Expedia Names Xavier Amatriain Chief AI and Data Officer to Lead Generative AI

Expedia Group’s hire of Xavier Amatriain as its first Chief Artificial Intelligence and Data Officer marks a decisive shift in how one of the world’s largest online travel agencies intends to compete in the coming era of generative and agentic AI. The appointment — confirmed by an Expedia Group release and reported across industry outlets — places an engineer who led core AI infrastructure at Google into the center of Expedia’s bid to turn first‑party travel data and large‑scale machine learning into a durable product and commercial advantage.

Background / Overview​

Expedia describes the new role as responsible for the company’s long‑term vision for machine learning, generative AI and data science across both consumer brands and its enterprise-facing businesses. The company says the move underscores a broader strategy: meet travellers where they plan, inspire trips with richer discovery, and reinforce search, personalization and customer service through AI. This hiring also follows a broader pattern in travel tech: OTAs and suppliers are embedding generative AI into discovery and planning tools, piloting agentic assistants that can take scoped actions on behalf of travellers, and experimenting with partnerships that push inventory and booking flows into new, AI‑driven touchpoints. Those initiatives include Expedia’s own Romie assistant and a raft of partnerships that place Expedia inventory inside third‑party AI interfaces. Independent reporting and company communications show Expedia actively piloting integrations with ChatGPT/Access Connectors, Perplexity’s Comet, Microsoft Copilot Actions and Google’s AI ecosystem.

---

Who is Xavier Amatriain — and why does he matter?​

A short professional sketch​

Xavier Amatriain brings two decades of experience building and scaling AI systems at major technology companies. Expedia’s announcement lists his most recent role as Vice President of AI and Compute Enablement at Google, where he led teams responsible for AI platforms and infrastructure used by products including Gemini, Search and Ads. Before Google, he held senior engineering positions at LinkedIn and Netflix, and he’s been connected to early teams behind startups that became influential in the AI ecosystem.

Why his resume fits Expedia’s stated goals​

The job Expedia created is explicitly about operationalizing AI at scale: platform architecture, data governance, model deployment, inference pipelines and cross‑product integration. Hiring an engineer who oversaw core AI infrastructure at Google signals Expedia wants the job done at the infrastructure and systems level — not just at the level of product experiments or point solutions. That matters because the economic and competitive value of travel data comes from linking real‑time inventory, user signals, payments, loyalty and personalization into models that can reliably influence bookings and customer satisfaction across millions of searches and reservations.

---

What Expedia is asking its first chief AI officer to do​

Expedia’s public announcement is deliberately broad but concrete in its aims: define a technical direction where advanced ML and generative AI create business value at scale, align AI efforts across product, B2B, marketing and operations, and build the technical foundations for “AI‑first” experiences. More specifically, the role will likely encompass:

• Architecting inference and training platforms that handle high‑volume, low‑latency travel queries.
• Building retrieval‑augmented generation (RAG) and grounding systems to reduce hallucinations in travel contexts.
• Harmonizing first‑party data from bookings, loyalty, and click/behavioral signals into privacy‑aware models.
• Operationalizing agentic assistants with clear consent, audit trails and human‑in‑the‑loop checkpoints.

Expedia’s framing stresses scale and first‑party datasets as a defensive and offensive asset: build unique models that leverage proprietary inventory and customer signals in ways that third‑party platforms cannot easily replicate.

---

Where Expedia already is on AI: Romie, Comet, ChatGPT and Copilot integrations​

Romie and EG Labs​

Expedia has been publicly experimenting with Romie, a roaming AI travel assistant that can join group chats, summarize plans, draft itineraries and help when plans change. Romie has been available in alpha through Expedia’s EG Labs and is an example of a hybrid model: conversational discovery plus direct links into the booking funnel. Romie’s scope — planning, shopping, booking and in‑trip support — illustrates the company’s intent to turn discovery into action inside an assistive interface.

Comet (Perplexity) and OpenAI/ChatGPT integrations​

In 2025, Expedia expanded its approach: a partnership with Perplexity to launch Comet, an AI travel browser that bundles conversational search and booking, and a formal integration with OpenAI through Access Connectors that enables an Expedia app inside ChatGPT. Expedia has also worked with Microsoft Copilot Actions to embed booking flows into Copilot experiences. These moves show a dual strategy: ship owned experiences (Comet, Romie) while also surfacing Expedia inventory inside powerful third‑party AI endpoints — an important distribution hedge.

