Why AI travel agents may be commerce’s final puzzle

Ixigo has positioned itself as one of India’s largest online travel agents, now wrapping its core booking platform with an AI-powered assistant called Tara that aims to understand users, assist them and complete tasks across the travel journey. In theory, that should move the experience beyond static filters and forms. In practice, the first live tests show how much work remains.

On a work trip to the Global Fintech Fest in Mumbai, with a budget set at 50,000 rupees (about US$523), Tara was asked to plan and book the travel. The assistant surfaced flights a week ahead of the event dates, missing that the fest runs from 8 to 11 September 2026 at the Jio World Centre and Trident in Mumbai. It then required the user to manually select the flight, enter personal details, choose seats and pay, with hotel options only appearing after the flight booking and needing separate payment flows.

Tara’s test: promise and limits of agentic AI

Rajnish Kumar, Ixigo co-founder and group co-chief executive, is candid about the trade-offs. He says Ixigo is not yet aiming for fully agentic discovery and booking, arguing that travellers still want to review and approve key decisions to avoid costly mistakes such as wrong flights or cancellation charges. He notes that Tara failed to infer the correct event dates because the assistant is restricted to curated sources such as Google Places and specific travel-related web searches — guardrails designed to stop the system from taking actions beyond its remit. As he puts it, giving AI more tools increases non-deterministic behaviour, which is hard to reconcile with the risk profile of high-value, non-refundable bookings.

One crisp takeaway for investors is this: AI travel agents are shifting from a user-interface experiment to an infrastructure challenge, and most of the value will accrue to platforms that can standardise messy travel data and payments at scale.

Can AI travel agents fix the industry’s plumbing?

The macro backdrop is supportive. International travellers spent a record US$2.02 trillion in 2025, according to World Travel & Tourism Council data, with travel and tourism contributing US$11.6 trillion to global GDP. Yet a McKinsey survey shows that while 90 per cent of travel executives say their organisations use generative AI in some form, only 2 per cent report that agentic AI is widely deployed across their businesses. That gap between experimentation and deployment is central to the investment story.

The constraint is less model capability than industry plumbing. Airline and hotel systems were built separately over decades, with inventory and prices that change in real time. Cancellation and modification policies often sit in plain-text formats, rather than standardised, machine-readable schemas that AI agents can parse reliably. Payments add another layer of friction, as any autonomous booking agent needs clear rules on approval, spending limits and liability before it can move beyond recommendation into execution.

Raunak Mehta, chief executive and co-founder of Singapore-based insurtech firm Igloo, argues that for AI agents to work reliably across travel, the industry needs far greater standardisation of inventory, pricing and policies, in formats that AI systems can read and act on. He points out that not every flight or hotel booking carries the same terms and conditions, which makes automated interpretation and enforcement difficult at scale.

What does the supply-access gap mean for investors?

Meanwhile, access to supply remains a gatekeeper. Airlines distribute fares via global distribution systems and New Distribution Capability pipes, both of which require formal approvals and commercial relationships. Hotels are even more fragmented, pushing AI-native start-ups towards aggregators or costly one-by-one supplier deals. Incumbent OTAs such as Ixigo face the opposite problem: stitching legacy systems together while designing conversational interfaces that are genuinely better than the forms and filters that millions of users already understand.

For investors, this structural friction is not all bad news. Global groups are already moving to capture the upside: Expedia‘s 2026 acquisition of Berlin-based AI trip-planning platform Layla aims to fuse conversational planning with deep inventory, pricing and booking rails. In India, Travelxp’s launch of Marco — an agentic AI trip concierge unveiled at Global Fintech Fest — shows how domestic players are targeting end-to-end planning, booking and payment within preset user limits, rather than stopping at itinerary suggestions.

As a result, the near-term risk lies in execution rather than demand. AI is already reshaping how travellers search and decide, often via general-purpose agents such as ChatGPT or Claude, which can plan trips but still send users back to OTAs or airline sites to transact. The platforms that can convert that intent into tightly coupled, standardised booking and payment flows should be better placed to monetise AI-driven demand. This trend mirrors broader AI investment surges across Asia, where firms are racing to move from pilot to production at scale.

For institutional investors and corporate strategists, the key signals to watch are straightforward: how fast incumbents like Ixigo push agentic features deeper into booking and payments, how far regulators and industry bodies move on data and policy standardisation, and whether end-to-end AI travel agents such as Marco can sustain their promise once they meet the full complexity of India’s travel market.

Quick answers