Conversational Commerce Take Two
After a few false starts, the next big thing in shopping is finally here
After years of hating on those customer service chatbots that appear in the corner of so many websites, ChatGPT made us all fall in love with conversational interfaces. It turns out that chat is pretty fun—and powerful—when large language models reach a certain scale.
The application for shopping is obvious: why get annoyed swiping through airline or restaurant sites when you could just tell a chatbot to fly from NYC to Miami for the weekend and book dinner at the most popular Mexican place in town? In fact, you can (sort of) do this today, courtesy of Expedia and OpenTable plug-ins for ChatGPT. While these plug-ins are not as sophisticated as being able to handle the nuances of my example, it points to a clear future when it comes to shopping. Clicking and tapping is slow; just telling a machine what you want is fast. Fast wins.
To get a bit wonky, conversational commerce is the next step in the evolution of shopping. The long-term trend is the reduction of friction—a trend I’ve discussed before both generally and as it applies to money. Friction, in the case of commerce, is one of the barriers that prevents transactions from occurring. Examples of friction include: finding the right item in stock and in the right size, finding the best deal, entering payment and shipping information, even signing a contract if one is purchasing a service. Friction is everything that makes it difficult to buy something.
Since people are lazy and want things easy, we’ve been on a 3,000+ year journey to make things faster, easier, and cheaper to buy things:
From the first permanent markets in Persia in 3,000BC to the Sears Catalog in the 1800s, malls in the 1950s, and more recently Amazon and mobile delivery services like DoorDash, people will embrace a new way to shop when it’s more frictionless than the old. Fast forward another few hundred years and we’ll get what we all really want: the Star Trek Replicator. Any item, instantly and at no cost, with no friction!
Conversational commerce is the next step in friction-reducing shopping innovation. As much as we all love e-commerce, the fact remains that some transactions are still difficult to complete through a website or app. For many categories of goods, the amount of customization, nuanced understanding, or the need to compare similar products means digital purchasing isn’t that simple; it’s a chore. That chore goes away when you can type in exactly what you want and have the machines be smart enough to provide the exact right product solution. I suspect that one day Amazon will be just a big chatbox.
But is that really Amazon’s future? The truth is that as powerful as chat-shopping is, we’ve seen this promised utopia before, in the form of Alexa. The original vision for Alexa was a helpful assistant who would take your impulse order: “Alexa, we need more toilet paper.” And boom, toilet paper at your door hours later. In hindsight Alexa was too dumb, delivery wasn’t cheap and fast enough, and humans weren’t quite impulsive enough, for that future to come to life.
But I’m optimistic that “Take Two” for AI commerce will be the one that sticks. Large language models (LLMs) are fundamentally more sophisticated than state-of-the-art AI circa 2014. Messaging with machines is much more natural than talking, and the success of WeChat in China means it is only a matter of time before similar experiences take hold here.
Where will conversational interfaces first take hold? The obvious applications are in the categories of products and services that were typically purchased, pre-internet, with assistance from humans. Think travel, telecommunications, automobiles and insurance. In each of these cases, significant research is required to make the right purchase, and the purchase process itself has lots of configuration, customization, and “paperwork”. These are all things that are now somewhat tedious on websites but would be much simpler if you could just tell a machine what you want, and have AI infer the rest.
But the applications extend to all product categories, especially when it comes to the challenge of product selection. Companies like Best Buy thrive, in spite of its commodity offerings, in part because the salespeople are experts. That expertise can be simulated with conversational interfaces. Likewise, products as diverse as cosmetics, over-the-counter pharmaceuticals, and even clothing could benefit from a conversation. Search “Monopoly board game” on Amazon and you get over 1,000 results—turning a trivial purchase into a twenty minute research ordeal. Search Rag & Bones Men’s Jeans size 30x32 and Google Shopping does a decent job returning results… but a few random brands still pop up, and I’m forced to slog through a product grid and filter and debate to find just the right item. A layer of intelligence above search results would certainly be beneficial.
Hunting through Amazon or Google Search makes me wish for the days when I’ll never have to spend hours on a product hunt. I look forward to just texting a bot.