What that means in practice​

• Consumers can begin planning in a generative AI chat and be routed to an Expedia booking flow or completed by an Expedia agentic interface.
• Expedia’s inventory appears in third‑party AI assistants, creating new demand channels but also increasing exposure to potential disintermediation.
• Combinations of in‑house assistants and partner integrations allow experimentation with different permission models, data residency and monetization levers.

---

The technical work behind the promise — and the hard problems​

Turning headlines into robust customer value requires solving several non‑trivial engineering and product problems:

• Grounding and provenance: Travel details are operationally sensitive. The industry demands RAG systems that cite authoritative providers (airline APIs, hotel availability feeds) and clearly show “last verified” timestamps to limit hallucinations and operational failures.
• Latency and scale: Search and booking flows require sub‑second response times across millions of daily queries; model inference must co‑exist with real‑time inventory checks and transactional integrity.
• Data governance and privacy: Many consumer AI services log and retain interactions; enterprise contracts and product tiers must define non‑training options and clear PII handling to meet enterprise and regulatory expectations.
• Agent permissions and auditability: Agentic assistants that can book, cancel or change reservations must implement fine‑grained consent UIs, secure credentialing and auditable action logs to manage liability and disputes.
• Commercial alignment: Integrations with platforms like ChatGPT and Google can drive demand but also alter economics; Expedia must ensure margins, attribution and partner terms align with long‑term strategy.

These are precisely the sorts of cross‑disciplinary systems problems that a leader with large‑scale platform experience is expected to navigate — tying model research to engineering execution and commercial outcomes.

---

Competitive dynamics: why OTAs need both product and platform plays​

The paradox of partnering with potential competitors​

Expedia’s strategy of embedding inventory into third‑party AI systems (ChatGPT, Perplexity, Copilot, Google) is pragmatic: reach consumers where they begin discovery. But it also introduces the risk that those platforms evolve proprietary booking capabilities that bypass traditional OTA flows, capturing margin or routing users to direct supplier sales.

• Platforms that control the search surface (Google, ChatGPT, Copilot) can make discovery seamless and may be tempted to integrate payment or booking capabilities that favor owned inventory or partners with privileged placement.
• OTAs’ defense is to offer superior experience, personalization and loyalty benefits that are portable across platforms — and to build their own action‑capable AI products that lock user intent and payment into Expedia’s control.

What winning looks like for Expedia​

• Offer actionable, trusted AI experiences that reliably convert discovery to booking while preserving user control and clear consent.
• Leverage first‑party loyalty and transactional data to personalize in ways a search provider cannot replicate without bilateral data sharing.
• Build flexible partnerships with clear commercial terms to monetize referral and agentic flows without ceding control of the end‑to‑end customer relationship.

---

Risks, governance and the regulatory horizon​

The travel use case highlights many of generative AI’s most pressing regulatory and governance questions:

• Consumer protection and liability: If an agentic assistant books a trip that fails due to stale availability or incorrect visa advice, who is accountable? Platforms, suppliers and OTAs will need to negotiate responsibility and remediation mechanisms.
• Privacy and model training: Many consumer models train on user interactions. Enterprises and high‑value travellers will demand contractual non‑training guarantees and data residency controls before routing PII through external models.
• Competition policy: Search and AI platforms that control discovery may attract regulatory scrutiny for gatekeeper behavior if their AI favors certain partners or channels.
• Operational safety: Travel advice has safety implications — incorrect health, timetable or entry requirements can cause harm. AI outputs must be grounded and include human checks for safety‑critical recommendations.

Industry analysts and internal trade research urge explicit audit trails, provenance displays, opt‑in agent permissions and clear consent flows as the minimum guardrails for agentic travel assistants. These are not theoretical cautions; travel firms already document instances where AI‑driven errors create tangible customer harm and reputational costs.

---

Practical implications for Expedia customers, partners and rivals​

For travellers​

• Expect more conversational trip discovery and itinerary drafting in apps and chat interfaces, with quicker ideation and personalized suggestions.
• Reliance on AI for final bookings will come with friction: explicit consent screens, confirmation emails and possibly higher verification for agentic actions.
• For now, critical facts (visa, health requirements, and flight connections) still need human verification when they affect safety or legal compliance.

For hotels and suppliers​

• Suppliers must make machine‑readable data available (structured APIs, schema.org markup, seat maps) so agentic assistants return accurate availability.
• Partners should audit cancellation policies and dispute flows; AI‑driven discovery increases the need for frictionless customer remediation.
• Distribution economics may shift; hotels and airlines should monitor attribution and renegotiate terms if agentic platforms materially alter referral or booking volumes.

