--- title: "The AI Booking Flow: Rethinking Commerce for Chat" description: "Traditional e-commerce is page-based: product, cart, checkout. Conversational commerce collapses the entire purchase into a single thread." canonical: https://nowah.xyz/blog/ai-booking-flow-commerce-ux lastModified: "2026-08-07T07:53:45.409Z" --- # The AI Booking Flow: Rethinking Commerce for Chat Traditional e-commerce is page-based: product, cart, checkout. Conversational commerce collapses the entire purchase into a single thread. There is a moment in every traditional e-commerce checkout where you lose people. They picked the thing. They want to buy it. They are ready. And then you send them to a cart page. Then a checkout page. Then a review page. Each page transition is a chance for them to reconsider, get distracted, or simply give up. Travel is the worst offender. Cart abandonment in travel booking runs between 81% and 87%. That means for every ten people who find a flight they want, at most two actually buy it. The other eight or nine disappear somewhere between "add to cart" and "confirm booking." This has been true for over a decade and nobody has fixed it. We think the problem is structural. The page-based commerce model was designed for browsing physical goods on desktop computers in 2004. It makes almost no sense for booking travel on a phone in 2026. So we threw it out and built something different. ## Page-based commerce is showing its age ![Illustration for this section](https://pics.nowah.xyz/website-media/product-017-img-1.webp) The standard e-commerce flow has four distinct stages: product page, cart, checkout, confirmation. Each stage is a separate page load. Each transition requires the user to maintain their intent through a context switch. Expedia's flight booking flow is a textbook example. You search for flights. You get a results page with hundreds of options. You pick one. You go to a review page. You enter traveler details. You go to a payment page. You enter payment info. You review everything again. You click "Complete Booking." That is six or seven distinct page loads between finding a flight and owning a ticket. Every single transition is a drop-off point. Research from the Baymard Institute shows that adding just one extra checkout step reduces conversion by roughly 10%. Expedia has at least four checkout steps after you select a flight. Do the math. The problem compounds on mobile. Each page load on a phone means a full-screen transition, a new scroll position, a moment of disorientation. The back button becomes a weapon of mass abandonment. One accidental tap and you are back on the results page, your selection gone, your traveler details erased. Start over. [Google Flights](/blog/best-flight-booking-2026-ai-vs-google) is faster, but it still operates on the same paradigm. Search page. Results page. Flight detail page. Redirect to airline or OTA for booking. That redirect is where Google loses most people, and Google knows it. They have been trying to bring more of the booking flow in-house for years, precisely because they understand that every page transition costs them. ## Conversational commerce collapses the funnel Here is what happens when you book a flight through Nowah. You open the app. You are in a conversation. You say something like "I need to fly to Barcelona next Thursday, coming back Sunday." The AI asks if you have a time preference. You say morning. The AI searches, and while it searches, you see it working. Three flight options appear as cards in the conversation. You tap one. A summary appears with the price breakdown, right there in the chat. You confirm. A [payment sheet](/blog/payment-sheet-native-vs-custom) slides up. You pay. Done. Booking confirmation appears as the next message. That entire flow happened in one screen. One context. No page transitions. No cart. No checkout page. No review page that looks different from every other page you have been on. The conversion advantage is not mysterious. When you keep someone in one context, they stay focused. There is no moment where the screen goes blank and reloads, no moment where they wonder "wait, did I pick the right one?" and hit back. The conversation is a continuous thread. Every piece of information they need is visible by scrolling up. Their intent, the options, their selection, the price, the confirmation, it is all in one stream. This is what we call conversation-native commerce. The purchase does not happen alongside the conversation. It happens inside the conversation. The [distinction matters](/blog/ai-agents-vs-chatbots-distinction-matters). ## Presenting structured data in chat ![Supporting diagram](https://pics.nowah.xyz/website-media/product-017-img-2.webp) The hard design problem with conversational commerce is not the text. It is the structured data. Flights have departure times, arrival times, durations, stops, airlines, prices, baggage allowances. Hotels have nightly rates, star ratings, locations, amenities, photos. You cannot present this information as a paragraph of text and expect anyone to make a decision. We solved this with rich cards that live inside the chat flow. When the AI presents flight options, they appear as compact cards with a clear information hierarchy. Price and schedule are primary. Airline and stops are secondary. Baggage and cancellation details are available on tap. The cards are designed to be