Are AI connectors solving a problem that does not exist?

The AI platforms need to monetize their traffic, since AI computing is exceptionally expensive! The quickest way to do so is to partner with well-established travel brands and to start earning affiliate commissions from them. 

To achieve that, AI connectors play a crucial role, but what kind of connectors?

The keyword here is “invisible B2B connectors” that service well established hotel and travel brands, but remain invisible in the background.

Ex. DirectBooker, an AI connector, founded by executives from Tripadvisor and Google Travel—Steve Kaufer, Richard Holden and Sanjay Vakil, recently announced supply agreements with five hotel chains including Best Western and Radisson Hotel Group. 

Google recently cited partnerships with global OTAs (Booking, Expedia) and global chains (Choice, IHG, Marriott, Wyndham) - all of these travel brands have one thing in common: huge loyalty programs and marquee brand names, making them appealing booking entities. 

Independent hotels? The only option is for CRS, Channel Managers and cloud PMS vendors to plug via AI connectors into Google’s UCP or to AI platforms like ChatGPT via MCP or A2A, allowing travelers to search and book rooms. Similar to how these vendors currently handle metasearch, they will charge the hotel a fee, which they can share with the AI platforms.

The question lands precisely where the industry tends to look away. Travellers do not wake up wanting orchestration. They want outcomes: the right option, at the right price, with minimal friction and maximum trust. The moment a connector introduces one extra decision — which system to link, whether it is safe, whether it will work — adoption collapses outside a narrow power-user segment.

Where connectors genuinely solve a problem is on the supply and enterprise side. Linking AI to PMS, POS, or P2P systems removes friction inside the business and lets staff act faster across fragmented data. That is real value — but not a traveller value proposition. Two distinct problems, repeatedly conflated in the same product narrative.

So the honest framing is not "are connectors useless?" but "who are they for?" If the traveller never sees the value, the winning model will be embedded distribution: AI experiences that already carry inventory, payments, policies, and support by default — vendor-neutral where possible, and invisible by design.

In travel, the best technology is the one the guest never has to think about. Connectors are a powerful B2B instrument; as a B2C promise, they risk solving a problem travellers were never trying to solve.

At least in their current form, I think AI connectors risk solving more of an industry problem than a real user need.

The idea of a unified journey makes sense, but it breaks when it assumes that travelers will actively connect tools, authorize systems, and manage infrastructure they’re not familiar with. Asking them to “connect the stack” simply adds friction.

There is also a separate issue with the interface itself. Chatbots are great for research, because they simplify discovery and comparison. But they remain a weak surface for booking. e-Commerce has already shown this, and it’s hard to imagine users fully trusting a conversational flow for something like a hotel stay, where price, policies, and confidence all play a role.

This might change, in part, with personal agents. If people start trusting them to handle broader aspects of their lives, they may also trust them to book on their behalf, especially for low-impact stays, like a stopover during a trip or a basic business stay. But that’s still a few years away.

Until then, connectors make sense only if they stay in the background. If users have to think about them, activate them, or even understand them, the model is ahead of its audience.

I think we're doing something slightly different: we're building for a capability that will disappear pretty quickly. To me, these in-chat apps are like computer-use agents: they're there to fill an immediate gap while we develop something better in the background. They're a bridge, one that will be left unused for travel discovery/shopping/booking applications as soon as personal agents arrive.

I suppose there may be some use cases for travelers to call up an OTA booking app within ChatGPT, though none come to mind. I think that if people want to use an OTA, they'd get a better experience by going to it directly.

But the real test will come when OpenAI and Anthropic release their personal agents later this year or early next year. Then, travelers will be able to ask their chosen agent to research and plan their trip on their behalf and get a short list of personalized itinerary recommendations to choose from. The issue, then, is whether hotels will develop the infrastructure to attract those agent travel requests directly...or let the agents shop at the OTAs who are aggressively preparing for them.

AI connectors are solving a real interoperability problem today, but they’re transitional infrastructure, not the long-term architecture for hospitality AI.

Agentic AI is shifting hospitality from human-operated software to machine-operated systems. Today, connectors provide the abstraction layer allowing AI agents to interact with fragmented hotel infrastructure through APIs.

However, AI agents are already capable of operating software directly from IBEs and PMS platforms to operational systems. Their dependency on connectors exists because most hospitality systems were designed for human workflows.

As agentic AI matures, agents will increasingly orchestrate workflows autonomously, researching options, modifying reservations, coordinating operations, and ultimately booking on behalf of travelers within a single continuous flow.

In this environment, the way connectors are implemented will become invisible to the traveler. Users will not search for or configure connectors themselves. Orchestration will happen dynamically in the background. 

This is why Apaleo has consistently invested in an open, API-first architecture and standards like MCP. The future is not another AI layer alongside the PMS, but an agentic service layer where AI can securely operate across the hotel ecosystem with multi-system execution capabilities.

The real question is not whether connectors matter today. It’s whether the industry is building for AI-native travel behaviour or maintaining legacy software assumptions.

