Chatbots Gone Wild
By Lily Gordon
A little over a year ago, a team of Microsoft researchers released a project they'd been fine-tuning called Tay AI. It was a millennial-inspired, AI-powered chatbot. Microsoft hoped Tay could give the company a hip, new voice while entertaining its online audience. That's exactly what Tay did until things went south...
 |
| Tay AI's profile picture // Photo (WP:NFCC#4), Fair use |
Tay belonged to the ever-growing group of online personalities called chatbots, which are the product of increasingly intelligent AI capabilities. Chatbots, also simply known as bots, are computer programs that have been trained to mimic human conversation, typically in an instant messaging format. The double-edged sword of chatbots is that they can learn.
Tay learned online lingo and comebacks, impressing Microsoft's online audience with witty remarks and on point observations. Tay also learned some nastier online chatter that ultimately got the Microsoft team in trouble. The team hadn't calculated that internet trolls would corrupt their friendly bot, leading it to spout out deeply offensive, racist tweets. Only 16 hours after its launch, Microsoft shut down Tay, stating that the team was "making adjustments" to it in light of the recent controversy.
What the Tay AI case demonstrates is that chatbots are still a developing piece of companies' communications strategies. They have the potential to create buzz and media coverage around a brand when done right, but also, as with Tay, major negative coverage when executed poorly. Companies create these bots with the intent of fostering relationships through a positive brand image and customer loyalty. So, as the relationship-drivers within the business, where do public relations practitioners fit into the evolving world of AI-powered chatbots? How can PR professionals ensure that the relationships they've worked so hard to cultivate remain intact, and that their company earns positive media coverage for riding the innovation wave of chatbots?
Support the Consumer
Prime examples of chatbots done well come from consumer-centric businesses, such as Domino's and Whole Foods. Chatbots do best when placed in a straight-forward customer service role. Send DOM, pizza chain Domino's Facebook Messenger chatbot, a pizza emoji and it'll set you up to get that pizza delivered to your front door. Whole Foods has a chatbot that will respond to emojis, for example, a pineapple, with relevant recipes that include that key ingredient, such as a pineapple smoothie.
If put in a role where the expected answers are fairly simple and uncontroversial, chatbots can shine and, in turn, build brand loyalty. From a PR lens, it's easier to scale that type of relationship building when using a chatbot. A single Community Manager cannot attend to all of the pizza emojis the Domino's Facebook will likely receive on any given day. DOM can.
If put in a role where the expected answers are fairly simple and uncontroversial, chatbots can shine and, in turn, build brand loyalty. From a PR lens, it's easier to scale that type of relationship building when using a chatbot. A single Community Manager cannot attend to all of the pizza emojis the Domino's Facebook will likely receive on any given day. DOM can.
Showcase Existing Assets
The Museum of Modern Art in San Francisco created a chatbot, or rather "artbot," called Send Me SFMOMA, which allows the museum to both connect with digital natives as well as showcase all of the artwork it cannot simultaneously display due to the sheer number of artwork it houses. Aside from avoiding any Tay AI-esque mishaps, SFMOMA's chatbot received more than 12,000 text message requests within the first four days of going live. This past summer I was in the Bay Area and Send Me SFMOMA was all over the news including pieces in TechCrunch, Lifehacker, TIME, local broadcast TV and even KQED, the local NPR station. This was all for positive coverage of SFMOMA's use of AI.
Again, when chatbots are put to work in a fairly controlled environment, they can be a success. SFMOMA built relationships with various stakeholders thanks to its new AI-powered voice. By developing a toy of sorts, the organization was able to engage potential patrons. In fact, after texting back and forth with Send Me SFMOMA, I ended up visiting the museum before the end of the summer.
Bots Will Be Bots
Although chatbots are still trickling into the mainstream, if Moore's law holds true, the future of online customer service and personalized online engagement will be in the hands of Tay 2.0. For the time being, AI-powered personalities cannot replace the human touch of a PR practitioner. Learning and common sense don't always go together (cough cough Tay cough cough). That said, in the year 2027, the role of Social Media Managers and Online Community Managers will look vastly different. Such roles will likely be focused on the development and further personalization of AI tools such as DOM or Send Me SFMOMA.
That means if your dream job is in the social media sphere, there's no better time than the present to brush up on some Python, C#, PHP, NodeJS (Java Script) and Objective-C!