Solving the "DoorDash Problem"
Nilay Patel from “The Verge” has recently been talking about the “DoorDash problem” for some consumer internet companies. What is the “DoorDash problem”? From The Verge:
“Briefly, it’s what happens when an AI interface gets between a service provider, like DoorDash, and you, who might send an AI to go order a sandwich from the internet instead of using apps and websites yourself.
That would mean things like user reviews, ads, loyalty programs, upsells, and partnerships would all go away — AI agents don’t care about those things, after all, and DoorDash would just become a commodity provider of sandwiches and lose out on all additional kinds of money you can make when real people open your app or visit your website.”
As you can tell, this isn’t really a DoorDash specific problem, rather emblematic of a larger problem for many consumer internet companies, especially marketplaces that rely on ads to make their overall economics work. Predictably, these consumer internet companies don’t think (or at least don’t want you to think) AI agents can really threaten their position. They believe their specific data, vetted networks, or brand trust create a defensive “moat” that AI agents cannot easily cross. Verge quoted C-suites from Lyft, Zocdoc, Taskrabbit etc who all sort of echoed this sentiment.
However, the CEO who I felt had the most cogent strategic argument to AI agents is Uber’s Dara Khosrowshahi. Notice this excerpt between Nilay Patel and Uber’s CEO:
Nilay Patel: So, I come to you and I say, “I want to be able to get cars from Uber.” What is the percentage toll? What is the extra margin you would have to charge me, in dollars and cents, to make it worth it for me to take the customer away from you in that way? Because when the customer opens your app, you get to cross-sell them into Uber One, or ask “would you like some food when you arrive?” There’s all of these other incremental opportunities you have to forgo if I take that customer.
Dara Khosrowshahi: So I have a weird philosophy on this. Initially, I charge you zero…People spend so much time trying to figure out what the economics might be when the first thing is to try it out. Is it going to be a good experience or not? Is your scheduling actually going to work, or is it going to be off and the driver has to wait for 10 minutes, which is terrible? Let’s just figure it out. Then, once you optimize the experience, we can measure. Are you an incremental consumer for Uber or are you totally cannibalistic?
If it’s cannibalistic, then I’m going to charge a lot of money. You can’t have any money because you’re getting the benefit and my content. You’re not bringing me any business at all. If it is incremental, then I would pay some take rate. Is it a 5, 10, or 20 percent take rate? It depends on the incrementality.
But I think so much innovation has slowed down because companies try to figure out the economics first. Figure out the experience first and then the economics. Listen, if I do a bad deal for a year, who cares? I’m going to renegotiate with you. I’m building stuff for the next 10 years. Success or failure isn’t going to be determined by my take rate being 5 or 20 percent in year one. It can set precedent, and precedents are dangerous. That’s why I would say to charge zero. Let’s try it out. Let’s see what the experience is. Let’s try to measure out what the value add is, and then the economics essentially will take care of themselves.
That really says it all! Nobody really needs to make a big claim whether any of these will work; the data can simply guide these companies where to strike the right balance over time.
The key word here is “incrementality”. Incrementality is why it makes enormous sense for Walmart to partner with ChatGPT as I suspect the people who shop at Walmart and the people who use ChatGPT daily may have a tiny overlap today. So, if people try to use AI agents to shop their groceries and Walmart pops up, they are highly likely to be incremental to Walmart. On the other hand, I suspect all of these ChatGPT DAUs are likely Amazon Prime members which makes it lot more cannibalistic than incremental for Amazon. And of course, if you’re using AI agents to order things from online, Amazon is not making its ad revenues which is essentially the key pillar for their retail business. As I mentioned recently, Amazon’s LTM advertising and subscription fees was ~$113 Billion which was 3.5x of its LTM operating income ex-AWS.
While Amazon has strong incentive to resist third-party AI agents roaming around its website, it does have compelling argument why such an experience is unlikely to resonate with consumers, at least at its current state. From Amazon’s 3Q’25 earnings call:
“search engines are a very small part of our referral traffic and third-party agents are a very small subset of that. But I do think that we will find ways to partner. We have to find a way, though, that makes the customer experience good. Right now, I would say the customer experience is not -- there’s no personalization. There’s no shopping history. The delivery estimates are frequently wrong. The prices are often wrong. So we’ve got to find a way to make the customer experience better and have the right exchange value. But I do think that the exciting part of this and the promise is that AI and agentic commerce solutions are going to expand the amount of shopping that happens online. And I think that’s really good for customers, and I think it’s really good for Amazon because at the end of the day, you’re going to buy from the outfit that allows you to have the broadest selection, great value and continues to deliver for you very quickly and reliably. And I think that bodes well for us.”
