• Transform magazine
  • June 25, 2017

Top

Open the pod bay doors, Hal

wow2AA.jpg

Artificial intelligence seems like something out of fiction. Yet its practical applications today make it an increasingly essential brand touchpoint. Brittany Golob examines branded chatbots

Science fiction is full of intelligent technologies that make humans lives easier, more efficient and more successful. Rosie in the Jetsons acts as household staff for the futuristic family while artificial intelligence (AI) inhabits and runs space ships in countless books. 2001: A Space Odyssey’s Hal is probably the most frightening example of what artificial intelligence (AI) could be.

In fiction, that future is still far off but in reality, that future may be closer than expected. In the pockets of over one billion people lives a chatbot by the name of Siri. Cortana inhabits Windows operating systems worldwide. Even Google’s responsive AI interface is helping bridge the gap between human beings and artificial intelligence.
For brands, forays into this emerging technology should be met with a sense of opportunity, and trepidation. Experts express little fear in the doomsday scenario posed by some science fiction writers of AIs taking over the world. However, the risk for brands lies in incapable execution of a chatbot or a poorly-designed AI interface that detracts from, rather than enhances the brand experience.

“Deployed efficiently and intelligently as an aid, chatbots can help augment a relationship with a brand, not least because they are increasingly prolific and increasingly practical in helping us access information, helping us make decisions and helping us open up opportunities. All of those things help us determine how we feel about and experience a brand,” Andrew Welch, managing director for London at global brand consultancy Landor, says.

But, Welch warns against the potential risks involved in the implementation of chatbots. He says focusing too much on the technology, and not enough on the content, messaging, tone of voice and usefulness of the AI will result in its ultimate failure. AI should be relevant and confident. If not, that will result in a poor brand experience and cause negative associations with the brand.

This second scenario has played out already. The shocking example is Microsoft’s Tay project which debuted in March 2016. The AI was meant to speak to Millennials, like Millennials, on Twitter. It was designed to learn from the conversations it had – not an unusual framework for an AI system. The problem was, brands typically control the input they give to AIs that learn from a body of information. Microsoft let the internet make its own corpus, resulting in Tay becoming quite the racist in only two days of operation.

The corporate vice president of Microsoft Research, Peter Lee, said in a blog post that Tay was based on the Xiaolce chatbot Microsoft deployed in China, which is still running successfully. But, he adds, “AI systems feed off of both positive and negative interactions with people. In that sense, the challenges are just as much social as they are technical. To do AI right, one needs to iterate with many people and often in public forums.”

In branded environments though, chatbots are not often used as public fora, but as problem solvers or efficiency- driven aids for both the internal audience and customers. Donna Peeples, chief customer officer at American business messaging developer Pypestream says companies should examine their processes and determine where a chatbot could provide the biggest impact. For things like call centres, this can be revolutionary. Chatbots can bring together numerous systems and swiftly provide either the caller or the customer service representative with the solution.

Internally, chatbots can work through the more repetitive internal communications tasks – like filling out pensions forms – or serve as an IT help desk for minor issues. IBM says that by 2018, at least 20% of employees worldwide will use some sort of automated assistance to augment their work. The technology for chatbots is still developing though. To provide brands with a secure chatbot, particularly in highly-regulated sectors like telecommunications or utilities, chatbots need to have guardrails. Peeples says chatbots can adhere to strict guidelines so that they are nearly scripted but are still able to solve problems and use natural language to have a conversation.

“I don’t see a way not to adopt it,” Peeples says. “The pace of global change and the technology it is enabling is only going to increase the variety and velocity of change. People and companies who don’t adapt as quickly as they can, if they’re not looking at disruption from the inside out, they’re probably being disrupted from the outside in, and they may not even realise it.”

