How AI transforms the travel value chain
The elusive search for ‘The Perfect Trip’ may finally end
There’s an open secret about competing in online travel.
Every online travel player is pursuing the elusive holy grail of managing the end-to-end trip for the traveler.
And everyone is failing.
Trips are bundles. Every trip bundles accommodation, transportation, activities etc.
The internet unbundles trips. But enduring profits in travel lie with the ability to rebundle trips.
This article looks at five key ideas:
- Rebundling is the key to enduring profits in travel, and yet today’s players across the travel value chain lack the right-to-rebundle
- AI creates entirely new opportunities for rebundling, and these opportunities follow vastly different paths in corporate travel and consumer travel
- AI changes the go-to-market strategy for today’s travel players
- For AI to effectively transform travel, we will need to move to a value chain where platforms and protocols co-exist
- Winners and losers in the future of travel
Let’s dive in!
But first, a small sidebar…
Platforms in the age of Generative AI
The core focus of my work this year is the intersection of platforms and AI.
I’m traveling on a speaking tour of the US coastal states during July and August 2024 to speak on this topic.
Current client engagements for the tour have been booked in Seattle, SF, LA, Portland, Boston, and New York.
If you’ve liked my work on this topic and would like to engage, write in to liz@platformthinkinglabs.com
Unbundling travel
Before the rise of the internet, travelers had limited information about far-out destinations. To address this, travel agents would curate package tours, bundling flights, accommodations, and other services.
The internet effectively “unbundled” trips into their constituent parts. Suddenly, travelers gained access to tons of information and the ability to book every aspect of the trip independently.
Today, online travel aggregators — Booking.com and Expedia — facilitate bookings for flights and accommodations. Platforms like TripAdvisor provide insights on lodging, dining options, and local attractions. On-demand rideshare services like Uber and Lyft offer convenient transportation solutions.
The problem with unbundling travel
But trips are still bundles.
By unbundling trips, the internet shifted the cost of rebundling back to the traveler.
As I explain in The ONDC conundrum,
The more the number of steps in the value chain you unbundle, the higher the coordination costs.
The more the constituents of a trip, the higher the overhead of rebundling.
Effectively, the internet provides choice for the traveler but increases overheads.
But that’s not the only problem with unbundling travel.
As I explain in How to win at Gen AI:
There is no sustainable value capture in unbundling. Unbundling unseats incumbents but doesn’t create scalable and defensible value pools.
That is achieved through rebundling. Rebundling involves bundling multiple unbundled capabilities into a cohesive customer-centric offering.
For most consumers, travel is both infrequent and seasonal.
- Owing to the low frequency, online travel players need to keep reacquiring the customer. As a result, Google Ads are the biggest marketing expense for most online travel players.
- Owing to the seasonality, all travel keywords get overbid during periods of high travel booking activity, making the economics even worse.
So there are really two ways to increase profits in travel:
- Target a customer base that is not infrequent and seasonal (e.g. corporate travel).
- Maximize capture of the trip so that you aid booking of not just one component of the trip but all trip components.
Rebundling travel
Why is rebundling important?
As I explain in Platforms or glorified distributors?
You gain power as an intermediary through rebundling.
Without rebundling songs into playlists, Spotify is merely a distributor.
In order to create a powerful position in the middle, intermediaries need to create new value through rebundling.
Rebundling can help shift power in the direction of the intermediary (and away from the producer) if consumer attention skews towards the new bundle.
And this is the unique opportunity that travel platforms should pursue.
Right to rebundle
If you own the right to rebundle the trip for the consumer, you can act as a central hub into which all other travel providers integrate.
Yet, this position has thus far remained elusive.
Online booking engines try to attack this problem. But while they have access to all the inventory, consumers engage with booking engines only during a narrow window in their consumer journey — at the specific point of booking. Hence, the best that booking engines can do is only cross-sell and up-sell other parts of the trip.
Owning the decision is critical to owning the right to rebundle. And by creating a reputation system for the travel industry, TripAdvisor has gained the right to own the decision. Yet, helping consumers choose hotels or restaurants or activities doesn’t automatically translate to aiding them across the more complex workflow of travel planning.
Further, the right-to-rebundle is won not only at the time of travel planning but more importantly, during the actual trip itself. A player that rebundles the trip should also be capable of assisting consumers in making changes to the trip on the fly. This might involve rebooking downstream accommodations when a flight gets rescheduled. Or creating consolidated claims sheets for claiming travel insurance. These disparate activities, again, increase coordination costs for consumers.
The right-to-rebundle accrues to the player that can
(1) effectively support key decisions during travel planning, and
(2) minimize coordination costs during travel execution.
