How to Make Your Chatbot Useful
The Two Main Must-Haves For a Useful Chatbot
Updates: this is the first issue with the brand new News Last Week section at the bottom. Check it out for a curated list of last week’s most exciting developments!
Prologue: I use chatbot as the popular term to describe Virtual Assistant, Conversational AI, and the more specific Marketing Chatbot space. These technologies all differ and offer special tradeoffs—which will be covered in future issues—but for simplicity, this article groups them all in the term “chatbot”.
In the last issue, I covered what makes a good chatbot, where I pointed out that being useful is the most important characteristic. This issue, we will go over some strategies you can follow in ensuring your bot is a useful tool for your users.
Why Utility Matters
Here’s the motivation for you to read your article: Retention and Quality. If you care about these for your product and chatbot, you should care about utility.
Retention
If your user finds the chatbot useful, they are more likely to use it again the next time they interact with your product. On the other hand, if they find it frustrating, they will use other channels to perform their actions. Worst yet, if they can not find easy ways to use your product through the chatbot, they might find alternative services from your competition.
Quality
It might be obvious, but let’s say it anyway. The quality of the chatbot reflects on the quality of your product. Even though it might be an auxiliary feature, a chatbot represents a natural way of communicating with your product. Now more than ever, your user has a human-like interface to your business. Do you want that human to be useful? Yes. So how do you get there?
How To Be Useful
The first step in making sure your chatbot is useful is understanding your users and their needs. Since you already have a mobile app or website, I will assume you already conducted this research and have a clear understanding of your user behavior. Armed with this analysis, it’s time to focus on empowering your users with a chatbot that’s useful to them.
The two main methods I use for ensuring chatbots are useful is to ensure that it covers my users’ most Frequently Used Feature and Most Difficult Actions.
Frequently Used Feature
Chances are that if your users access a feature every time they open your app, they will want to access it through the chatbot as well. Implementing those frequent features is important because your users will expect them. If you’re a banking application, showing a user’s balance is a must-have. How would your users trust the chatbot if it can not display what they consider the bread and butter of your product? This means that your product’s most popular features are table-stakes for your chatbot’s utility.
It follows that these features should be the first competencies you develop for your bot. Not only will they cover the majority of the cases your users will need, but they will also act as a guiding light for what your app delivers to the customer. By developing these features in a conversational paradigm you will quickly notice whether your users visit your application to extract information or to take action.
Let’s take the banking industry for example. If you run a consumer bank, chances are that your customers will be checking their account balances every time they interact with your services. Whether to check on their new automated paycheck deposit or to perform a monthly audit of their credit score, most consumer bankers utilize online and mobile banking for staying up to date with their financial status. On the other hand, if you run a business bank, the majority of your customer services will be to deposit, withdraw, or transfer large sums of money, in addition to extracting billing data to generate financial reports. So while these two banks both provide similar institutional services, their chatbots will end up with drastically different Frequently Used Features.
Each chatbot will have a specific breakdown between Informational and Actionables skills depending on your users’ behavior
I’ve found that these features can commonly be split into two categories: Informational or Actionables. Informational features consist of extracting data and relaying it to the user. In our banking example, this was a simple balance query “What is my balance”. The second category of features, Actionables, revolves around performing an action on behalf of the user, like making a money transfer. After developing the frequently used features, the breakdown of Informational to Actionable services your chatbot must provide will become apparent. While the distinction between these two categories may seem arbitrary, future issues will cover how their technical implementation and conversational considerations can differ dramatically, and what that implies to your development process.
Most Difficult Actions
Once you have your users’ most popular features covered, it’s time to focus on their pain points. These paint points can manifest in two ways:
The Difficult Necessities: actions your users find cumbersome but perform anyway
The Forgotten Gems: features that are unutilized due to their difficulty or obscurity
The Difficult Necessities
The features in the first category tend to be essentials ones that are designed to be multi-step and involve many confirmations and form entry. The reason they are complex may be due to the technology used to implement them or regulatory restrictions at the time. In most cases, it is because that feature was developed at a time when interactive, single-page applications were not as popular as they are today. A classic example of this in the chatbot business is money transfer. On my personal bank account mobile app, I have to navigate four different pages before I can wire money from one account to another. And it’s always different based on the transfer type, which can be annoying. Chatbots (especially conversational ones) are good at solving these sorts of problems because they are not created on a “page-by-page” workflow, but can take your input all at once. For example, instead of specifying the source account, recipient account, money amount, and transfer date all in separate pages, you can simply say “transfer 20 dollars from my checking account to my mom’s account now”. A decent chatbot will pick all of the information you provided and ask you to confirm the transfer or clarify from which checking account (if you have multiple), etc.
Chatbots can help facilitate the difficult necessities of your product for your users
These multi-page, cumbersome flows are hindering your user’s ability to utilize the essential features of your service. Because they can not avoid doing them, your users are associating this difficult journey with your brand, which is not desirable. Chatbots can help alleviate these issues by offering alternative workflows that allow the users to perform actions at their own pace. The confident and quick-witted consumers will try to finish a transfer in one go, while the more timid customer will likely want to converse with the chatbot in a more structured and turn-based approach. Both ways, your customer is getting a responsive experience where they feel helped rather than hindered. In that way, chatbots can help facilitate the difficult necessities of your product for your users.
The Forgotten Gems
The second category is harder to define and might be highly subjective. The reason being that this category will encompass all the features that you believe make your product stand out, but your users either think of them as a gimmick or simply do not know they exist.
An example that we’ve mentioned multiple times now is money transfer. It could be that your money transfer policy is truly revolutionary or extremely inexpensive compared to the competition. Your users might not take advantage of this feature because it’s complex or simply not promoted in your product home page. Including these hidden gems in the chatbot increases its channel distributions, thereby giving it more space to breathe in and increasing its consumption. Why does it matter? you may ask. Won’t people just ignore it on the chatbot as they did on the main app? No. Chatbots are inherently contextual and can only do so many things at once. Therefore, their scope tends to be highly defined and limited. Instead of fighting for shelve space with your website’s other content, it is now being promoted on the metaphorical shop’s window display. In this way, you can elevate your products unique offerings through your chatbot, and highlight the unique selling point to your users.
You can use the chatbot as your product’s second introduction to your customers, and help your users find ways it can be useful for them. In some cases, they will discover new features they didn’t know you offered, and in others, they will be prompted to use something they have been thinking about for a while.
News Last Week
Hooray! You made it to the end of the newsletter! Your prize is last week’s most exciting news in the chatbot industry. There seems to be a general uptick in virtual assistant adoption in conferencing software, surely prompted by the surge in online communication due to Covid-19.
A GPT-3 backed AI wrote the top article on HackerNews.
AudioCodes AI can now transcribe, summarize, and act on your meetings.
Cisco Webex announces an integrated AI Assistant to help your meetings.
Verizon is partnering with Google Cloud to automate customer service centers.
US Bank releases a native Virtual Assistant on its mobile apps.