The year 2016 was one of chatbot hype. Facebook opened its Messenger platform to chatbots and all hell broke loose. Automated text-based virtual assistants mushroomed for everything from ordering stuff to help desks.
It became apparent that despite advances in natural language processing, chats with humans were too nuanced to go beyond simple initial queries. People usually engaged the bot in complex interactions until it broke down and produced hilarious responses.
General purpose chatbots were the first to die, including Facebook’s own “M". Attention shifted to automated voice-based virtual assistants like Alexa which worked within a narrower context. Chatbot-makers also started targeting large enterprises in areas like e-commerce and banking where the volumes of customer queries were overwhelming call centres.
Even these were too rudimentary to handle anything more than barebones stuff. Even though a plethora of AI startups came up to support conversational commerce, armies of human agents worked behind-the-scenes to do most of the work for the chats.
But 2020 appears to be another inflection point where B2B (business-to-business) startups hope to prove new models of enterprise chatbots work. Some of them have started to scale up.
TRANSACTIONS TAKE HOLD
Bengaluru-based Yellow Messenger claims to have 80 enterprise clients, 20 of which are abroad as it forays into South-East Asia. Its annual recurring revenue (ARR) jumped five-fold from $1 million in 2018 to $5 million last year, according to Raghu Ravinutula, co-founder and CEO.
He says it’s still early days as only 15-20% of enterprise customer interactions have moved to automated chats, but the shift is accelerating as bots get better and faster and are able to handle bigger loads than humans.
A whole range of enterprise software products have come up for everything from CRM (customer relationship management) to email marketing, sales and HR. Most of the customer interactions for these are happening on voice and text chat channels. “Our view is that virtual assistants and chatbots are just the front-end automation layer and they need to be integrated with back end tools for customer support, marketing and other functions. The company that does that well will dominate this space," says Ravinutala.
He says chatbots have entered a “transactional, semi-intelligent" phase, after the initial period of simple decision-tree models and FAQ (frequently asked questions) content. They’re transactional in the sense that they can solve a customer’s problem or enable a shopper to buy something.
They need to be trained to enable these transactions from customer queries, so in that sense they’re not fully autonomous. But they’re smart enough to hand over a conversation to a human when it’s beyond the bot’s ken.
“Last year is when enterprises started seeing significant tangible results from chatbots in reducing customer support cost, converting more customers and higher growth from the same marketing spend," says Ravinutala.
This is reflected in Yellow Messenger’s 2019 year-end numbers: the number of times its bots loaded crossed the one million mark compared to 10,000 a year earlier; the bots exchanged more than 4,00,000 messages daily compared to 4,000 a day the previous year. With 3 million unique users on chatbots powered by Yellow Messenger last year, it claims to be the world’s largest AI-based conversational channel for enterprises.
Going forward, chatbots will need to become more artificially intelligent so that they can learn better on their own. This will reduce the resources required for training and retraining the chatbots.
Another area where chatbots need to improve is their emotional quotient. They need the ability to pick up cues about the state of mind of the human they’re interacting with. Lack of empathy with a customer’s anger, frustration or sorrow breaks down many machine-human interactions.
With more conversational data becoming available to enterprise chatbots, they will become better at handling these chats over time.
Saurabh Sharma, head of digital innovations at Bajaj Finance, is seeing this unfold first hand. The company deployed its BLU virtual assistant powered by Yellow Messenger on its Experia self-service channel. It handled millions of conversations with a high level of accuracy and generated new sales and revenue, he says. This led to chatbots being deployed across multiple channels, including voice bots integrated with Google Assistant and Amazon’s Alexa.
“It’s a brigade of virtual assistants, from special purpose bots to advanced AI bots. The BLU army is learning and growing every day," he writes in a LinkedIn post.
Chatbots had their comeuppance after the initial fanfare. But now they’re finding their feet in the enterprise software market.