Here’s what you need to know about Chatbots

Messaging platforms are coming of age, and chatbots are all the rage now, they are touted as the next app-like wave in the communication industry. Is this just hype, is the fear of missing out causing this to be a gold rush? Is it worthwhile for marketers to walk down this path to pursue the next big thing? Let’s explore.

Artificial Intelligence (AI) and Machine Learning (ML) are advancing by the day, there’s a layer of it being added to everything we know; smart cars, smart homes, smart shoes and of course messengers have become platforms on which commerce and brand communications are being built.

There are three key drivers shaping the market for chatbots. Firstly, large communication technology companies from the likes of Facebook, Amazon, Microsoft and Google are making it conducive for building chat-based products. They have opened up their respective messaging platforms for developers to build and innovate on top of it. Even Apple has announced APIs for Siri, a move to allow developers into what has been a walled garden so far.  So it seems chatbots are no longer standalone products but part of an ecosystem – just like apps tht are built and distributed in app stores.

Secondly, there’s a rising usage of Messengers over traditional SMS; recently, Mark Zuckerberg announced that 60 billion messages are sent over Facebook Messenger and WhatsApp every day, together they are three times bigger than the 20 billion SMS sent. WeChat in China has set an impeccable precedence to show the world how a thriving metamarket can be built within a chat app. WeChat is more like a virtual hypermarket on top of which other businesses have built stores through chat apps.

The third macro factor driving the rise of chatbots is the Gen Next’s growing preference to chat based interfaces. Messaging has surpassed social media to become the preferred method of communication for millennials. Mary Meeker’s research shows that 69 percent of millennials prefer web/internet chat, social media, and messaging to other channels such as talking on phone or IVRS. In the past decade we have seen brands enabling customers to do everything online, from ordering pizzas with emojis, to banking with hashtags, we have seen it all. Chatbot is the emerging new world order.

All of this gives genuine reasons for marketers to consider investing their time and efforts to building chatbots and make it an effective means to allure a new wave of customers their way. There a few reality checks and challenges to be considered though, namely:

Chatbots can’t make your business awesome: Chat apps/ bots are not a magic wand one can wave to make a product or service work.  In the foreseeable future, chatbots are only likely to be another customer touch point or an interface to existing business, it can perhaps slowly ease the load on IVRS and Email based customer support. Imagine an airline company using chatbots to assist in ticketing and flight related questions, the bot is only as effective as the company’s routes, pricing strategy, ability to fly on time and efficiency in handling baggage without losing etc., adding a bot interface cannot offset the dissonance caused by a lost baggage or a delayed flight!

The multi-agent problem: Chatbots are slowly evolving into an enabler, a digital concierge. But that evolution comes with a subset of challenges, one of them is the multi-agent problem. Consider this; Google’s upcoming messenger Allo allows users to add @google in a group chat and invoke it for getting answers or to get things done.  Similar to functionalities in Amazon’s Alexa and Apple’s Siri. Now if a user instructs the bot to book movie tickets at Inox, it has to talk to another agent (bot) that can complete the transaction, let’s say there are three agents available – inox, bookmyshow and fastticket, which one is the best to book? Bots will have a hard time determining this. Bots have limited world knowledge, they suffer from curse of choices, they grapple to understand and decode context and are under severe pressure to delight users by giving them exactly what they asked for. All of this is not just a developer’s problem, but a brand manager’s nightmare.

Compatibility and seamlessness: Besides the multi-agent problem, there’s another task for marketers to grapple with. To ensure a seamless experience for people using chatbots and smart assistants. Can we ask Amazon Alexa at home to buy a movie on Google Play, pay using Paytm and stream it on an Apple Macbook? Considering Apple won’t even let you live stream their keynote address on a Chrome browser, we can be sure there’s a long way to go before the industry agrees to usability heuristics for bots and works towards creating a common protocol (like Bluetooth, Wi-Fi) for all bots to work collaboratively for providing people an ultimate experience – like JARVIS in the Ironman movie series! 

Training the parrot and everything else: Pregnant teen girls in Australia are taking up smoking with an aim to deliberately reducing baby weight. What does this news has to do with bots? The warning label on cigarette packs were originally meant to deter people from smoking but they distorted the warning to their convenience, took up smoking deliberately as a quick hack to reduce fetal weight and thereby hope to ease labor.  This is a classic example of selective distortion in the area of communication studies.  Communication is a tough business, the way a message is written is not necessarily how people perceive it.  Just like humans have selective distortion problems, bots have context and natural language processing (NLP) problems.  Microsoft’s Tay, an AI powered chatbot went horribly wrong when it started hurtling hate sentiments like ‘Hitler was right I hate the jews’ (sic) among other tweets unquotable here.  In fact, it was programmed to learn from humans by browsing the web, it diligently did so, but the online trolls won this round.  In a world where screenshots of a brand goofing-up go viral, not all brands can’t afford such costly experiments.

AI and ML greatly depend on the training data aka what you teach the parrot to say.  There have been giant strides of progress in this field, Google’s TensorFlow, IBM’s Watson are a few examples, but given the diversity of languages, cultures, and our own distortions and bias, there’s a long evolution period for bots to grow human-like, metamorphosis from their current status of a chatty conversational assistant to replace humans and perhaps build their own Skynet!

For now, the only jobs chatbots can take away are that of a few helpless souls working at call centres with pseudo American names and a fake accent.  

Originally published in The Economic Times

Sreeraman Thiagarajan is Vice President (digital) at Publicis Beehive, as an Entrepreneur in Residence he drives innovations through the Publicis Drugstore initiative. Views expressed are personal.

Follow him on twitter at @sree_raman or on LinkedIn for 25 more free articles on marketing. 

Arrange a Conversation 

Browse

Article by channel:

Read more articles tagged: AI, Featured

Enabling Technologies