<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=283259072089923&amp;ev=PageView&amp;noscript=1">

Blog

Tommy Hilfiger Bot: Artificial Intelligence Gone Wrong

Posted by Aditya Subbarao on Oct 11, 2016 10:11:39 AM

Last month Tommy Hilfiger released TMY.GRL, its chatbot on Facebook Messenger, which seems to be the very first shopping assistant bot from a large clothing brand.

The bot was launched in an effort to promote the new Gigi Hadid fashion collection to a more digitally connected consumer base. Using TMY.GRL a consumer can browse through a series of looks or collection of products. They can then choose to learn more information about those fashion styles and add them to their shopping cart.

While this may seem like a cool experience for consumers to intimately engage with a clothing brand and shop for specific products, my experience was quite the opposite. And it all has to do with one thing: artificial intelligence (AI).

Brands like Tommy Hilfiger are experimenting with the power of AI to better understand a consumer’s habits, context, and intent in order to deliver the most relevant and engaging content, products, information, and overall experience. But what brands fail to realize is that experimenting with AI in the context of 1-to-1 conversations can often do more harm to a brand’s reputation rather than create a better experience.

I would like to walk you through my most recent experience using TMY.GRL by breaking it down into the following three categories:

1. Guided Product Discovery
2. Product Shopping
3. Human Comprehension Using AI

In each case, I will highlight both the positive and negative interactions I experienced via chatbot as well as my expected and ideal interactions.


Guided Product Discovery
The idea of being able to shop within a collection is useful since it captures a consumer’s specific interest in that collection and contains the experience within the most relevant products. When I make the decision to shop certain looks, I expect to browse for products featured in the images I am viewing.

Discovery_Process.png

TMY.GRL does a very good job with this since there are only about 7-8 looks to choose from each containing 5-6 products. This selection process closely resembles how a consumer would shop in a retail store by traveling to a section of a store featuring a certain fashion and asking a customer service rep about certain styles and size available for purchase.

When I want to learn more about an individual product, I am given info on the inspiration of the style, which materials were used, and what sizes are currently available.

Product Shopping

If a product I liked was available, I could add it to my cart, similar to how many ecommerce stores operate today. And once I was ready to complete my transaction, I was redirected to the Tommy Hilfiger mobile site with the selected items automatically added to my mobile checkout cart.

Shopping_Process.png

Although at first glance this experience looks quite seamless, my biggest issue is that if you went back and forth between the web and chatbot interfaces to add new products, the previous chat sessions would not recognize products that are already in your cart. Also, there is no way to edit or remove a product once it is added to your cart unless you manually make changes on the brand’s mobile site. Thus, the shopping experience is arguably more difficult via chat than simply within a mobile web browser.

Another area for improvement was towards the end of my conversations with TMY.GRL when I found out that products I was interested in buying were either sold out or not available in my size.

      Discovery_Fail_1.pngDiscovery_Fail_2.png

The chatbot tries too hard to be sympathetic and instead left me wondering when the item will be back in stock; In a few days, a few weeks, or perhaps even a few months? This would never be an issue in a physical store where what you see is what you get. And even a brand’s online store is usually connected to real-time inventory to avoid this problem.

Therefore, all retail chatbots should be connected to a brand’s ecommerce system to overcome these issues and ensure that all product conversations are consistently relevant to consumers. That's how we like to do it at Parlo.

Human Comprehension Using AI

Relying on AI to understand a consumer’s behavior and respond to any request is the ultimate objective in order to completely replace a brand’s customer support staff with chatbots. However, today’s AI functionality can only understand the most basic human inquiries. Anything that deviates from the promoted products or the brand is easily lost in translation...or worse...misinterpreted.


My experience using TMY.GRL’s AI started off okay with simple hellos and goodbyes. However, once I started asking about certain products, the chatbot would only address half my inquiry.

AI_Red_Dress.png

Yes...those are dresses. But are any of them red? Nope. Try again TMY.GRL.

Then I proceeded to get a little more direct with it, which obviously led to some confusion.

AI_Human_Understanding.png

I was quite impressed to see the occasional humorous response.

AI_Positive_Humor.png

But that impression quickly turned into frustration when I tried to leave the conversation, but the bot insisted I keep exploring and kept pushing more irrelevent content to me.

AI_Fail_1.pngAI_Fail_2.png

The one positive I took away from TMY.GRL’s AI was that she was programmed to always guide the consumer’s conversation flow towards a “yes” or “no” response about the promoted collection.

AI_Reflecting_Questions.pngAI_Reflecting_Questions2.png

In fact, I actually preferred how the chatbot deflected my confusing questions/remarks and guided me back to the product collections. But unfortunately, I would often see the bot attempt to respond like a fully functioning human, making it look foolish and continuing to waste my time in the process.

Final Recap

For the first retail chatbot of its kind, TMY.GRL did do a good job promoting product collections and driving consumers to a checkout experience. However, there were still several points of friction via chat that I could have avoided had I just gone to a website or clicked on an email promotion. Also, I believe that the bot’s Artificial Intelligence emphasized Artificial more than Intelligence since my questions were constantly misunderstood.


At Parlo we pride ourselves in creating more delightful and seamless consumer experiences via chatbots. Our approach is to design chatbots that supplement rather than replace a brand’s existing marketing, commerce, and customer support processes. And we do this by focusing on how to be more helpful rather than more human.


Want to create better customer experiences for your brand? See our previous post where we discussed how chatbots are changing the customer service experience, one haircut at a time. 
Subscribe to our email list for more tips, tricks, and updates on the world of chatbots. Or request a demo of our platform to learn more!

Topics: ChatBots