How to design brands for chatbots and virtual assistants
If you believe the hype, the way people deal with services and brands is going to involve a lot more conversation, and lot less clicking and tapping. Currently, if you want information about a product, service or brand – or to get it to do something for you – you’ll likely use a website or app. The same is true to take part in marketing campaigns and experiences.
Brand AI – chatbots you send messages to like Domino’s Facebook Messenger Bot and intelligent assistants you speak to like Amazon Alexa, Apple Siri and Google Assistant – are designed make it easier to find the information you want easier. Ask them a question or issue a command in a human way – “Siri, when’s my next meeting?” or “Alexa, play me the new The XX album” – and they’ll save you the time and effort of looking at lists of options or finding the right app. They’re not ideal for everything – browsing is better done the traditional way – but the more certain you are of what you want, the more useful they can be.
And they can be used purely for marketing/fun, as with this – ok, rather dumb – Facebook Messenger quizbot for the latest Resident Evil movie.
AI’s like these affect how consumers perceive brands in the most obvious way – by making things easier, people like the brand more – but their personality can also affect how people perceive the brand behind them,
“When we think of AI – or at least when I do – what we want to is ultimately to humanise a company, product, or service to create a desired perception [of that brand],” says John Creson, principal at the San Francisco office of Designstudio – which has branded the likes of AirBNB, Deliveroo and the Premier League. “Just having a voice to communicate with, it certainly does make it a more intimate human experience – compared pushing buttons and going through apps.”
John says that AIs fundamentally change the way you communicate with brands – from a purely transactional exchange to more of relationship that you trust.
There is a question of who you’re having the relationship with though. While you interact with chatbots on Facebook inside that social network, the brand can use imagery in the conversation to ensure you’re fully aware it’s the brand you’re communicating with. With intelligent assistants that speak and listen, it can feel like you’re always talking to them rather than the brand. You have a relationship with Alexa or your preferred gender of Siri, even if you’re speaking ‘to’ Uber or Airbnb or Tesco.
John wonders if this matters if the primary objective of communicating with a brand in this way is to engender trust on the part of the consumer.
“Say I wanted to make some changes in my financial planning,” he says. “Would I need to have a relationship with the bank in that moment? Or would my relationship with the likes of Alexa be so solid that it would understand my financial goal, and therefore I would trust implicitly that it would interface with the bank in a way that was in my best interest?”
Would we ever trust virtual assistants to choose our banks for us – trusting them to make the right choice for us and not be influenced behind-the-scenes by commercial deals done between their makers and specific banks? If not, then the place of the AI is just to live up the customer’s image of the brand created through traditional branding, marketing and advertising – replicating the persona of a store assistant or call centre staff.
The difference that an AI has over most real people is that through the data they have on you, they ‘know’ you very well indeed.
John says that “the ability of these things to really quickly be able to give you more personalised, preferred recommendations will seem pretty amazing.
“My relationship with Nike will be consistent in the way that I want to relate to it – but that could be very different from the way that you want to. It’s based on your personalised preference. The ability to be kind of personal at scale and have the variability in terms of being a little more aligned to you, I think is fascinating.”
This leads onto the question of – within the constraints that you have – how do you decide what kind of personality do you give a chatbot? Do you ask a bot to represent the brand using the same parameters you ask of a human representing the brand – whether in a store, call centre or on social media?
“For me, it starts fundamentally in the same way that we would do it today,” says John. “You define what the purpose, values and personality of the brand are. Then comes to decisions about ‘How does it interact? Is it sassy? Is it reserved?'”
And if bots can adapt to your personality, then they could adapt theirs to yours while maintaining the overall brand voice – like a department store assistant who changes how formal they are depending on the customer.
We’re some way from being able to have a meaningful conversation with a bot in a the same way we do with a human. And there’s some question over whether we want to – or do we want to just bark commands at them like an overindulged aristocrat, or use the robotic shorthand we use for Google searches currently, when we type ‘Best phone’ to end up here rather than ‘What’s the best phone for me?’.
“Right now they sound so like robots. My experience with Siri is that you find yourself kind of talking in a similar kind of way to it, as you’re frustrated that it doesn’t understand what you’re saying. Alexa changed: that you heard it, you talked to it, and it responded – and you suddenly started to communicate with it [using natural language].
“I think that’s going to be interesting territory. I know with the movie Her, [filmmakers] explored that blurring of the lines. That’s going to be something that clearly will make some people very uncomfortable – but other people will jump right in. And over time it’s going to be the way by which we interact with brands in greater ways: not all brands, but a certain group of brands that are what you might call more of the iconic brands or the super brands – which will have a greater role in our lives.”
For us to be able to talk to bots and virtual assistants using truly natural language though, they need to perform sentiment analysis on what we say to detect tone – even sarcasm. And that requires a level of artificial intelligence levels above just understanding the words we say on a purely literal level.
John says until AIs can do sentiment analysis consistently, it’s best to keep it simple. People will respond better to a bot that doesn’t get when you’re being sarcastic – as that fits with their previous experience of bots – than one that thinks you’re being sarcastic when you’re not.
Brands have bigger concerns about AI than misinterpreting what users say. They’ve all heard about Tay – the millennial chatbot created by Microsoft that started spewing bigoted and white supremacist comments within hours of its release.
The way to assuage these fears is for brands to start small and simple – as most projects are at this stage. And as the projects – and risks grow – then to put systems in place to deal with any bad behaviour, just as companies do with real employees. Except, unlike with real people, you can turn them off completely if something goes horribly wrong.
However if AIs get so intelligent that, as in Her, they are essentially people – we may have to re-evaluate that.