Decision support and coordination are key to winning the right-to-rebundle.
Rebundling travel in the age of AI
To design a solution for rebundling of travel, let’s revisit where travel is currently bundled.
Since before the rise of the Internet, travel was bundled by travel agents offering package tours. The problem with these bundles was that they were not customer-centric. They were bundled by travel agents and could be customised (to give the semblance of personalisation) but only within production-side limits (e.g. you can have any hotel as part of your package as long as the agent has a lead-generation deal with them).
With AI, we now have new vectors of rebundling, which can effectively combine the abundance and choice offered by today’s travel inventory with the ability to create consumer-centric bundles or trips.
As I explain in How AI agents rewire the organization:
Let’s say you’re responsible for managing corporate travel.
- An LLM can help generate a list of interesting destinations and an itinerary that meets certain constraints.
- An AI agent can operate with far more complexity and look up the top-rated hotel with available rooms during a specified period, within a specific budget and complete the task of making the reservation.
- An autonomous AI agent can take this several steps further by learning about your context over time and finding and booking the hotel that best meets your travel preferences and constraints.
How exactly does this work? What’s special about autonomous AI agents?
Autonomous AI agents provide the first instance of goal-seeking, self-learning, path-finding, path-adjusting technologies.
First, AI agents are goal-seeking:
In order to accomplish this goal, an agent must (1) scan the environment, (2) plan and the tasks involved, and (3) act out the plan leveraging other agents and digital resources.
Second, autonomous AI agents learn continuously to get better at goal-seeking.
Much like rational human actors, agents can reflect on their actions and results and refine their path towards goal-seeking in response.
Finally, agents can call agents, paving the way for hierarchical organization of work.
Agents can call agents; it’s agents all the way down. A travel-planning agent can call a hotel booking agent and an activity planning agent and coordinate across the two (and more) to accomplish its overall goal.
Unbundling and rebundling the travel agent
The key to rebundling travel is to start by looking at the legacy bundler — the travel agent.
Think of what determines effectiveness of a travel agent:
- The managerial capabilities required to coordinate travel planning and execution
- Ease of access to the options needed for planning and execution
The extent to which travel can be rebundled through AI agents depends on two key factors:
- The sophistication of the AI agent
- The degree of interoperability and open access to travel options
Compared to other industries, the travel distribution value chain started opening out towards distribution of inventory back from the days of the Global Distribution Systems.
But the decision to travel doesn’t just involve booking hotels or flights (inventory), it also involves evaluating and considering a whole range of rich information around this inventory — reviews, special features, customisation options etc. Global Distribution Systems weren’t set up to distribute this rich content.
A lot of this rich content was embedded in the managerial capabilities of the travel agent. A good, well-informed travel agent would provide rich context, another one would not.
However, with API-based distribution, open access to travel options is increasing all the more.
Today, travel APIs increasingly distribute rich content in addition to inventory. This content can be served across the customer journey — most importantly, in all the interesting parts. After all, booking tickets is the least engaging (and possibly most onerous) step in the traveler’s customer journey.
The right to provide decision support and coordination support to the traveler stems from gaining a right to play in the non-booking steps of the customer journey.
Transforming consumer travel in the age of AI
The effects of unbundling are most directly observed in case of consumer travel. An AI assistant can start rebundling this by creating a conversational interface to travel booking. In this way, an AI assistant can start gaining the right to customer relationship.
As an AI assistant gains the right to relationship, it establishes itself as a decision making hub. This grants it negotiating power to get other travel providers to open out their APIs and provision content and capabilities to this hub.
Once this plays out, an AI agent can most effectively rebundle travel by working across these resources opening out in the ecosystem.
Transforming corporate travel in the age of AI
Corporate travel involves larger workflow integration. Players that already provide this workflow (for instance, Concur) are already established in a hub position. Other players in the travel ecosystem often connect into this hub position. Where expense management and policy compliance are key considerations in corporate travel, players that serve these considerations gain the right to customer relationship and the right to the core data identifier (e.g. the traveler account ID), using which they rebundled all other workflows (through API integration) around this core position.
These incumbents corporate travel solution providers are well-positioned to leverage their dominant hub position and bundle an AI agent on top of it. The customer relationship is already owned by the workflow hub. The agent simply gets bundled into this position to strengthen the incumbent’s position.
Convergence
Eventually, if both decisions and resource deployment shift to agents, we will increasingly see convergence between how consumer travel is booked and how corporate travel is booked.
If you want to understand how this larger pattern plays out across industries, have a look at:
How to win at Vertical AI
Augmenting and eliminating agents in consumer travel
AI travel planning assistants will initially augment travel agents, and eventually seek to eliminate them.