For industry competitors and platforms​

• Expect a two‑track arms race: platform owners will push for action‑capable AI that keeps users in their ecosystems, while OTAs will invest in proprietary AI experiences, loyalty hooks and commercial mechanisms to capture booking value.
• Strategic partnerships (like Expedia’s Comet partnership with Perplexity and ChatGPT integration) are a pragmatic way to participate in multiple ecosystems while maintaining direct customer touchpoints.

---

How to measure success — short‑term metrics and longer bets​

Success for a hire of this nature should be measured across both product and business KPIs.
Short‑term measurable indicators:

• Conversion lift on AI‑driven journeys (discovery → booking conversion rate).
• Reduction in support costs via AI‑assisted service automation.
• Uptake rates and retention for AI features in EG Labs and staged rollouts.
• Error and hallucination rates in production flows (issues per 10k sessions).

Longer‑term strategic indicators:

• Share of bookings sourced through Expedia’s owned AI experiences (Comet, Romie) versus third‑party AI channels.
• Retention and incremental revenue from loyalty members who engage with AI experiences.
• Commercial terms and margins on bookings routed from third‑party AI platforms.
• Regulatory and risk posture: number of incidents requiring remediation and the robustness of provenance, audit trails and dispute resolution.

---

What remains unverified or speculative​

Several public claims and projections about AI in travel are plausible but not yet fully verifiable in the public record. These should be treated with caution:

• Any headline ROI figures quoted in vendor announcements (e.g., “X% uplift” in conversions) are often pilot metrics that do not generalize without full implementation details and independent validation.
• The trajectory from pilot to global, production‑grade agentic booking at scale contains many operational hurdles: reliable grounding, multi‑party contracts, payment security and fraud prevention.
• Specifics about which models train on traveller interactions and which contract tiers include non‑training guarantees vary by vendor and region; where contractual non‑training clauses matter, buyers must secure them in writing.

When such statements are material to business decisions, the prudent path is to request independent audits, pilot data, and contractual guarantees rather than rely solely on vendor PR.

---

Early verdict: a strong hire, but execution is the hard part​

The appointment of Xavier Amatriain gives Expedia a credible and experienced leader to consolidate AI, compute and data efforts at a critical juncture. The hire signals that Expedia understands AI is not just a set of product features but an engineering and systems challenge that touches infrastructure, data governance, legal and commercial strategy. Bringing in a leader with a history of building foundational AI platforms is therefore consistent with the company’s stated ambition. However, the ultimate test will be execution: turning Romie‑style pilots and third‑party integrations into reliable, profitable, and safe production experiences that maintain Expedia’s margins and customer trust. That requires a sustained investment across engineering, partnerships, legal and customer service teams, plus a careful stance on data contracts and agent permissions. The travel industry’s “agent era” promises new convenience; it also raises fresh questions about accountability and market power that Expedia must address proactively.

---

What to watch next​

• Product: how quickly Romie and Comet move from alpha to broad availability, and whether Expedia layers booking permissions into those experiences without sacrificing safety.
• Partnerships: whether Expedia negotiates stronger commercial safeguards as its inventory appears inside ChatGPT, Copilot and Google AI Mode.
• Regulation: any policy actions that clarify liability for agentic commerce or impose provenance requirements on AI search and booking assistants.
• Metrics: published pilot results or independent audits that quantify conversion, error rates and customer satisfaction on AI‑driven journeys.

These milestones will indicate whether Expedia’s strategy — invest in platform scale, own customer intent and distribute through partnerships — is converting technical leadership into sustainable commercial advantage.

---

Practical takeaways for travel technologists and IT leaders​

• Prioritize provenance and RAG frameworks for any travel‑facing AI; users and regulators will demand clear sources for operational facts.
• Negotiate non‑training and data residency clauses for PII and enterprise flows before enabling third‑party model integrations.
• Design agentic features with explicit consent, scoped permissions, and audit logs to reduce liability and increase user trust.
• Treat energy and compute economics as a first‑order design parameter: inference costs at scale matter for margins and product pricing.

---

The hiring of Xavier Amatriain is a meaningful signal: Expedia is committing to unify platform engineering, data science and product AI under an executive with the scaffolding experience required to execute at scale. The strategic challenge ahead is not simply engineering models but embedding them into trustworthy, auditable, and commercially aligned travel experiences — a task that will determine whether the next wave of travel AI amplifies the OTA’s role or hands new parts of the customer journey to platform gatekeepers.

---

Source: travelweekly.com.au Expedia raids Google for first chief AI officer