There's absolutely no way to generate any reasonable amount of bookings from a channel that requires installation of an app (unless you're a global recognizable brand with a large pool of loyal customers or spend millions on marketing).

We've all been through this with mobile apps. This is no different.

Existing iteration of apps withing AI platforms like ChatGpt and Claude is just that - iteration. There have been talks about simplifying the "connection" process and invoking apps based on the context of user query.

And THAT will make a big difference. And we don't know when and how exactly this will be implemented but what we do know is that AI platform functionality (especially with OpenAI and Claude) changes faster than your girlfriend's mood.

But also: if apps are invoked based on context - how will the prioritization happen, which app will be chosen out of myriad of apps in the marketplace (ChatGpt already offers dozens in our space)?

And that is a question for a separate discussion.

Old habits die hard. It will be many years after the major advent of change where we will still see people undertaking processes they preferred before the AI age. Leaders and laggards as the story goes.

In the medium term I don't think it is reasonable to expect the traveller to undertake these mechanics to gain an outcome. As per the above, some will. But there is a continued assumption that people know what they're doing and how to do it efficiently, like some old travel or hospitality hand. The reality is that in the majority this is not the case.

The point at which this level of intelligent automation will come is when the technology itself is smart enough to draw the lines between data and execution. It will form the necessary joins and hooks needed to achieve efficiency. No humans involved! Ahh, the bliss.......!?

Which brings us back to the question sitting out there for all of us in hospitality land.

When it makes those connections on its own, which one will it choose..........

Will it be yours?

AI connectors often try to solve a problem travellers don’t actually feel: most people want simplicity, not more setup or control. Expecting users to connect unknown systems to tools like ChatGPT adds friction rather than removing it. Where connectors do work is behind the scenes, integrating systems in the hospitality tech stack. For consumers, the winning approach is invisible integration, not user-managed connections. If connectors succeed, it will be because users never have to think about them at all.

The question here is not be whether AI connectors are necessary, but why AI platforms are pushing branded connectors in the first place?

 The current connector model pushes branded integrations and user-managed connections that most travelers neither understand nor want to think about. That creates unnecessary friction in an experience that should feel simple and intuitive.

 Travelers are not looking to orchestrate backend systems or manage a collection of hospitality tools inside ChatGPT or Claude. They simply want the best, most relevant travel options presented seamlessly.  This is a fundamental Google did with traditional search.

 Where AI connectors become valuable is not as consumer-facing brands, but as invisible technology that helps surface hotel availability and rates without adding complexity. Without that kind of infrastructure, travelers will continue gravitating toward OTAs because they appear to offer the most complete view of inventory.

 Some of the best technology in travel is the technology travelers never have to think about. AI should reduce friction, not introduce more of it.

Both positions have some truth to them: travelers don't want to connect to new tools, but the current experience in AI platforms is lacking. In ChatGPT or Claude, you can do great research for a trip, but you are very far from completing the process: information is coming from non-validated sources (travel providers don't have any control), there's is no live data of any type (for example, hotel rates), and LLMs tend to equalize all providers, playing a "this is nice, but this is also nice" game. 

Also, we have been here before: we use apps in our phones all the time, simply because they are a much more efficient way to get things done. 

In summary, the issue is not that connecting to new tools is bad, it's that we need a better user experience to connect and use them. ChatGPT and its peers need to do a much better job of including apps and connectors into user conversations.

Simone's right that the connector model assumes a traveler who doesn't exist. Most people booking a trip have no interest in linking apps to their AI assistant, and asking them to do it makes the experience worse than just opening Booking.com.

There's also a distribution asymmetry that gets ignored. A traveler might install a connector for an OTA they already use, or for a chain whose loyalty program they're enrolled in, because the brand is familiar and the relationship already exists. Independent properties and B2B platforms operating inside the hospitality stack don't have that recognition, so the connector sits unused regardless of how good it is.

The whole pattern of building apps on top of chatbots also looks like a transitional phase. It's the equivalent of browser plugins, which never scaled into a real distribution layer despite years of trying. I don't see an App Store forming on top of ChatGPT or Claude either, because the platforms have every incentive to keep the relationship with the user and none to hand it to third parties.

For hotels the conclusion is uncomfortable but familiar. Whoever owns the agent layer keeps winning, and right now that means OTAs, chains, and metasearch.

Complexity vs. Capability: Travelers want outcomes—a booked room, a dinner table—not the "opportunity" to manage the backend plumbing of the hospitality stack.

At Alliants, we believe the future isn't about asking users to orchestrate systems they don’t care about. It’s about intelligence that works in the shadows. If the traveler has to think about the connector, the technology has already failed them. We should be building for the human who wants to travel, not the user who wants to configure a platform.

It’s a hype on a hype. I get vendors telling me that hoteliers need to add a MCP server to their already complicated tech stack. These are short calls. I remember the days of WAP (early mobile channel) and Nokia were pushing “WAP servers”, for about £40k each. No one bothered. WAP came and went, the IPhone was launched in 2007, and mobile got absorbed into internet standards. 

