To Bot or Not: How to Make a Virtual Assistant Truly Efficient
Chatbots seemed to spring up everywhere in the years 2016–2018, especially in the US. To this day, new chatbots surface often enough, yet disappear just as quickly. Many leading international brands have abandoned their chatbot experiments, preferring to return to traditional methods of communication with their audiences to improve customer experience. Here I will try to explain why this happens and give a few guidelines for the successful implementation of a chatbot.
Global studies and reports unanimously agree that, in the near future, chatbots will be an integral part of our daily lives, able to solve practically any problem. But public opinion on these assistants is split into two camps: those who consider chatbots to be effective, while others see them as a pointless waste of time and money. The latter seems more prevalent, and among these skeptics are companies whose chatbot experiments ended in failure. Few of them open up about such mishaps, but they are, unfortunately, a common occurrence.
Top 4 Most Popular Bot Design Articles:
Why Chatbots Fail
Some businesses achieve immense success thanks to the many benefits of bot technology, yet many brands only manage to harm their reputation after ‘hiring’ a chatbot. This is primarily due to errors in planning and project implementation, as can be seen from the following rebuttals against the two most prevalent arguments against the use a chatbot.
“A chatbot requires too much time and financial investments. It should be saving these resources, not wasting them.”
For any business, a chatbot means additional burdens and responsibilities, and one should hardly expect immediate results from the launch of a chatbot. No project exists on its own; there must always be someone within company who is focused on achieving the brand’s objectives and reaping the most benefit from each decision and action within the project.
Without such determination, a chatbot will quickly die on you. Furthermore, due to lack of engagement, communication with the contractor regarding the terms of reference or the knowledge base of the bot’s responses stalls, and deadlines constantly shift. Eventually, even agreed-upon measures stop being effective.
“Customers complain about the chatbot and leave, which harms our company’s reputation.”
If you acquire a chatbot just for the sake of having one, launch it and run with it, it will only lead to disappointment. The most common train wreck conversations with chatbots look like this: in answer to almost any question, the bot redirects the client to a live operator or replies so obscurely that it borders on incoherent. As a result, the client, just as before, ends up speaking with a human consultant or turns to competitors in search of other solutions.
There’s no magic trick here. If the chatbot is unable to answer the customers’ questions, then it simply wasn’t taught to do so. This can be due to several factors, ranging from a series of small missteps to the choice of an inept contractor.
It ultimately boils down to this: developing a chatbot is a difficult task. Inserting a few strings of code to appear as a pop-up window on the website won’t cut it — the bot needs to undergo training. After all, a chatbot is a type of artificial intelligence: it requires knowledge about the world, about itself and its intended area of expertise, as well as basic communication skills and access to regularly updated content.
If full-time developers or the chosen contracted company fail to provide the bot with everything it needs, it will be impossible to create and maintain a full-fledged virtual assistant, much less an efficient, cost-effective one. Negligence won’t only yield a faulty chatbot, but can also drive customers away from the brand.
It’s important to note here that, apart from business-related matters, users often try to talk to chatbots about irrelevant topics, and this is perfectly normal.
Communicating with a bot is uncharted waters for virtually anyone, and many are at a loss when faced with a conversation with AI. Imposing images of Skynet-like superintelligence cultivated by movies and literature come to mind, which prompts people to test just how smart this instance of artificial intelligence really is. If it turns out that the chatbot can’t manage anything above primitive replies, the client is left disappointed.
Of course, it isn’t essential for a brand’s chatbot to be able to make small talk when its primary focus is to perform its basic functions correctly. However, practical experience shows that to get the highest competitive advantage, a company needs to do more than simply provide its customers with useful and accessible information. The interaction with the chatbot should be pleasant and, in some ways, even addictive, the way online gaming and entertainment services are. A brand can only get the most benefit from a chatbot if the latter can keep the clients engaged.
The Chatbot Roadmap
First and foremost, the company must determine whether it truly requires a chatbot.
For instance, it may be completely unnecessary when a company wishes to automate a business process which doesn’t entail communication with its clients. An example of this is a navigation function: this can be done with a chatbot as an interface which launches the search, but the chatbot itself isn’t essential here — the quality and navigation capabilities are.