As I explained in AI-led growth:
With AI-led growth, a sufficiently sophisticated AI sales assistant, empowered with product knowledge (trained by the company) and customer knowledge (trained by resellers) can take over the human intermediary.
This plays out in four stages:
- Stage 1 — Assistant augmentation: Converts product knowledge (about travel options and choices) into a ‘travel plan assistant’ — an AI assistant which assists travel agents in planning travel.
- Stage 2 — Salesforce expansion: The number of travel agents dramatically increases as learning curves go down for new travel agents who do not need to have deep historical knowledge.
- Stage 3 — Assistant learning: This widening base of travel agents using the AI assistant rapidly create data about different customer needs as they work with the assistant to address them.
- Stage 4 — Salesforce contraction: Once sufficient customer data has been captured, the assistant — evolving to an AI agent — is sophisticated enough to replace travel agents and serve the customer directly
At various points across these four steps, the actual ‘product’ form of AI may take one of two shapes:
- Replacing the current channel (prospective customer interacts with AI assistant) or
- Embedding in the current channel (AI assistant augments travel agent).
Rewiring workflows in corporate travel
Things play out a little differently in corporate travel, where the travel booking decision is embedded in a payments and compliance workflow. Unfortunately, the players that manage this workflow do not have a unique right to serve the decision.
As I explained in AI-led growth:
With AI-led growth, a sufficiently sophisticated AI ‘decision support system’, empowered with product knowledge (trained by the company) and customer knowledge (trained by end users) can take over the legacy workflow.
This plays out in four stages:
- Stage 1 — Workflow augmentation: The AI assistant is embedded into the travel planning and booking workflow.
- Stage 2 — Decision interface emerges: The AI assistant starts enabling an important and underserved decision in the workflow. Where this happens, usage may start shifting from the workflow interface to the AI assistant. Users increasingly start their journey with the assistant.
- Stage 3 — Assistant learning: The more travels planned and booked, the more this assistant is trained. This widening user base using the AI assistant rapidly creates data about different user needs and the assistant learns from them as it works alongside them. As the AI improves and usage shifts to the assistant (and eventually an agent), the rest of the workflow starts getting reorganised around the AI agent.
- Stage 4 — Decision interface dominates : As the AI agent improves, it may absorb more of the workflow.
Imagining an interoperable future for travel
A truly custom-centric trip can be achieved only if it is — well — centred around the customer.
The problem to that — today — is that every platform serving travel use cases has their own little island of user data. Expedia has some, Airbnb has some, and TripAdvisor has some. No one has a full view of the customer’s trip.
In order for agents to be truly successful, we need to move to a model where travelers increasingly manage their own ‘data passport’.
As I explain in The false bundles of the platform economy:
For user-centric AI agents to start making decisions on behalf of users, we need user-owned data systems (e.g. Solid Protocol) where multiple AI agents compete to gain right to mediate services to the user and provide the best recommendations without compromising user privacy.
This would involve a perfectly competitive market of AI agents, all competing to create the perfect trip for the traveler.
AI Travel — Winners and Losers
This brings us back to my favourite question.
Where do power and profits shift? Who wins and who loses?
Travel agents: LOSE — In the short run, we may see them winning as they are augmented by AI assistants. In the long run, they lose. Not necessarily because they are entirely replaced, but because their sales commissions will go down. Their ability to charge a premium gets eroded.
This is an application of the following thesis:
AI won’t eat your job, but it will eat your salary
Travel providers: WIN/LOSE — The corollary to the above is that hotels and airlines that rely on sales and now can increasingly engage customers through AI assistants are positioned to win as their cost of customer acquisition and retention dramatically falls. But that is if they can stop acting like Air Canada.
This is an application of the following thesis:
AI-led growth
AI travel plan assistant providers: WIN — These players stand to benefit from the shift to AI-led growth. However, to get this right, they will have to get industry stakeholders on board. And that is where vertical advantages will play a key role. AIs that already have deep vertical advantages will be more likely to be selected by enterprises that care about not compromising their customer experience.
This is an application of the following thesis:
How to win at Generative AI
Corporate travel workflow incumbents: WIN/LOSE — Corporate travel workflow providers will need to assess where they stand. If they already serve the most important user decisions within their workflow, they can use AI to improve their positioning. If the most important user decisions in the workflow are currently underserved, they can seize that opportunity with AI. The ones who do can extend their workflow advantages to now win with an intelligence advantage. The ones who don’t may risk getting displaced by AI-led challengers.
This is an application of the following thesis about workflow incumbents:
How to lose at Generative AI!
If you enjoyed reading this article, join my mailing list at platforms.substack.com to receive a carefully curated set of ideas every week