“Deployed efficiently and intelligently as an aid, chatbots can help augment a relationship with a brand, not least because they are increasingly prolific and increasingly practical in helping us access information, helping us make decisions and helping us open up opportunities. All of those things help us determine how we feel about and experience a brand”

To adopt AI technology in a seamless and useful way, brands should consider not only where the chatbot sits within the organisation, but what it sounds like and how it interacts with the user. Decisions have to be made about naming, the voice of the bot, its tone of voice and conversational ability.

“I think where brands were once defined by their visual expression and human interactions, we’re now entering a world where brands are going to be increasingly defined by their digital voice and digital interactions,” says Lippincott’s Dylan Stuart. “That could become one of the main channels through which one would interact with a brand. As a touchpoint, it needs to be developed with that in mind.”

Stuart says brands should consider how a chatbot reflects the brand character and purpose. It could be a friendly human-like Rosie-type bot or it could be a simple interface through which humans can solve problems. Naming a bot, particularly with a gender-neutral name, can allow for a stronger connection between the user and the bot. But, a name isn’t necessary in every case, says Welch.

In creating the character of the chatbot, language is most important. The chatbot’s tone should reflect the way a brand wants to develop its connection with customers. It should also be flexible while still adhering to guidelines to avoid a Tay-style disaster. But one of the risks is that a bot may serve to distance consumers from the brand. “If you’re trying to build an emotional bond with your brand and you position it as not a true mediator, that is going to potentially weaken the interaction people have with your brand,” Stuart says.

The biggest success story thus far is Siri, which is not only a useful tool, but a sort-of friend that can have sort- of conversations. Siri has humourous scripted responses for certain questions and has enhanced the connection between iPhone and iPad users and the Apple brand.

Yet, getting the language right has been a huge part of Siri’s success. And it’s not an easy thing to do. Even without considering chatbots, developing a brand’s tone of voice is difficult, and consistency is key. Brands should approach tone of voice not as a set of values that characterise the company, but as a set of practical guidelines and writing styles that can help the organisation express itself more consistently. This tone of voice can be applied to call centres, legal teams, social media managers and marketing professionals, as well as digital tools like chatbots.

For chatbots, though, there are limitations. “You can’t really programme a bot to be clear or ‘everyday’ or down to earth,” says language consultancy the Writer’s creative director Nick Padmore. “But you can programme it to say ‘but’ instead of ‘however.’” Thus, chatbots should have clearly defined language styles, as well as the ability to use natural language. Creating consistency though, may be a challenge for companies with more distinctive tones of voice – the Virgins and Innocents of the world.

Ikea, which used a chatbot called Anna to direct users around its vast website, took the bot out of commission in August after 10 years of service. Anna’s managers found that impersonating a human was not necessary and doing so only meant Anna would get asked irrelevant questions. According to DMI, which launched the mobile version of Anna, half the questions she received were sexual in nature.
Guardrails are indeed important. “Consistency,” says Padmore, “is the most important thing.” But he adds that chatbots shouldn’t try to emulate humans, lest humans are duped by the bot. Ultimately, chatbots should be integrated seamlessly into the brand experience. Peeples says they can be reactive – call centres, internal comms – or proactive by informing customers about a service failure, for example, and offering a solution, thus avoiding customer calls.

But the branding of the bot – in terms of language, messaging and tone – is necessary to achieve this, says Welch, and content is the most important aspect. The chatbot should be programmed with the right content to solve problems to be useful in order to augment the relationship with the brand. He warns that unless this is done effectively, the chatbot will be irrelevant. “I don’t think any customer would be forgiving if the content I was using was useless or of no utilitarian value, however cool or new the technology is,” Welch adds.

The physical and the digital spaces are converging. Kids born in the last two decades already exist in a world where interactions with people and with brands are not viewed as online versus offline, only as different forms of engagement. Technology research company Gartner says 89% of companies are expected to compete primarily through customer experience, up from 36% in 2012.

Though the technology in AI needs to develop further in order to reflect the science fiction versions of technological integration between humans and machines, that gap is smaller than ever before.