Tatiana Tantri Indah Sari
The Skift Data + AI Summit recently convened industry leaders to move beyond the superficial discussions of artificial intelligence adoption, instead delving into the profound chasm between ambitious AI strategies and their often-stark operational realities within the travel sector. Rafat Ali, CEO & Founder of Skift, set the tone for the day, emphasizing that while "everyone in this room has a strategy for AI," the efficacy of these strategies remains a significant, open question. The summit aimed to address the friction points where corporate ambition collides with practical implementation, where the imperative for speed challenges the necessity of trust, and where established business models face tectonic shifts driven by AI.
The core premise of the summit was to confront the "say-do gap" in AI, a phenomenon where organizations often possess detailed plans and impressive slide decks but lag significantly in real-world production. This disparity, Ali noted, is where the most critical decisions reside, shaping the future trajectory of travel companies. The event was meticulously designed around five pivotal tensions, each serving as a focal point for sessions, supported by disquieting data, and framed by probing questions for speakers. These tensions, while lacking facile solutions, underscore the urgent need for leaders to cultivate the judgment required to act decisively in the face of uncertainty – a central theme of the day’s proceedings.
A Deliberate Design for Decision-Making
Breaking from traditional conference formats, the Skift Data + AI Summit structured its agenda around specific decisions rather than broad topics. Moderators were tasked not merely with introducing speakers but with articulating the precise tension confronting the audience, extracting unique insights from panelists, and synthesizing their practical implications. The day unfolded in three distinct acts, mirroring the lifecycle of AI implementation within an enterprise:
- Act I: Build the Stack – Focused on the foundational requirements for production-grade AI, examining whether existing infrastructure is robust enough to support ambitious AI initiatives. This segment probed the technical backbone necessary for scaling AI effectively.
- Act II: Run the Business – Explored the operational challenges encountered when AI pilots transition into full-scale organizational integration. Key questions revolved around what aspects of the business break during this transition and, crucially, who assumes ownership of this complex transformation.
- Act III: Drive the P&L – Addressed the ultimate business impact of AI, scrutinizing how artificial intelligence contributes to revenue generation and cost optimization, and enabling leaders to discern tangible financial returns.
Throughout these three acts, the five overarching tensions were interwoven, ensuring that fundamental questions—such as the perennial debate between pilots versus production, speed versus trust, and build versus buy—were revisited and answered from diverse perspectives. This intentional friction, featuring operators at varying stages of AI maturity and with differing strategic bets, aimed not to forge a consensus but to challenge assumptions and recalibrate expectations, fostering a more nuanced understanding of AI’s practical deployment.
Unpacking the Five Tensions Driving Travel AI
The summit’s agenda was a rigorous exploration of the challenges and opportunities presented by AI, with sessions designed to dissect each of the identified tensions. The discussions underscored the complexity of AI integration, highlighting the strategic imperatives facing the travel industry’s leadership.
Tension 01: Pilots vs. Production
One of the most stark revelations concerned the disparity between AI pilot projects and successful production deployments. Carnival Cruise Line’s experience of 100 pilots yielding only 6 in production is not an anomaly; MIT research corroborates this trend, reporting that 95% of organizations realize zero ROI on generative AI projects. This data paints a clear picture: the skills and methodologies that foster early-stage innovation (curiosity, rapid prototyping, minimal governance) often diverge sharply from those required for scalable, production-ready AI.
Consequently, a significant majority of travel organizations have yet to meaningfully scale AI. Skift Research indicates that fewer than one-quarter of industry companies have widely adopted generative AI, with a mere 2% reporting similar progress for agentic AI. This "say-do gap" represents a critical competitive battleground. Firms that master the transition from experimental pilots to robust production systems will gain a compounding advantage, potentially leaving competitors behind. Sessions within "Act I – Build the Stack" directly addressed this, with Hilton discussing "The Stack You Have Won’t Scale What You’re Building" and Marriott + Google Cloud exploring "How Are Leaders Using AI to Solve Real-World Problems?" These discussions focused on the architectural and operational shifts required to move from proof-of-concept to enterprise-wide deployment, including considerations for data governance, security, and integration with legacy systems. Startups like Flight Science, BizTrip AI, and TakeUp also presented their solutions, highlighting areas where agile innovators are bridging gaps that larger entities might overlook.
Tension 02: Speed vs. Trust
The imperative for rapid AI deployment often clashes with the fundamental need to build and maintain traveler trust. A sobering statistic reveals that only 2% of young leisure travelers are willing to let AI book their trips independently. This low level of trust, even among digitally native millennials and Gen Z, underscores a significant hurdle. AI applications in pricing, customer service, and personalization are inherently reliant on consumer confidence. However, the pressure to innovate quickly can lead to visible governance gaps, eroding credibility and fostering skepticism.
Industry players are already adjusting strategies, pulling back on fully autonomous AI, integrating human oversight, and re-evaluating what "ready" truly means when speed and trust pull in opposite directions. The debate "Should AI Agents Control the Customer Journey?" featuring Cloudbeds, directly tackled this tension. Later, Booking.com’s "AMA: Running AI Like Infrastructure" session likely touched on how a global giant navigates this delicate balance, especially given the low percentage of travelers willing to delegate booking decisions entirely to AI. The implications for brand reputation and customer loyalty are immense; a misstep in AI deployment could have long-lasting negative consequences. Maintaining transparency about AI’s role and ensuring robust error-handling mechanisms are becoming paramount for companies seeking to leverage AI in customer-facing functions.
