Marriott and Google Discuss AI Production Challenges

At the Skift Data + AI Summit 2026, Colin Coleman, SVP - Enterprise Data, Analytics, and AI at Marriott International, and Joff Romoff, Global Head - Travel & Hospitality at Google Cloud, participated in an Ask Me Anything moderated by Seth Borko, Head - Research at Skift, per Skift.

According to Skift, both speakers framed the primary enterprise AI challenge as getting pilots into production, saying people and workflow constraints, not technology, most often block that jump. Skift records the following evidence quoted from the session, attributed to Romoff and Coleman:

• •Romoff said a large European hotel company cut call response time by about 75% through conversational call-data automation.
• •Romoff said Air France reduced development time about 35% by using Google Cloud's Gemini to centralize fragmented data sources.
• •Romoff said the average travel prompt is about 15 words.
• •Coleman said, "This isn't about automation," and added that automating an existing task can lock in broken workflows rather than re-engineer for desired outcomes.

Companies attempting to move GenAI pilots into production commonly face nontechnical bottlenecks such as messy, distributed data sources and workforce practices that assume legacy task boundaries. Centralizing data and building repeatable data pipelines are frequent prerequisites before model or prompt engineering delivers measurable ROI.

Reporting in the session highlighted travel-sector variation: Skift notes cruise and aviation are already converting AI into revenue, while lodging's consumer-facing layer, including Marriott's conversational search work, is still arriving. Industry-pattern observations indicate sectors that standardize partner data flows and expose operational metrics early tend to realize faster, measurable gains from ML projects.

For practitioners: track adoption of centralized data fabrics, metrics like call-response time and development-cycle reduction, the evolution of prompt lengths in travel use cases, and organizational investments in workflow redesign and skilling rather than purely model tuning. Observers should also watch production use cases that embed AI into multi-stakeholder processes across booking, operations, and customer service.