No, hoteliers don’t need extra AI connectors. The big PMS players have already launched IBEs that can surface in a chat window. They pass bookings to the brand’s website. The PMS provider runs a MCP server which links to multiple Gen AI providers. Gen AI chat providers will just be ano sales channel that a PMS/CRS/CM links to, unfortunately probably with extra commissions.The better IBEs and PMS players have added Gen AI summaries into their products. Guests will continue to mostly find our hotels in Google and Bing, with some referrals from chat engines.

Focus of guest service and operational excellence.

The second a traveler is asked to connect apps, manage systems, or understand how things work behind the curtain, the whole idea of “simple AI” falls apart.

AI was supposed to make things easier. Instead, we’re sometimes just replacing old friction with new, tech‑heavy friction and calling it innovation.

Most travelers are not trying to be more efficient. They’re trying to get on with their lives. They don't care how a booking engine is used, nor do they know what a CRS does in the background.

Expecting people to connect tools they’ve never heard of to an AI assistant assumes a level of technical interest and patience that just doesn’t exist in the real world. Real progress happens when all of that disappears.

AI should quietly decide which systems to use, when to use them, and why, based on what the traveler wants — not ask the traveler to do the work. If the tech starts showing through the experience, we’ve missed the point.

The future of travel isn’t about teaching people how the technology stack works.

It’s about making sure they never have to think about it at all.

Agree. The idea sounds compelling: connect booking engines, PMSs, loyalty systems, and travel tools directly into ChatGPT or Claude to create one seamless booking journey.

The real question however is simpler: why would a consumer actively connect a hospitality infrastructure app they have never heard of — or care about the brand of — to their AI assistant?

For strong consumer brands like Booking.com, Airbnb, or Accor, connectors may make sense because users already know and trust them. But for invisible B2B infrastructure companies, AI apps currently feel more like PR for hoteliers than something that will genuinely drive more business.

We built one too. Interesting experiment, but probably not where the long-term value sits.

The real opportunity is elsewhere: making inventory, rates, availability, attributes, and differentiated product data directly consumable by LLM ecosystems — similar to what Google Hotel Center did for search and distribution.

That is where the competitive advantage starts.

In an AI-driven discovery world, structured and differentiated product data matters far more than another connector nobody will activate.

The real game may not be AI apps. It may be GEO in the absence of direct access for true data feeds of ARI.

Complexity is the Killer in B2C

I believe we are currently overengineering a problem most travelers simply do not have. Complexity is the biggest killer in B2C adoption. Consumers only embrace technology when it removes friction, not when it introduces another layer of setup, permissions, and integrations.

The connector logic may work perfectly in a professional environment. Connecting Outlook to your CRM, or integrating sales tools into your daily workflow, creates value because these are systems people use constantly. But travel is different. Most people book one or two trips a year. They do not want to manage APIs, activate connectors, or decide which booking engine should be linked to ChatGPT or Claude.

The customer knows Booking.com, Expedia, or Tripadvisor because they are consumer brands. But nobody wakes up thinking, “I should connect my AI assistant to a hotel middleware platform.”

The future of travel AI will not be won by the platform with the most connectors. It will be won by the platform that hides complexity best. Travelers want outcomes, not orchestration. If users need technical understanding to make the experience work, then we are not building for mainstream travelers — we are building for a small group of tech enthusiasts.

AI connectors are not solving a non-existent problem. They are solving a very real enterprise challenge: legacy fragmentation. Hospitality still runs on layers of systems that were never designed to behave as one experience. The issue is not whether connectors matter. They do. The question is who should experience them.

Travelers should not be asked to understand the hospitality stack. They should not need to know which booking engine, CRS, loyalty platform, payment layer, service system, or content source is being invoked behind the scenes. If the value of a connector depends on the consumer actively managing it, then too much of the architecture has been exposed.

The real opportunity is not “connecting your travel tools to AI.” It is AI-enabled orchestration that disappears into the experience itself.

From an Oracle Consumer perspective, the winning model is one where enterprises use AI and connectors to unify identity, inventory, service, and fulfillment, while travelers receive better recommendations, more relevant offers, and smoother outcomes.

So yes, we should be skeptical of connector strategies that assume consumers want to manage complexity. Of course, they do not. But invisible, intelligently orchestrated infrastructure is exactly where much of the future value will be created.

The question here is not be whether AI connectors are necessary, but why AI platforms are pushing branded connectors in the first place?

The current connector model pushes branded integrations and user-managed connections that most travelers neither understand nor want to think about. That creates unnecessary friction in an experience that should feel simple and intuitive.

Travelers are not looking to orchestrate backend systems or manage a collection of hospitality tools inside ChatGPT or Claude. They simply want the best, most relevant travel options presented seamlessly. This is a fundamental Google did with traditional search.

Where AI connectors become valuable is not as consumer-facing brands, but as invisible technology that helps surface hotel availability and rates without adding complexity. Without that kind of infrastructure, travelers will continue gravitating toward OTAs because they appear to offer the most complete view of inventory.

Some of the best technology in travel is the technology travelers never have to think about. AI should reduce friction, not introduce more of it.