Tension 03: Restructuring vs. Readiness
The travel industry is reorganizing work around AI at a pace that far outstrips its efforts to prepare its workforce for this transformation. Only one in five travel and tourism workers believe their organization is adequately prepared for AI adoption. Data from BCG and NYU further highlights this skills deficit, showing that only 2.9% of travel workers possess AI skills—the widest gap across any major industry. Compounding this, a significant 36% of hotel and airline employees report receiving no formal AI training whatsoever.
While some travel companies have initiated AI upskilling programs, these successful instances required substantial, deliberate change management investments—resources that most organizations have yet to allocate. AI adoption, therefore, increasingly resembles a comprehensive organizational transformation rather than a mere technological rollout. Simultaneously, CFOs are citing AI in earnings calls as a justification for headcount reductions, seeking tangible proof of efficiency in profit margins. This creates a precarious situation where the human element of change management, if neglected, could undermine the very acceleration of AI it aims to achieve. The session "The AI Ownership Problem Nobody Has Solved Yet," featuring Amex GBT and TUI, directly addressed the complex human capital implications of AI integration, including defining new roles, reskilling existing teams, and managing employee anxieties about job displacement. Executive Guided Roundtables and the "Roundtables: What We Heard" segment by Skift Research further provided platforms for leaders to share their challenges and nascent solutions in managing this critical human-technology interface.
Tension 04: Build vs. Buy
The fundamental strategic decision of whether to "build" AI capabilities in-house or "buy" them from external vendors presents different implications depending on an organization’s scale and strategic objectives. Expedia’s decision to hire a Google AI VP to build extensive in-house capabilities contrasts sharply with Hyatt’s adoption of ChatGPT Enterprise. For global platforms like Expedia, owning more of the AI stack can offer advantages in terms of better margins, clearer differentiation, and greater operational flexibility. Conversely, for others, an ambitious build strategy can lead to distraction, escalating costs, and project delays.
A critical unanswered question for many companies is discerning which AI capabilities are genuinely differentiating and which are effectively commoditized. The optimal build-vs.-buy decision is often fluid, shifting with a company’s position on the ever-evolving AI maturity curve. The summit provided a unique forum for leaders who have made divergent bets to stress-test their strategies against peer insights. For instance, "What Startups Are Solving That Others Won’t" highlighted innovative solutions that could be acquired, while "What Investors Are Actually Funding in Travel AI" provided a perspective on market trends influencing this decision. Companies like Travelport, with its "Brand Talk," would likely present its value proposition for "buying" solutions that integrate seamlessly into existing travel ecosystems, offering a counterpoint to the "build" philosophy.
Tension 05: Control vs. Visibility
AI is fundamentally reshaping how travelers discover and book trips, creating a sharp tension between protecting existing direct booking channels and embracing new AI-driven distribution paradigms. The extensive use of AI in trip planning has more than doubled, challenging traditional travel brands that have spent years optimizing their direct booking experiences. AI systems do not typically direct users to a brand’s website for browsing and comparison; instead, they generate a curated shortlist based on their ability to find, parse, and trust available data. If a company’s data is not structured to be machine-readable by AI systems, it risks becoming invisible in this new recommendation-driven landscape.
This shift means the new "shelf space" is less about search engine rankings or OTA placement and more about whether AI can interpret a brand’s offerings clearly enough to recommend them. The industry is transitioning from a "search/scroll/compare" model to an "ask/shortlist/decide" paradigm. The tension is stark: brands must either expose their data and restructure operations to capture AI-driven traffic or risk protecting their first-party booking funnel at the cost of diminished visibility. Priceline’s "Data Lessons from Deploying Consumer-Facing AI" session likely delved into the intricacies of optimizing data for AI systems and managing the balance between direct engagement and AI-mediated discovery. Mews, with its focus on "AI + Revenue: Why Architecture Is Essential," also contributed to this discussion by highlighting how underlying data architecture dictates a brand’s ability to participate effectively in an AI-first distribution ecosystem.
Navigating the Future: Judgment Over Roadmaps
The Skift Data + AI Summit concluded with a clear message: the "say-do gap" in travel AI is not only real but widening. While some companies are genuinely innovating at the model and orchestration layers, many more find themselves somewhere between merely reorganizing existing structures and simply renaming old initiatives with AI buzzwords—often without a clear understanding of where they truly stand.
Rafat Ali’s closing remarks reiterated that the summit’s purpose was not to provide a prescriptive roadmap. Instead, by presenting concrete examples—such as how Expedia manages the deployment of its 1,500 AI agents, how Booking.com addresses the trust deficit among travelers, and how Marriott and Google Cloud define production-grade AI infrastructure—the summit aimed to cultivate something more profound: the judgment necessary for leaders to accurately assess their organization’s position and make informed, decisive next steps. Attendees were challenged to leave not just with an understanding of the five prevailing tensions but with the clarity to identify which one demanded their immediate and focused resolution. The journey through AI in travel is less about finding easy answers and more about building the strategic acumen to act effectively without them.
Dwi Wanna
Walid Sujud Trenggono
Chandra Hamzah
