Bots and Brands: AIâs Customer Relationships Depend on Trust

by Daniel Giordan
Partner CX & Interaction Design, Wipro Digital
@giordan_UX

The author is pleased to have this article originally published by Information Age

Artificial intelligence and thinking computers are a prominent plot device for Hollywood. Movies such as 2001: A Space Odyssey, Alien, and Terminator come to mind, as well as current day examples such as I, Robot, Ex Machina, and Transcendence. These pop culture examples underscore the fears and challenges we face as we attempt to program natural behavior and interact with intelligent technology.

I think about this whenever I read the growing list of examples of how bots are becoming more pervasive, especially after seeing an article that proclaimed “the bots are taking over.” In reality, bots do not carry such hyperbolic doomsday results. The truth is that it’s very hard to teach a computer to think and process nuanced information, but smart bots that can personalize interactions are crucial for brands to successfully leverage artificial intelligence.

Bots’ rise to popularity

For those of you just coming up to speed, a bot is a software application, which runs automated Internet scripts that gather, analyze, and file information. One of the earliest bots, created in 1997, was created to track stock market trends and supposedly could predict future events through keyword analysis. Today, these automated bots, also called agents, are being created for myriad uses by some of the largest companies on the planet.

Current-day examples include (source: Engadget):

• CNN's bot, with which users can message topics like "zika virus" or "politics" and it will respond with related stories and information.

• HP built a tool that lets users print a photo by sending it to the company's printing bot on Facebook Messenger. Once a user sends a photo, the bot responds conversationally (i.e. "Hey, nice photo,") and gives printing options.

• eBay is launching a platform that allows buyers to receive price alert updates within Messenger so that they never lose an auction.

• HealthTap is now offering the expertise of its network of top U.S. doctors instantly via Facebook Messenger as a new, convenient and simple way to access health information. Anyone can type a question and receive free answers from trusted doctors anytime, anywhere.

• Spring is launching Spring Bot, a personal shopping concierge powered by the Facebook Messenger Send/Receive API, Send to Messenger plugin and Zopim live chat.

• Fandango's bot for Facebook Messenger will provide fans quick and easy access to movie information, trailers, show times, theater locations and a link to advance ticketing for all theatrical releases on Fandango.

These examples are not much more than glorified alerts and simple data and platform integration. However, the goal of these efforts is to create applications that can solve problems of much deeper complexity, and will resonate with users on a more personal level.

The key to success is getting personal

Having spent four years creating agents and semantic personalization solutions for a major retailer, I can say I’ve spent a ton of time thinking about this issue. I’m a UX guy, not a programmer, so the bulk of my time was spent creating interfaces and contexts that would encourage humans to engage with these helpful bot agents. Our team played a critical role, because user acceptance and adoption is a key factor for setting user expectations and driving the “learning” aspect of these automated tools.

The biggest challenge in building an effective bot is to solve the issue of personalization. The bot has to know and understand you (who you are and what you want) in order to process and retrieve things the way you want it. This means understanding your needs and expectations in all their complexity so that your experience is both satisfying and delightful.

Learning ‘human nature’

The most meaningful bot activities happen when they help us solve for complex decisions rather than a simple search string. Solving these problems requires a bot to know you and your situation: like a trusted personal assistant who anticipates and makes initial decisions on your behalf, the bot knows the types of answers or information you are looking for. Bots are also meant to be pervasive and ubiquitous, running in the background and surfacing whenever the appropriate data surfaces.

How does a real personal assistant learn these things? They interact with you, they watch what you like and don’t like, and take your feedback and apply it through an aggregate knowledge base that grows more nuanced over time. My teams worked on agents that tried to solve problems such as, “help me plan my daughter’s wedding”, or “teach me to play guitar.” Bots may not be there quite yet, but eventually, they will learn enough to make complex and interconnected decisions involving a wide range of variables, revealing the true promise and potential for bot technology.

Personalization happens when a relationship is built over time

Yet these solutions will never come to fruition until we solve the challenge of giving these bots the appropriate data so they can learn who we are and what we like. Yes, bots can correlate initial data such as my geolocation, initial profile data such as gender and address, and then cross-check and crowd-source data to make assumptions based on 'people like me' ­– but these approaches will not be as accurate as most of us would like. This is all the more challenging given the aforementioned media-inspired paranoia and distrust around artificial computing and computer-based personas. We need to solve this on an emotional level, as well as technical and logistical.

The companies that are solving for personalization are going to be the leaders in creating meaningful bots, because personal knowledge allows the bot to be more intuitive and insightful. Companies need to look for ways to drive repeat engagement, whether it’s frequent purchases and search queries, surveys that reveal deeper insights, or the creation of groups and communities that aggregate people with common attributes and desires, which will allow crowd-based learning.

Brands must earn trust, or die

At the heart of all of this is the need to create a construct that people want to use. People need to understand bots and engage with them repeatedly. No matter how intuitive your system is, the real benefits of bot personalization unfold over time. You may hire a savvy personal assistant, but that person will be exponentially more effective after working with you for a year than on day one. And getting this long-term traction requires a brilliant experience that meets the core needs of users. As users’ relationship with their bots deepen, they need to feel comfortable sharing their personal preferences, not be shocked or suspicious when it proves to be insightful, and be willing to build rapport with the interface, the brand, and the technology.

Does this sound familiar? For most CMOs, this is exactly the kind of dynamic they want to establish between their audience and their brand. On a broad level, the bot becomes an extension of the brand, perhaps even the face of the brand. Don’t believe me? Consider the chaos and brand implications of the failed Microsoft TAY chatbot system.

Successful bots are not a one-team effort

So if someone walks into your office and tells you they have engineering working on a bot for your company, make sure you have the right marketing and UX people engaged, and that you dovetail the project with your personalization, analytics, and customer service teams. Because in the end, the bot, AND your brand, are in service to your audience, and are only effective if you can drive engagement.

Failure to do so won’t result in an AI takeover or the appearance of a Terminator… but it will mean your efforts will not be valued, and that your creation (and brand) could be ignored.