What makes Chatbots unsuitable for complex actions?
A sequential chatbot can seriously hamper your aspirations around providing a better CX
Chatbots, today, have emerged as an extremely useful way to enhance, among other things, Customer Experience (CX) and increase the overall operational efficiency of organizations. They have the capability to support and scale business teams and provide the scope to automate a lot of manual processes of today.
But isn’t it a little too early to bank wholeheartedly upon the efficacy of chatbots to give a leg up to your CX, operations or even sales efforts.
The idea of automating and scaling personalised communications is very appealing to organizations. But try to remember of the times when you were at the receiving end of automated responses. How did it feel to converse with a bot, knowing for a fact that all you’re going to get are canned responses, without any human touch?
In case that didn’t turn out to be a very likeable experience for you, chances are that you’re among the 73% of users who’d never engage with a chatbot again.
Now, chatbots are mainly of two types,
- Based on AI, where they’ve the capability to learn while communicating
- Sequential or Rule-based where they’re programmed for only specific pre-defined scenarios
Anything which requires a deeper and more contextualized involvement, such as, providing high level recommendations or screening candidates for recruitment, cannot be achieved by regular sequential chatbots. AI chatbots will be better equipped to save the day for these use cases, as they can factor in contextuality while responding and even learn from user behaviour as they go about carrying out the processes.
But what if you do not have that choice? What if your only option is to use regular sequential chatbots to solve your challenges?
Let’s take a look at the various ways a sequential chatbot can seriously hamper your aspirations around providing a better CX.
- The Curious Case of User Intent
For chatbots, understanding user intent is the top priority followed by immediate resolution. Now, this could be anything from supplying useful external intelligence to providing access to personal information to giving perfect recommendations. It’s only when a chatbot is able to accomplish these things, that it’ll be able to give the best results.
But chatbots are built on a decision-tree logic and their responses depend on keywords identified in user’s inputs, which might not always display the precise intent of the user. This leaves very little scope of accommodating deviations away from a programmed script and as such, can result in unsatisfactory experiences and even loss of revenue, if employed in complex scenarios.
- That Formidable Case of Attention Span
With a drastically dwindling attention span of millennials today, chatbots are posed with a real first-world problem — keeping the users engaged while at the same time, providing enough value for them to stick.
If you think about it, most of the chatbot conversations don’t go beyond 2 or 3 messages. A user asks a question and the chatbot answers it — Case closed! But the situation starts getting more complex when there’s an actual problem to be solved and not just one-line questions.
Capturing the context of the conversation, nuances or even user’s sarcasm is hard or impossible today with sequential chatbots. Even more straightforward items like plurals need to be thoroughly defined or imported to give the chatbot a full understanding of the situations it could encounter with users. On top of that there’re other issues as well.
For example, how long should a chatbot wait to respond, if the person is in the habit of writing multiple short messages while texting? Replying midway might mean a not-so-clear understanding of the problem entailing convoluted answers. Waiting for too long to finish might result in the customer getting annoyed and quitting the platform.
- The Millennial Way of Texting
If there’s one thing that has truly evolved in the last 10 years, it’s how people text. Abbreviations — un-recognized abbreviations — spelling mistakes — Hinglish (the seemingly most popular Indian language today) — typos — the local slangs, and what not.
Our ways of texting has changed radically and although it’s still okay while talking to humans, we really don’t care if our words are suitable enough for a machine to understand.
But machines are not the only ones at fault. This is exactly how millennials talk today!!
We want the support agents to show empathy for our problems and ensure us of prompt resolutions. And this is achieved by the combination of two things — words & emotions. With the right set of words, one can actually calm down one’s irate customers, in a jiffy. But what if there’s a chatbot trying to do the job?
Chatbots can’t showcase sympathy for the problems or exuberance for the resolutions (something which caused Microsoft, a great deal of bad publicity). How much do you think people are going to enjoy such types of conversations?
- Escalation Workflows
I’ve already talked about how chatbots are programmed for only repetitive actions and can answer only what has been stored in their database, without applying any intelligence. Any out-of-the-database question would mean your chatbot dropping dead. But what then? Your customer is still waiting on the line for their resolution.
Well, you ask your human crusader to take charge but by the time your crusader says Hello! to the customer, he / she realizes that the line has been disconnected. Congratulations! You just lost a customer and potentially, some revenue as well.
Chatbots, very often, do not have a proper escalation workflow to let humans take over when they fail. And when they do, the transition from a bot to a human is often marred with interruptions.
Hence, chatbots need to not only have defined escalation workflows but the transition needs to be very seamless as well, so as to not hamper the customer experience. Because if the customer is left wanting more, without any further useful responses, they might forever close their doors on you.
- Poor Memory
Chatbots have often been accused of having a poor memory. They’re unable to memorize past conversations and as such leave the users having to type the same queries multiple times. And this, pretty obviously, is an annoying situation for just about anyone.
Now, it goes without saying that having a powerful and intuitive memory can be of great help to chatbots, but sadly, this isn’t the case yet for our rule-based butlers. While speedy resolutions and understanding user intents might be the topmost priority for chatbots now, the capability to understand and take lessons from past instances, for better articulation of solutions, can go a long way in establishing chatbots as a reliable technology partner.
Any new technology needs a good amount of time to perfect itself and reach the required proficiency levels where it can walk side-by-side with humans. And similarly, even chatbots will also need to go through multiple advancements to be able to understand the nuances of humans.
But let me assure you by saying that the work has already begun!
The problems aforementioned are primarily with sequential or rule-based chatbots only. The game is totally different when you bring AI chatbots into the picture which have a much higher efficacy in handling customer engagements today and give you far lesser pain points to dwell on.