Among the emerging technologies in the news at present, chatbots are easily one of the hottest. At a recent survey, it was found that nearly 8 out of every 10 businesses across the world have plans to start using chatbots by the end of this decade – with these bots already put to use in some form or the other by many of the respondents. The chatbots, which simulate human interactions on the basis of machine learning, are a part of the rapidly growing application-to-person (A2P) messaging industry. In fact, the value of the global A2P market has been estimated to reach a humongous $58 billion by 2020. The demand for customized, efficient, AI-powered chatbots is very high – and the onus is on software and app developers to cater to these requirements. In here, we will discuss certain basic rules of thumb for building and optimally using chatbots:
Select the right bot platform
For smooth deployment and best performance of a chatbot, you need to build it on a suitable bot platform. Fortunately, there are many such developer-friendly, multi-featured frameworks currently available – with the Facebook Messenger Platform, Chatfuel, PandoraBots, Chatty People and the Microsoft BOT Framework being some of the most popular ones. Wit.ai is yet another bot platform that has found widespread favour among developers – with 100000+ developers onboard by this April. Both free as well as paid platforms are available (it is always advisable to start small), and there are differences in the degree of coding required, tools used (universal or domain-specific) and whether programming with SDKs is supported. Chatbot-makers have to select the platform that would be best suited for the precise type of bot application they are planning to create.
Avoid open-ended questions
A chatbot, for all its merits, is not a human being. It can ‘converse’ with end-users within a ‘fixed set of rails’ (the programming is done accordingly) – and whenever an unexpected response is received, there is every chance of erratic responses and resultant embarrassments. To keep these bot interactions streamlined, most queries to the users should be presented with alternative response options, or clear, single-phrase answers (i.e. ‘what is your shirt size?’ or ‘Are you looking for this jeans in a) blue or b) black?’. Do not let the chatbot ask any questions for which the probable answers cannot be anticipated.
Trigger emotions with images/videos/gifs
A long, boring text-only conversation with a chatbot is not a particularly exciting proposition for anyone (even if the bot under question is very prompt). Researches have shown that introducing a visual element in the conversation – with the help of rich media content, like HD photos and videos – can emotionally appeal to customers, boosting up their engagement levels. This also enhances the convenience factor of end-users – particularly when they are on a large online shopping portal. Instead of having to search for the right section and category, buyers can check out images and videos of items right inside the chat box.
Provide data analysis capabilities
One of the key drivers of the ‘chatbot revolution’ (2016 was dubbed as the ‘year of the chatbots’) has been the rapid improvements in artificial intelligence (AI) standards. With advanced machine-learning capabilities, chatbots can now collect and analyze customer data – helping in the creation of the correct promotional and marketing strategy for each user. Make sure that your AI-powered chatbot can indeed handle user-information – and provide personalized browsing/chatting experience to customers. The bot should be able to ‘learn’ from the previous browsing behaviours of a user, to provide contextual chat services and easy access to the stuff that the user is interested in.
Bot downtimes should be minimal
A chatbot should be (unlike human customer representatives) ‘always online’. This 24x7x365 availability is a big advantage of chatbots for businesses. What’s more – since chatbots (in an ideal scenario) is available at all times, they can easily replace humans at lower levels/first interaction points with customers. While setting up your chatbot on your chosen platform, find out the likely performance and the possible downtimes of the bot server. The effectiveness of a bot crucially hinges on the quality of the server on which it is hosted. If a chatbot frequently remains ‘down for maintenance’, it is not of much use.
The security factor
By the end of last year, the total volume of financial transactions performed through Google Assistant, Siri and other mobile digital assistants was nearly $2 billion. People are also increasingly providing personal information – like bank account details, card information and online orders/purchases – to chatbots. Developers, on their part, have to ensure that their bots deliver cutting-edge security assurance to customers – through proper data encryption, authentication and authorization standards, and use of access tokens (if applicable). Keep in mind that hackers are getting more sophisticated over time – and to keep the risks of security breaches at an arm’s length, the artificial intelligence of your bot needs to evolve even faster.
Chatbots should help in marketing and sales
For big companies and brands, chatbots should serve as ‘intelligent’ members of the marketing and sales teams. To enhance the chances of a bot’s success, you need to program it in a way that it can respond to common buyer queries, provide localized responses (to questions like ‘What’s the weather?’), and can influence the final purchase decisions of the end-users. The overall ‘buyer journey’ should happen in a streamlined flow – with the bot handling questions, displaying products/services, motivating users to make the purchase, and ensuring a smooth and secure payment process (directly within the bot). Chatbots can also help in the online promotional campaigns of brands, by building traffic levels and engagement in the social media pages of businesses.
Allow humans to step in
There are no limits to how a user might interact with a chatbot. You can do all the research you want – but even then, it would be next to impossible to second guess and predict all types of potential inputs from individuals. That’s precisely why these AI bots should always be backed up by human support – ready to step in and take up the conversation, whenever the bot seemingly runs into troubles. There can be an unexpected reaction to a piece of news/information, or buyers might ask for more detailed information (beyond the chatbot’s capabilities). In all such instances, a human assistant should be able to take over. Chatbots that collaborate with human employees work well – those which try to completely replace them and work on a standalone basis are likely to fail.
