specific challenges.

Effective business start-up services

Both Wayfair and Expedia point out that the real strength of LLMs is in the programs they focus on that provide measurable impact. Wayfair uses artificial intelligence to enrich the product catalog, improving metadata accuracy independently. This not only improves workflow but also improves customer search and customer experience. Tan also highlights another revolutionary project: using LLMs to analyze ancient artifacts. With original machine manufacturers no longer available, gen AI helps Wayfair reduce technical debt and uncover new innovations for entry-level machines.

Expedia has successfully integrated gen AI into customer service and improved logistics. Naidu shares that an AI-based tool designed to summarize calls ensures that “90% of travelers can reach an agent within 30 seconds,” which helps improve customer satisfaction. In addition, GitHub Copilot has been used around the world, speeding up code creation and development. These operational benefits emphasize the importance of aligning gen AI capabilities with clear, valuable business use cases.

The role of hardware in gen AI

Hardware considerations for LLM upgrades are often overlooked, but play an important role in long-term stability. Both Wayfair and Expedia currently rely on cloud infrastructure to power their gen AI operations. Tan says Wayfair continues to evaluate the growing number of cloud providers like Google, and monitor the demand for data that may be needed to make real-time use more efficient.

Expedia’s approach also emphasizes flexibility. Treated especially at AWSThe company uses a proxy layer to forward the service to the most appropriate computing environment. The system balances performance and cost control, ensuring that inventory does not increase. Naidu points to the need for this evolution as AI-powered business applications grow in complexity and require more control.

This broad overview reflects many of the enterprise computing trends, reminiscent of the transition from data centers to microservices architectures. As companies like Wayfair and Expedia expand their LLM capabilities, they demonstrate the need to balance cloud scalability with emerging options such as edge computing and traditional chips.

Education, leadership and change management

Posting LLMs isn’t a technical problem – it’s a cultural one. Both Wayfair and Expedia emphasize the importance of strengthening organizational readiness to adopt and integrate gen AI tools. At Wayfair, comprehensive training ensures that employees in all departments can adapt to new functions, especially in areas like customer service, where AI-generated solutions require human supervision to match the company’s voice and tone.

Expedia has taken the initiative further by establishing a Responsible AI Council to oversee all major decisions related to gen AI. This council ensures that deployments are aligned with ethical guidelines and business objectives, building trust across the organization. Naidu emphasizes the need to rethink metrics to measure the success of gen AI. Traditional KPIs often fail, prompting Expedia to adopt precision and remember strategies that best align with business goals.

These cultural changes are critical to gen AI’s success in the business environment. Technology alone cannot drive change; change requires a workforce that is equipped to use gen AI technology and a management system.

Lessons in effective development

The Wayfair and Expedia events provide valuable training for any organization seeking to develop LLMs effectively. Both companies show that success depends on identifying common business cases, maintaining flexibility in technology decisions, and fostering a culture of adaptability. Their hybrid approach provides an example of combining creativity and innovation, ensuring that gen AI investments produce tangible results.

What makes developing AI in 2025 an unprecedented challenge is technological and social change. The hybrid, flexible, and robust data systems that define today’s AI deployments will lay the foundation for new innovations. Businesses that build this infrastructure don’t just develop AI; they will develop resilience, flexibility, and competitive advantage.

Looking ahead, cost issues, real-time capabilities and development needs will continue to shape the business landscape of AI. As Naidu clearly states, “Gen AI and LLMs will be a long-term investment for us and have set us apart on the journey. We must remember that this will require prioritizing investment and understanding the use cases.”