Chatbot technology is nothing new. Since the first chatbot was developed in 1966 (that’s right, bots are as old as The Beatles), the foundations of bots have been slow to evolve beyond keyword-based replies and natural language processing.

What is new and interesting, however, is the advent of messaging platforms like SMS and Facebook Messenger that provide accessible channels for brands to deploy bots. This new development comes at the same time many shopping center operators seek to (a) reduce guest services overhead and (b) appear on-trend with tech as brick-and-mortar retail steps up the competition with e-retailers.

In this article, we offer four key takeaways from the Rise Of The Machines in the shopping center space, including what you and your property can do to better engage with guests at scale.

1. Bots are far from failure-proof

Although we idealize computer programs as highly precise beings immune to the chaos and emotions of a human brain, even the best bots often fail to grasp the point of an open-ended guest message. In fact, all current chatbot vendors are selling a version of a simple keyword-based auto response enhanced with the ability to identify the meaningful parts of guest input (in a nutshell, Natural Language Processing).

Take the keywords that might lead a chatbot to send info about Sandy’s Sandwiches. A keyword-based approach would scan for words like “lunch” and “sandwich”, and ignore punctuation and extraneous words like “the” or “hello.” This keyword strategy would send the same info to each of these guest messages, resulting in fine, bad, and terrible guest experiences–

• I’m looking for a lunch place– what’s the best spot for a sandwich?
• We already ate lunch at Sandy’s Sandwiches and don’t want anything heavy, but a nice place to sit for a light bite… Any recommendations?
• I need lunch, but have a terrible wheat allergy and can’t eat anywhere sandwiches are served. Can you help?

Some operators work around this by forcing guests to send limited inputs (as in many Facebook Messenger bots), but that technology is no better than the telephone trees we get forced into when we call a massive corporate helpline. That impersonal agenda is far removed from the experiential direction brick-and-mortar retail needs to head.

2. Bots must be transparent about use case and identity

Inviting a guest into a “customer service” live chat experience only for the guest to realize “Concierge Sara” is actually an apathetic robot could potentiall be deeply insulting, particularly when it comes to guest complaints or negative experiences. Bots must be open that they are indeed bots by clearly branding the service as AI– the truth will come out anyways within 1-2 messages.

Similarly, a chatbot’s use cases should be presented as clearly as possible. If your bot is only meant to promote items on sale for a holiday, there should be no chance a guest opens the chat experience expecting to learn about accessible parking. To minimize the fallout of implementing an AI chat solution, transparency is key.

3. Human fail-safes are vital

With even the best bots regularly failing to understand the exact meaning behind a guest message, having a human on hand to avoid further frustration is a must-have. Without a human fallback, guest will continue to bombard the bot with messages when the bot fails to deliver the right information, and the bot will continue to regurgitate bad info.

Brands must offer chat guests a clear path to human help, and not leave guests hanging in a failed AI experience. While trying and failing is so often the key to success in life, this is not true in the emotional world of customer service.

4. The best solution will not be conjured from thin air

Today’s mall chatbots are typically built by a third-party vendor predicting typical guest questions, bulk importing keywords, and shipping first drafts into Facebook Messenger. However, the guesswork involved with this approach fails to result in a satisfactory guest experience.

Instead, chatbots should be built to answer the most accessible and appropriate guest inquiries, making up, say, 10% of total guest messages (with the other 90% routing to human staff). From there, bot builders and mall operators should partner to gradually increase that number, always defaulting to human aid when a chatbot is not confident in its ability to respond.

Kipsu is doing exactly this – using the 55,000,000 messages in our database to understand guest questions and construct staff responses, and allowing human staff to respond easily and intelligently using Kipsu’s powerful conversation tools. With Kipsu’s head start in the space, we will enable our industry to finally burst past the 52-year-old basics being reconfigured into new messaging channels.

‍