Fashion and electronics are the two top categories users are browsing on eBay's Shopbot, according to the e-commerce companies latest insights.
The Facebook Messenger-based chatbot, which launched in October 2016, aims to satisfy consumer demands for richer and more contextually relevant search enquiries. At the time, RJ Pittman, chief product officer at eBay, said: "Combining AI with eBays breadth of inventory and unique selection will enable us to create a radically better and more personal shopping experience for virtually anyone that owns a mobile phone.
Now, more than nine months on, the team is finding those who use the chatbot are nearly three times more likely to do so to ask questions on specific products versus browsing eBays inventory for inspiration and discovery suggesting both engagement and retention are high.
I caught up with Pittman to hear more about the progress Shopbot has made, as well as to understand his views on the future of AI and conversational commerce.
RA: Can you share an update on eBay ShopBot what has it achieved and how are you seeing users interact?
RJP: eBay ShopBot has created an opportunity for eBay to reach a new group of shoppers on one of the largest social messaging platforms Facebook Messenger, which has over 1B monthly active users.
We have been testing a number of new personalized shopping techniques in beta (including eBay ShopBot) whereby shoppers can save both time and money. In eBay ShopBots case, shoppers can enlist the bot to do the heavy lifting of deal hunting and discovery for them, through a simple conversation or a photo.
eBay ShopBot uses deep learning algorithms paired with natural language understanding (NLU) and computer vision to help users express their shopping needs naturally. eBay ShopBot can even create a size profile so that all of our curated collections and browsable inventory is presented in just the right size.
RA: Theres been a fair amount of criticism that such bots aren`t the best user experience, what are your thoughts on what needs to improve?
RJP: There is definitely still work to be done by both the messaging platforms and brands to ensure bots are advancing, to make the conversations with shoppers meaningful.
The user experience and interface of messaging platforms were not initially built for search, discovery or purchase. This is an area where improvements can be made. The messaging canvas where chats occur can be laid out in new and different ways to support product browsing when shopping versus using a single column flow for sequential chat messages. Also, we want to make it easier to discover and learn that great shopping bots exist on these large platforms.
As the bot industry matures, so too must the platforms that support them. Organizing a directory, catalog, or bot 'store' not unlike an App Store is on the horizon. This will help tremendously with the exposure and more casual adoption of bots in messaging.
Bots are still a nascent platform for commerce, but the more bots learn from user behavior, the faster they can evolve and adapt to create a seamless and natural experience.
RA: How are you seeing the technology becoming more intelligent?
RJP: True artificial intelligence is the key. NLU is vital to capturing shopping intent, from which a rich shopping experience can occur.
The vast majority of bots today simply pass keywords from a user's message as input into a search engine, which then returns search results in a messaging window. This method does not build trust or confidence between a bot and a user. We need to develop a relationship with each customer, a personalized experience that will strengthen the more a user interacts with the bot; this requires deep learning, in addition to natural language understanding. We are seeing growth here, but its still the early innings.
Also, computer vision algorithms continue to improve the more things they can 'see' and learn from. At eBay, we have billions of descriptions, billions of images, and billions of items. We are learning very quickly how to recognize the worlds products visually.
RA: If Gartner predicts that by the year 2020, 85% of customer interactions will be managed without a human, what do you think needs to improve to make that viable?
RJP: In order for this to occur, we need to arm our commerce systems with more self-service information, to enable our customers to find answers on their own. This is not a challenge exclusive to AI, but rather a combination of automated customer service technology and a more comprehensive knowledge base that covers the majority of issues that customers face today. This two-pronged approach will help accelerate our path toward automation and self-service delivered through bots.
RA: A big theme were hearing discussed around AI is the idea of empathy and emotional intelligence how do you think that can come into chatbots?
RJP: If chatbots are powered by the same AI that is investing in creating algorithmic human consciousness, they would benefit from these advancements. Chatbots powered by keyword search engines will obviously not benefit in the same way.
Put simply, a chatbot is one of many endpoints for AI. Once the AI is capable of reasoning and empathy, it can be expressed through a variety of end points or interfaces, such as IoT devices in your living room or in your car, as well as a chatbot. All boats will rise with the tide as we continue to develop more human-friendly AI capabilities.
RA: Can you tell me more about the Google home version of eBay ShopBot?
RJP: Voice interfaces are very powerful and will play a role in our future. There are simply too many use cases wherein access to information by looking at a screen is either not possible, not safe, or inconvenient.
We are in the early stages of this new frontier, but we are excited about the prospects of bringing the eBay shopping and selling experience closer to our customers everyday lives. The Google Home prototype is just that. Initially it will help you to find out what something is worth, just by asking: "Hey Google, let me talk to eBay." From there, consumers all over the world can quickly take inventory of just how much value they are sitting on with respect to items they could readily sell on eBay.
And of course, with eBay ShopBot, shopping with Google Home will be possible with just a simple conversation. These devices will help make our online shopping and selling experiences more convenient keeping us up to date with great deals and finds, letting us know when an item is scheduled to arrive and even alerting us when we've received a great offer on something we listed for sale. All without needing to log into a phone or computer.
RA: Whats your broader vision for AI-driven commerce?
RJP: AI will finally enable truly personalized shopping in the way that today we see delivered by the best human sales assistants those who build trust and confidence with their shoppers. For nearly 20 years running, I've yet to see online recommendations come close to the quality and efficacy that can be delivered by sales assistance in physical retail stores.
While AI will certainly unlock many new capabilities in this regard, it isn't the only thing holding back breakthroughs in personalized shopping online. When companies reach large scale, they sometimes lose sight of key attributes, such as the intelligent use of all their data, and a more conscientious effort to be customer-centric. Investment in better data science and data management tools, paired with the low cost of storage, substantially increases the surface area from which we can derive new insights for our customers through deep learning and broader AI capabilities.
RA: Do you believe this sort of interaction is going to be the norm in the future?
RJP: There is a permanent role for conversational commerce in the future. My view is that the new norm will be situational; some shopping occasions will take place without a screen (e.g. voice interfaces), some will be a hybrid experience, which could include chatbots, some will be in virtual and augmented reality and some through traditional 'self-serve' online experiences without the need for assistance or conversation. The science of commerce converges on new tipping points of capability that will truly change the customer experience. It is an exciting time to be in e-commerce.
Rachel Arthur is a business journalist, innovation consultant and the founder and editor of Fashion & Mash, a daily news site covering the intersection of fashion and technology.