We all hate calling helpline numbers. IVR (interactive voice response) is painful. Listening to the million options that you don’t care about is frustrating.
IVR: Marathi madhe suru thevna sathi krupiya ‘1’ daba. To continue in English press ‘2’. Hindi ke liye ‘3’ dabaiye.
You press 2.
IVR: We bring you a host of new services. (Some promo spiel goes on and on and on.)
For billing press ‘1’. For tariff plan-related information, press ‘2’. For data related information, press ‘3’. To manage your account, press ‘4’.
Now let’s say that you’ve lost your phone, want to cancel your sim and are panicking. What’s worse, you’ve already been on the phone listening to useless information for over a minute!
For sim loss, MNP and other information, press ‘5’.
You finally press 5.
The high cost of having a human being at the other end of the line is what makes IVRs so tedious. Companies want few people to speak to the agents. But speaking to the agents is a better experience than IVR, especially if you get a sharp person to resolve your problem. But more often than not, the agent’s low-quality English, a bad network and the person’s inability to understand what you want can be quite exasperating. So what do we do? Say hello to the chat-bot.
The ideal chatbot should be able to understand what you’re asking and give you the information you need. Companies take note of the usual things people have problems with and program the bot to customise its replies for different queries. This is ideally supposed to be win-win for both the customer and the company. The customer can just type in what he or she needs to know and the company doesn’t need to employ a human being do this job.
But things are never that simple. Just like people talk differently, have different accents and pronunciations – they also phrase sentences differently and have a different understanding of the grammar. Most websites in India are now in English, the laptops have English keypads and even those who use Hindi online have to use the Roman script. This makes things very complicated.
Taking all this into consideration, when it was time to improve service for Kosha’s online customers – I chose Facebook messenger. The app appears on the website, aping the chatbot’s look but there’s a human behind it. From Kosha’s experience, the chat window enables the customer to reach us directly. That kind of direct access can’t be imitated with the phone and only to a limited extent with social media. Through chat, we’re able to gauge what kind of questions people have and what destinations they’re travelling to. This helps with both operations and communication strategy.
What I’ve realised is that clients love to talk and be heard. I personally think the human approach works well, especially for the volume we’re at. We’re documenting all the questions we’re getting on chat and will work towards automation 12-16 months from now.
After speaking to fellow entrepreneurs who’ve made a decision between HI (Human Intelligence) and AI (Artificial intelligence), I realised that I probably did the right thing. Some of my contemporaries said the usage of the chatbot on their apps is extremely low even though they’ve provided all sorts of services – two-factor authentication, cost-calculations, and ability to book and cancel. But the worst part of their experience has been that those who use the chatbot end up with incorrect bookings! While the customer is probably messing up while replying to the bot, the bot ends up being rejected. Hello again – human customer service!
The main intent with bots was to deal with a customer waiting time – a problem that I highlighted earlier. But while the bot has all the answers, it, unfortunately, doesn’t have the ability to understand all the questions. Another problem with chatbots is that, like most machine learning algorithms, natural language processing also requires a huge dataset to work with. This creates a chicken and egg problem where your service won’t get better till it doesn’t have experience and it won’t get experience till the service doesn’t get better.
Chatbots are cheaper than human customer service only if it is your first time with customer service implementation and you’re starting out with a bot. However, if you already have a working setup, switching to a chatbot system becomes a hard sell. The best way to go about it (in case you already have a setup) is to migrate to the chatbot in steps. There is also the additional problem of people spamming the bot. All this spam adds to the dataset and your algorithm won’t improve by replying to it. In order to ensure that your data set is similar to a genuine customer’s requests – you’ll need human intervention. Hiring machine learning professionals is much more expensive than hiring college students and training them for a few weeks.
Most of the chatbots used by Indian companies are one of the below categories.
This type of bot can’t exactly decipher the meaning of what you’ve typed. It just matches the words you’ve typed to the options it has listed. Axis Bank’s ‘Aha’ seems to be in this category. If you type “Can you help me in investing my money?” It will say, “Some related queries – how does bill pay facility help me in paying mobile bills? I want to recharge fast, can you help? How does bill pay facility help me in paying my DTH bills?” The bot has basically matched my question with questions that have the word ‘help’ but none of its responses are relevant. And Axis Bank does have a wealth management facility called Burgundy – which is even listed as a separate tab on their website menu. But the bot is clueless.
A lot of people would rather message than call as this kind of service helps them. You’re chatting/texting an agent instead of speaking to him or her on the phone. Whether the costs of this are cheaper than having a phone line, your cost of employment probably remain the same, but the internet and data usage may be cheaper than having a commercial phone line or call centre and your agent’s time may be better used chatting with 3-4 customers simultaneously. I’ve used com’s chatbot to send gifts to friends in other cities and it worked flawlessly.
You’re most likely to come across this message: ‘We’re away at the moment but please leave your contact details and we’ll get back to you.’ These bots don’t really answer any questions but they create a database of interested customers for the company.
There have been all sorts of predictions about chatbots streamlining customer service. Headlines have screamed ‘Chatbots to slash business operation costs by more than $8bn by 2022’. Some research has also suggested that between 75-90% of queries in healthcare and banking will be dealt by chatbots in the US. Has all this happened in the US? No.
The chatbot’s connection with sales is not a linear one in my opinion. However, there is a lot of data in the public domain that will show you that a chatbot’s sales conversion is higher than a website’s. The difference is sometimes shown to be more than 30 times.
But I don’t buy into that argument because the customer on the chatbot is in any case more engaged than the average website visitor, and so it is natural for sales to be higher. Having the bot won’t increase online sales in a manner that you can measure, but having a chatbot can help in the sales process by the following –
- Reducing the time customers spend seeking information
- Data collection and insights – keep a track of what customers are most interested in knowing, can their questions lead you to highlight another part of your business or alter a certain part of your product or service. It’s very difficult to keep track of this kind of information when agents are talking to customers on the phone. There is so much that is not documented from that conversation because the agent is operating with a prepared script.
Taking the shopping assistant online. A good chatbot can be an ideal shopping assistant. But AI in this sphere is only developing and even globally, only a few have got it right
To Read Our Daily News Updates, Please Visit Inventiva Or Subscribe Our Newsletter & Push.