To ensure that the development of the chatbot isn’t a waste of time and money, follow the algorithm below.
Step 1: Analyze the Target Audience
Divide your customers into groups by the type of communication channels they use and by the questions they ask most frequently in said channels. This will show which part of your audience will be using the chatbot and which purposes it must serve.
As a result of this analysis, you may find that your brand is excluding the younger clients who visit the website in search of something specific or wish to ask a certain question, but don’t understand how to do it. Most of them may not decide to call the company directly if they don’t find an answer because they’re more comfortable with text communication. In this case, a chatbot can be a good alternative. It can consult clients directly on the website or on WhatsApp, Facebook Messenger and other IM services.
Pro Tip: Don’t direct your clients to communication channels they rarely use.
There are several ways to find out which questions potential customers have:
- Consult the sales department;
- Analyze chats and social networks;
- Conduct a survey of current customers;
- Analyze your competition.
Step 2: Define the Chatbot’s Core Functions
Decide whether the bot should be able to collect additional information about the clients, determine their location, conduct surveys, collect feedback, check order status, etc. And keep in mind that the development of the different functions directly depends on the chosen platform — a website, a messenger, an online chat or an internal system.
Step 3: Design the Chatbot
Just like books are often judged by their covers, chatbots are judged by their designs.
Research shows that if there’s a face behind the chatbot, along with name and an intriguing backstory, customers react much better to it. It’s not surprising, given that the alternative is a bland, faceless chat window. Another thing to consider is the bot’s style of communication.
It can be business-like or more informal and talkative; this depends on brand positioning in general.
These tweaks lead to people projecting personality onto the chatbot and unconsciously seeing them as a living person, which serves to put them much more relaxed and open during interaction.
Step 4: Develop the Knowledge Base
The knowledge base, aka the ‘brain of the chatbot’, its database of answers, as well as the list of potential queries in different variations are all essential components of chatbot development.
We’ve established that in addition to basic knowledge about the company and its products, the chatbot should be able to engage in small talk (speak about itself, the weather, politics, etc.) and have the ability to respond to both complements and insults. With these skills, the chatbot can placate an aggressive or dissatisfied customer, and even if the latter ends up switching to a human operator anyway, they will be in a much better mood than after a failed conversation with a text box.
Step 5: Develop the Chatbot
This can be done using ready-made platforms or by creating a unique chatbot. The choice between the two depend on your priorities.
The first method is fast, easy and cheap, but suitable only for chatbots intended to execute standardized tasks (e.g. placing orders for pizza or booking a cab). These bots are a straightforward matrix of questions and answers, lacking true intelligence and the ability to understand the context of the conversation. The development of a more complex virtual assistant takes time, effort and is significantly more expensive. Nevertheless, it ensures that the chatbot is tailored to the company’s needs, and its knowledge base will be, for all intents and purposes, unlimited.
Step 6: Do a Test Run, Repeatedly
Even if you’ve considered every little detail during the development of the chatbot, there’s always a chance you missed something important. It’s crucial for the chatbot tested before it’s properly launched. Depending on the number of errors detected, testing can take up a week or even a month.
Either your employees or a target group of current customers can talk with the chatbot and take note of its errors. This can also reveal hiccups in its interface and point to new ideas for enhancing its performance.
Step 7: Don’t Abandon the Product
After the chatbot is launched on all selected platforms, it must be constantly monitored for errors in its dialog logic so they can be promptly fixed. This can also help to identify popular queries that aren’t included in the chatbot’s knowledge base, to further train the bot and to collect and analyze client feedback.
If you skip this stage, the whole venture will be rendered meaningless. Past experience shows that chatbot projects which forgo monitoring are soon abandoned, with no second chances or further support. After all, chatbots really can do more than we expect from them. But they can only truly be efficient if we first define an effective and clear-cut strategy for their development.
This article is written with the help of Nataliya Lemeshevska, CMO at Nanosemantics Lab.
Don’t forget to give us your 👏 !