compared at a glance. Three cards, stacked vertically, with consistent formatting so your eyes can jump between price on card one and price on card two without hunting. This is something chat interfaces traditionally do poorly because chat is designed for variable-length content, not structured comparison. We spent months getting the card layout right. Hotel cards are a different challenge. Hotels are more subjective than flights. A flight from New York to Barcelona either works with your schedule or it does not. A hotel in Barcelona involves taste, neighborhood preference, proximity to where you want to be, photo quality, review sentiment. We lean heavier on AI synthesis for hotels. The agent does not just show you three hotels. It tells you why each one fits what it knows about you. "This one is in the Gothic Quarter, which you liked when you visited Madrid. Walking distance to the beach. Strong reviews for cleanliness." The key principle: structured data in chat should feel native to the conversation, not like a webpage that got crammed into a [message bubble](/blog/ai-message-bubble-is-canvas). The cards flow in the thread like a message from the AI, because that is what they are. ## Payment within conversation The trickiest moment in any conversational commerce flow is the transition from browsing to buying. You have been casually chatting with an AI about flights to Barcelona. Now you need to hand over your credit card. The tone shifts. The stakes change. Most chatbot commerce implementations fumble this by doing one of two things. They redirect you to a separate checkout page, which kills the entire "one context" advantage. Or they try to collect payment information through the chat itself, asking you to type your card number into a message box, which feels insecure and awkward. We use native [payment sheets](/blog/payment-sheets-in-chat-booking-flows). When you confirm a booking, the operating system's payment sheet slides up. On iOS, that means Apple Pay or a saved card. On Android, Google Pay or saved payment methods. The sheet overlays the conversation without leaving it. You authenticate with Face ID or fingerprint. The sheet disappears. The conversation continues with your booking confirmation. This works for two reasons. First, the user never leaves the chat. They see the conversation behind the payment sheet, which maintains context and reduces the "where am I?" feeling that separate checkout pages create. Second, the payment sheet carries the trust of the operating system. Apple Pay and Google Pay are familiar. Users have already decided they trust those systems. We are borrowing that trust rather than asking users to trust a new checkout form. The price breakdown appears as a card in the conversation before the payment sheet. Subtotal, taxes, fees, total. Clear and visible. The user can scroll up and see the flight they selected, the dates, the times, all right there. Nothing hidden. No surprises at checkout, which is one of the top reasons for cart abandonment across all e-commerce. ## Post-purchase in-thread In traditional e-commerce, the confirmation page is a dead end. You see an order number, maybe a summary, and then what? The flow is over. The next interaction is an email. In conversational commerce, the confirmation is a message. And messages can be followed by more messages. Your booking confirmation appears in the chat, and the AI immediately follows up. "Your flight is booked. Want me to find a hotel in Barcelona for those dates?" The conversation continues. The commerce continues. This is where conversational commerce has a structural advantage that page-based commerce cannot replicate. After a traditional booking, the user has to start an entirely new search flow for the next thing they need. New search form. New results page. New context. In chat, the next purchase is one message away, and the AI already knows everything about the trip so far. We see this in our data. Users who book a flight through the chat are significantly more likely to immediately book a hotel than users on traditional platforms who complete one booking and then have to restart the entire process for accommodations. The friction reduction is not subtle. The confirmation message also serves as a reference. Users scroll back to it days later to check their booking details. It is a permanent part of the conversation history, not a confirmation email buried in their inbox. The trip itinerary updates automatically in the Trips tab, but many users prefer the chat as their primary reference because the context around the booking is preserved. They can see what they asked for, what options they considered, and why they chose what they chose. ## How WeChat proved chat commerce works WeChat did something in China that Western tech companies have been trying to replicate for a decade. They made chat the commerce platform. Not an add-on to commerce. The platform. Within WeChat, users can browse products, pay merchants, book services, and send money to friends without ever leaving the chat app. By 2024, WeChat Pay processed over a billion transactions daily. The entire commerce flow, from discovery to purchase, happens inside the messaging interface. What WeChat proved is that the psychological barrier to buying within a chat is much lower than the barrier to navigating to a separate commerce site. Users are already in the chat. They are