Chatbots? Those dumb automated things on messenger? Seriously? Yes. While chatbots may not be in the same league as apps, they can do a lot more today than the ones you probably tried out for fun when they first emerged. From ordering you a pizza to counselling you for depression, bots have long surpassed the stage where you could dismiss them as just a fad. To give you an idea, the global chatbot market is expected to reach $1.25 billion by 2025, growing at a CAGR of 24.3%, according to a new report by Grand View Research, Inc. And why wouldn’t it? There are quite a few factors that have pushed them to popularity:
- A dominance of messaging apps: Chatbots reside within most of the apps you already use – Facebook messenger being a notable name. And if you look at numbers, usage of messaging apps has surpassed usage of social media this year.
- App fatigue: Numerous studies have indicated that people actually use few apps, despite phone storage and processing power increasing every day. Repeated updates are not helping either.
- Chatbot support by big names: Thanks to the support from companies like Facebook and Microsoft, it is now possible to develop bots for Facebook Messenger, Skype, Slack, Telegram and some other popular messengers.
- Lower costs: Chatbots are easier to develop and maintain compared to apps, being almost entirely server-side applications. On top of that, the same companies mentioned earlier have made available quite easy to use development tools that utilise their sizeable research into the area.
I’m convinced! I want to work with Chatbots
Hold your horses! We know that impressive investment statements and industry forecasts look quite lucrative – but a better way to decide whether you want to be a part of this industry is to understand which skills are needed for the area. And to do that, you need to know what chatbots are made of.
When we come to the backend, there are quite a few systems working in tandem. “There are APIs which consume the message, databases which provide easy access to data and an intelligent layer of Natural Language Processing which processes the message and figures out a reply to the user”, Swapan explains, “These systems and algorithms can be built using various languages; for instance, we mainly use Python at Haptik. For NLP you can also write your own algorithms using various different open source libraries such as scikit-learn, nltk and many others.”
Apart from technical skills, there are a couple other areas that are quite important to a good chatbot experience – an education in linguistics, psychology, or sociology could also provide a perspective that is needed for a good, and immersive, conversational experience.
|What about voice bots?
The smart assistant segment has been quite popular because of consumer devices such as Google Home, Amazon Echo and more. Voice-based bots are not too different from chatbots – in fact, they mostly comprise of an additional layer of services and APIs that convert text to speech and vice versa. The key to understanding different accents and speech speed itself is a distinct machine learning problem that deserves its own piece.
But where do I work?
Globally, chatbots are entering every segment of technology imaginable. But when it comes to India, the market is quite nascent, which is a good thing. There is promise in a few areas at the moment. “The most popular bot use-cases include marketing, sales and support,” says Beerud Sheth, CEO and Co-founder, Gupshup, “Marketing bots are helping marketers qualify leads and route to right salesperson and more. Sales bots are enabling transactions through chat and also help with product discovery and evaluation. The support bots mainly automate recurring support queries and help with handing over to human agents when required.”
|Where do I learn?
All the skills mentioned in this article can be learnt separately, but there are quite a few overarching areas that can be covered in well-designed courses. Swapan says, “Andrew Ng’s – Machine Learning course on Coursera is a great way to learn the basics of ML. Once you are good with the basics of ML, diving deep into different NLP algorithm will help you get an edge in the industry. Understanding Natural Language Tool Kit (nltk.org) is a great way to learn as well.” Another good skill to have is system architecture design, for which you can find a highly recommended course from MIT here. And if you want to get your hands dirty right away, without too much coding, you aren’t short of options either with Dialogflow from Google and Messenger Platform from Facebook for basic bot creation, for free!
If you’re planning to enter this field, it is highly likely that you will end up working for a company that has its own chatbot development platform. After leaders in the field like Haptik and Gupshup, even upcoming startups have established goals of coming up with their own bot-developing platform. Indian-founded, California-based Botworx.ai, that raised $3mn earlier this year in venture funding, claims that their system is flexible and can work with third-party NLP engines in order to do things like use a voice recognition processor for voice input, or an NLP engine for an unsupported local language, or even more enterprise-usage scenarios, such as interfacing with popular NLP offerings from Facebook, Amazon, etc., and even CRMs and more.
What lies in the future?
Even with chatbots being highly promising as a prospective technology paradigm, there are already quite a few specific roadblocks in its way. For instance, discoverability on platforms like Facebook, Slack etc. is an issue, along with big platforms like WhatsApp still not being open to chatbots (that might change soon though with WhatsApp for Business). Additionally, no matter what forecasts might say, bots today are not equipped to surpass human interaction. Most of them can’t even hold their own in a simple conversation that extends even slightly beyond their intended use case. And these roadblocks are important because they might decide whether chatbots get discarded as a fad or are bettered and accepted by users at large.