I have tried AI agents and my initial underwhelming reaction still persists today. But as I said before, there is a pretty clear distinction between the full autonomous agentic AI and conversational AI experience. And I am a much more ardent believer of conversational commerce than truly agentic commerce.
My friend Liberty was telling me yesterday that he’s looking into buying luggage for his next family trip. He shared the below search result on Amazon; you can see why it would be a much better experience if he just provided some key aspects he’s looking in a luggage (price range, size, color etc.) and then ask the AI to give him three options which he then can click to buy or go to the specific website himself to complete the purchase. At least I am certainly not there yet that I would just ask a prompt to any AI and let it make the final purchasing decision, including payments. The reality is there are so many contexts that reside in our head that can be difficult to write down in minute details while prompting to AI, but AI can certainly narrow down the decision process from ~25-30 similar looking luggage to just three and allow me to pick one.
Given this context, I do think OpenAI’s yesterday’s launch of “Shopping Research” is very appropriately named. I would love AI agents to do research on my behalf to help in shopping decision process, but not take the final decision itself. Such a feature should pose an immediate threat to Google’s networking business which has already been declining in revenue for the last 13 consecutive quarters.
The core Google search or Amazon’s advertising can also be under pressure if ChatGPT does become the “superapp”. However, customers seem to trust retailers’ on-site agents 3x more than third-party agents such as ChatGPT, but that can change over time. Of course, Amazon shouldn’t be actively helping ChatGPT to become a superapp by making all their inventory available to shop on ChatGPT given much of it is going to be cannibalistic for them. The two most important factors for e-commerce online is selection and delivery speed, and it is difficult for anyone to match Amazon on either front.
OTAs also have this “DoorDash problem” although there are important nuances to it. Unlike many other marketplaces, Booking makes less than 5% of their revenue from ads. Airbnb doesn’t even have any ad revenue. However, ceding the customer relationship to ChatGPT can make Booking’s “connected trip” dream difficult to execute. Last week when Google launched some AI features for travel, all the OTA stocks took bit of a beating as investors started to entertain the possibility that Google may end up making OTAs irrelevant. Skift later reported that Google clarified that it was not their intention:
“In a follow-up with Skift, Julie Farago, Google’s vice president of engineering for Travel and Local, sought to clarify the company’s intentions. “Google has no intention of becoming an online travel agency,” she said.
“For many years, we’ve helped travelers compare their options across different providers and complete the booking with the partner of their choice. Our work to enable agentic booking for flights and hotels builds on this framework.”
Farago stressed that “Google won’t be the merchant of record. Partners will continue to service the booking and manage the customer relationship.”
In my Booking Deep Dive, I also made the case that the risk to OTAs are likely overstated. From my Deep Dive:
While agentic booking remains a mirage, conversational booking or planning the entire trip on Google or ChatGPT seems like a real possibility. But even if that happens, I am not sure that’s necessarily a big negative for Booking. Booking already lives in a world where ~35-40% of its customers don’t start their travel journey on their app or website. Instead of fighting for those customers on core Google search, they will just have to do that within the confined box of ChatGPT or Gemini. Is this fundamentally a different paradigm? Even if it’s different, is it necessarily a negative development for Booking?
On one hand, you could say if we move away from the search environment where Google has ~90%+ market share to a relatively more bifurcated chatbot experience, I wonder if that may lead to some CAC deflation compared to status quo. On the other hand, if consumers become so habituated with general chat bots that the starting point of planning a trip shifts more to indirect channel, Booking’s direct or organic traffic may already have peaked. While you can start planning the trip on Booking app itself, ultimately the question may come down to integration and personalization. Which app or company will have much richer context of your personal preferences? ChatGPT or Booking? If you plan the entire trip in Gemini, perhaps Gemini will be able to embed your trip details automatically in Google Calendar, use Google Maps directly from the Gemini app environment, and utilize Gmail to easily find booking confirmation or flight codes. It is not difficult to imagine how such an integrated experience may seem much better than Booking’s connected trip experience. Nonetheless, Booking can respond to such a headwind by dialing up the benefits of their Genius loyalty programs as long as doing so is accretive for them from margin perspective. As you can imagine, it is simply too early to know how any of these may evolve.
Airbnb should be more immune from some of these headwinds though. Most of their exclusive listings don’t have their own websites, so they will remain reliant on Airbnb in driving their bookings. Airbnb itself will come up with their own AI-powered search and if unique listing is what you’re looking for, your best bet will continue to be to go directly to Airbnb app or website.
Given this context, I did make a small change in my portfolio yesterday which I will mention behind the paywall.
In addition to “Daily Dose” (yes, DAILY) like this, MBI Deep Dives publishes one Deep Dive on a publicly listed company every month. You can find all the 65 Deep Dives here.
Current Portfolio:
Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.