Focus on the end-users
Empathy and understanding are big factors when it comes to making a chatbot. Do not simply create a chatbot for the heck of it – cramming it with loads of confusing, complicated features and showing off your coding prowess. Instead, try to guesstimate what the end-users would like a chatbot to do, and design accordingly. As a starting point, list down the most time-consuming tasks (maybe browsing through products, maybe the payment procedure, etc.) and find out how these processes can be made faster and more user-friendly through a chatbot. In addition, a good chatbot should also double up as a convenient self-serving assistant – allowing people to update/make modifications to their account details without any hassles, as and when required. A chatbot that is focussed to help customers does not, generally, need major overhauls later – and that is another important advantage.
Good copywriting and a sense of humour
App developers often make the folly of emphasizing too much on the technical aspects of a chatbot (coding, framework, features supported, etc.) – neglecting the most important thing, the words used in the bot, in the process. Remember, the principal purpose of a chatbot is to give users a ‘feel’ that they are conversing with actual humans – and if the copywriting is not good, this objective will not be met. In early-2016, Tay.ai – a Microsoft chatbot for Twitter – was in the news for all the wrong reasons, and was dubbed as ‘racist’ (no mean feat for something that isn’t even human!). Also, it is important to program a chatbot in a way that it can show a nice, subtle sense of humour. For instance, the ‘Arti’ bot responds to ‘Will you marry me?’ with a smart ‘Let me see the ring first…’ riposte, while if ‘BOTbot’ is asked ‘Are you Skynet?’, it snaps back by asking the buyer whether (s)he is ‘Sarah Connor’!
Note: A nice sense of humour, along with an ability to handle angry, frustrated users (who can turn abusive) makes a chatbot more likeable and easy to interact with.
11. Make it an expert for basic repeated queries
In the United States, businesses can save up to $23 billion from their total salary expenses on customer service representatives – by employing chatbots for handling common, repeated, easily predictable queries from first-time customers. Make it a point to find out the nature of such repetitive queries that your business is likely to face, and use a chatbot that can respond to such questions independently, and without any problems whatsoever. A properly-made chatbot should be expert at performing repetitive tasks and general low-involvement activities (e.g., providing preliminary information about business/product). This expertise, in fact, is a big motivating factor for brands starting to adopt chatbots for day-to-day customer interactions.
12. Provide a personal touch
On the traditional graphical user interface (GUI) of business websites, users come across plenty of mechanical, call-to-action words and phrases (‘Order Now’, ‘Click Here’, ‘Add To Cart’) etc. Chatbots bring an entirely different mode of communication to the table. Developers need to equip bots with high-end natural language processing (NLP) capabilities – so that customers can ‘speak’ to it just as they would with any other fellow-human being. The chatbot should be programmed to add social norms and etiquettes in its conversations (‘Hello’ and ‘Thank You’, and ‘Goodbye’) as well. People prefer interacting with other humans rather than with faceless robots – and while chatbots are, in essence, the latter, their conversations should give off a nice personal vibe. Audiences would love it.
13. Pay attention to design matters
Designing the UI of a chatbot is a lot different from creating a website or making a mobile app. After the toolkit/bot platform has been selected, customize the font and typography (if the selected platform allows that) – and create a designated section or panel where users will be logging in. The bot must be proactive while greeting users (first-time visitors might have no idea on what they should ‘tell’ the bot). Focus on maintaining a streamlined conversational flow – by letting the bot understand and categorize different types of questions, validate user-data, receive and analyze critical inputs, and maintain an ‘agile conversation process’ at all times. For most chatbots, it is a good idea to present the features chronologically (new features after every successful interaction), instead of dumping all the bot features at one go. Users should never face any confusion while trying to strike up a conversation with a chatbot.
Note: If a chatbot supports speech recognition, less focus is required on the visual UI. Developers have to be more concerned about maximizing the accuracy of voice recognition in the bot.
14. Be careful while selecting channels
Developers have to do their research before finalizing the messaging platforms/services that their chatbots would support. This decision has to be taken after giving due importance to country specific features. Apart from standard text messaging, the popularity of other IM platforms have to be analyzed carefully. Facebook Messenger, Line and WeChat have very high user-bases in the United States, Korea and China respectively – and hence, chatbots in any of these countries have to support the concerned IM platform. There is a possible risk-factor in this context too – as the number of channels supported by a bot increases, the volume of separate APIs to be handled also goes up…and that can lead to errors. In general, the SMS messaging platform and the top 3-4 other messaging channels in a country should be supported by a bot there.
Keep in mind that pre-programming bot interactions and responses is different from email replies, and using a standard mail response in a bot would not be a smart idea. Developers, in fact, typically reiterate the importance of having exclusive chatbot-first use cases in this regard. Chatbots with AI support can (apart from maintaining customer interactions) facilitate real-time internal communication within businesses. You also have to enhance the discoverability of your bot, by submitting it to chatbot directories, creating Facebook advertisements, actively seeking feedback from bot communities, and bringing it on Product Hunt bots category.
With Chatfuel, it takes all of 15 minutes (at times, less!) to create a chatbot, and that too, without any coding. The chatbot technology will grow stronger in the foreseeable future, adoption rates will soar – and you need to follow the above points to ensure optimized performance from these bots.