Another A.I. offering out of the IBM Watson stables, this time the ELF Chatbot at the Mall of America in Minneapolis. The Mall boasts over 520 stores and restaurants, an indoor theme park, a 1.3 million gallon aquarium and its own wedding chapel. Forty million visitors a year walk the 4.2 million square feet each year. And they now have a chatbot. 1080bots keeps up with the latest ecommerce chatbots to help our clients create brilliant omnichannel experiences.
E.L.F. or Experimental List Formulator is a chatbot created by agency Satis.fi to showcase IBM Watson’s Artificial Intelligence Technology. The bot begins by asking a few simple questions and provides buttons for a quick and easy response. The bot’s logo pays homage to Snapchat, a curious feature for a Facebook Messenger bot.
As a good bot should, ELF introduces itself as helping with activity ideas for visitors to the Mall of America. It asked a few questions such as children or adult and how long we’d be spending.
It then asked what we like shopping for and offered two options. As this was IBM’s Watson, I decided to skip the buttons and give it the ‘Chelsea Boot’ challenge to see if it could quickly get me to a shoe store. Not a major feat to be able to resolve the word ‘boot’ to a shoe store, but ELF immediately faltered. How did the chatbot so quickly lose its place?
I scrolled back up and pressed the ‘Fashion for Him’ button but ELF was still confused and asked me to type “ELF”. I mis-typed deliberately to see if it was able to sense simple errors, but it wanted me to be precise. We began again, but pressed the button. ELF was now lost and was unable to respond with the correct next question. I tried again with the natural language term, “Shoes” should be easy as there are “Shoe Shops” in the mall.
We went back to pressing buttons ‘ “Shopping” followed by “Fashion for Him” and the bot returned a suggestion for coffee and another to try the mobile app. Neither had active links, so I pressed the more ideas button only to receive a couple more dead-end bullets. Shame as I might have wanted to shop. Once more I tried to coax Watson with some natural language, but alas, this was not a game of Jeopardy.
Not believing the experience, I went back to the Fortune Magazine article where it said we could type natural language like “Tom Ford” or “Sunglasses”. ELF is also available on the Mall of America site itself so we tried that.
We were greeted identically and so we followed the dialog in the buttons. One of the suggestions in menswear was for menswear store Askov Finlayson so we typed that in. ELF could go no further and let me know I could chat to a live human.
I returned to the Facebook Messenger interface and asked to speak with a concierge. Chris came on the line and I let him know I was looking for Chelsea Boots. Chris got back to me within minutes with a number of shoe shops. I was more specific and asked for high-end stores and he was good to get back to me. Meanwhile, I was looking at the online store directory myself and getting similar results. Were we using the same app, Chris and myself?
Meanwhile, on the Mall of America mobile site, I was searching for Chelsea boots, nothing there, but I did try men’s boots where I was given a list of stores.
By now I was unsure of the benefits of using a chatbot, between the live person and the mobile site, I was not getting anything better. I went back to the mobile site but I think I’d just broken it.
After experiencing another IBM Watson chatbot, we should take stock of what was good and what we’d do differently:
As I said, at 1080bots we’re reviewing ecommerce chatbots to help us build brilliant conversational commerce experiences for retailers and brands to engage their customers. If you want to figure out what functions you should hand off to a chatbot, drop us a line at email@example.com