already engaged. The transition to purchase is a tap, not a journey across multiple pages and forms. But WeChat's model has a limitation. It is merchant-initiated. Brands push products into feeds and mini-programs. The user still browses and selects from catalogs, just within the WeChat container. AI changes this in a fundamental way. Instead of browsing a catalog within chat, you tell the AI what you want and it finds it for you. The discovery step, which WeChat handled with mini-programs and merchant feeds, the AI handles with natural language understanding and real-time search. This collapses the funnel even further. WeChat took commerce from five pages to two steps. AI takes it from two steps to one conversation. The AI-powered model also handles something WeChat never fully solved: complex, multi-variable purchases. Buying a t-shirt on WeChat is straightforward. Booking a multi-leg international trip with specific seat preferences, baggage requirements, and hotel standards is not something any mini-program handles well. But it is exactly what a conversational AI agent is built for. The more complex the purchase, the greater the advantage of [conversation over](/blog/why-we-chose-conversation-over-search) catalog browsing. ## The conversion advantage Let me lay out why single-context commerce converts better, because the logic is worth being explicit about. Every page transition in a traditional checkout flow does three things. It resets the user's visual context. It introduces a loading delay, even if just half a second. And it creates a decision point where the user can bail. The first one matters more than people think. When you navigate from a results page to a checkout page, your brain has to reorient. Where is the price? Where are the dates? Where do I click next? Even if the checkout page is well-designed, you spend cognitive energy processing the new layout. That is energy that should be going toward completing the purchase. The second one, loading delay, compounds on mobile networks. A checkout page that loads in 200 milliseconds on Wi-Fi might take two seconds on a spotty 4G connection. Two seconds of blank screen at the moment of maximum purchase intent. That is not a minor issue. Studies consistently show that each additional second of load time reduces conversion by 4-7%. The third one, the decision point, is the killer. Every time the screen changes, the user implicitly asks themselves "do I want to continue?" Most of the time, the answer is yes. But each time you ask the question, some percentage says no. With four or five page transitions, you are asking the question four or five times. The math of sequential small losses is devastating. In a conversational flow, there is one context. The chat. The user never leaves it. There are no page transitions. There is no reorientation. There is no loading delay between selection and payment. The decision to buy happens once, not repeatedly at each step. The travel industry's 81-87% cart abandonment rate is not a fact of life. It is a consequence of a commerce architecture that was designed for a different era. A single-context conversational flow does not eliminate abandonment entirely, people still change their minds, but it removes the architectural causes of abandonment. The people who leave are leaving because they decided not to buy, not because the checkout flow lost them. ## What this means for the future of booking We are not saying that every product should move to conversational commerce. For simple purchases, a well-designed one-page checkout is hard to beat. You do not need an AI conversation to buy a book on Amazon. But for complex, high-consideration purchases like travel, the conversational model has real structural advantages. Travel involves multiple variables, subjective preferences, and sequential purchases that benefit from context continuity. The conversation handles all of this naturally, in a way that page-based flows cannot. The industry is starting to recognize this. We are seeing more travel companies experiment with chat-based booking flows, though most are still hybrid implementations where the chat hands off to a traditional checkout page at the moment of purchase. That handoff is the exact point where the advantage is lost. If you are going to build conversational commerce, you have to commit to keeping the entire flow in the conversation. Otherwise you are just adding a chatbot to the top of the same broken funnel. We committed. The booking flow in Nowah, from first message to confirmation, happens in one thread. No redirects. No separate pages. No context switching. And it works. Not because we are smarter than the teams at Expedia or Booking.com, but because we started from scratch with a different assumption about how commerce should work when an AI agent is doing the heavy lifting. The page-based commerce flow was the best we could do when the computer was a passive tool waiting for user input. When the computer is an active agent helping you make a decision, the flow should reflect that relationship. One conversation. One context. One purchase. That is what AI booking looks like when you build it from the ground up instead of bolting it onto the old model. --- Nowah is an AI travel agent that searches and books real flights and hotels through conversation — no filters, no thirty open tabs. [Plan your next trip](https://app.nowah.xyz).