Automation is a very popular term. It allows us to be creative, to think outside the box, take care of our business and enjoy the benefits of entrepreneurship lifestyle. I was introduced to the benefits of artificial intelligence while trying to automate a very annoying part of my work.
Before becoming a chatbot developer, I was a running a small resume writing firm. Collecting clients’ data was one of the most tedious tasks. Online forms didn’t allow clients’ to truly express themselves, while Skype calls and live chat were always stealing a few hours from my work day.
Chatbot was the perfect solution. A program that will talk to clients, collect their data, while providing a fully personalized experience.
Now, after three years of chatbot development, I still look for perfect solutions, but my views have become much broader. I would like to share my experiences, with entrepreneurs, who, like me, are always missing one or a few hours to make their workday complete.
How can chatbots help us?
Having a few extra hours each day is a great gift. After automating some of your tasks, you will wonder why you haven’t found out about chatbots earlier.
These useful code scripts have been around for more than a decade. We could find chatbots on mIRC, Windows Messenger, and ICQ, back in the day. They were spamming chat rooms, asking us for “ASL” and posting suspicious links and XXX content.
Today, chatbots got more serious. They represent the leading automation force in customer service, sales, HR, digital marketing, and various other fields.
For example, the U.S. rail company Amtrak managed to increase their revenue per booking, for more than 30%, by replacing traditional customer service forms with the chatbot called Julie.
If the Amtrak case study is not enough for you, I have also created a list of chatbot features. Ideally, every chatbot developer should carefully analyze client’s needs and come up with a list of unique functionalities. Still, if you are indecisive, this can serve as a good starting point:
- Customer service – Welcoming users to the website, navigating them through company’s offer, answering simple inquiries, requests, and complaints.
- Support – Solving simple issues, guiding users through troubleshooting and connecting them with live operators in case of emergency.
- Sales – Presenting company’s products and services, basic price negotiation, running loyalty programs, issuing and accepting discount codes.
- Marketing – Generating leads, collecting consumer data for buyer personas, broadcasting messages.
- HR – Interviewing candidates, collecting information and selecting candidates by their previous work experience, educational degree or some other criteria.
- Employee training – Presenting lessons in the form of online chat, testing employee’s knowledge, helping them to acquire new skills.
- Productivity – Sending notifications, conducting basic office assistant tasks.
Can I build a chatbot by myself?
On top of this list, I should also add that chatbots aren’t that difficult to make. There are dozens of free automated platforms, with plugins, wide variety of integrations, natural language processing (NLP) and easy to use interface. That is why today, many entrepreneurs decide to build bots by themselves.
In this tutorial, I will present some of the best benefits of this technology, by creating a simple chatbot for the Killer Startups website. The chatbot will be called T – 1000 (I am a big Terminator 2 fan), and it will welcome website visitors, generate leads, collect their data and sell ad space. So, let’s start!
Let’s create a chatbot!
Choosing the platform
I will create T – 1000, with the help of Snatchbot, a completely free automated platform. The chatbot building market is very diverse. I use various environments for my projects, depending on the needed integrations and features.
In this case, I have chosen Snatchbot because it offers a free website deployment and uses “if” clauses for building the dialogue flow. This way, the platform gives us a great insight into its backend work. Other popular automated platforms are Chatfuel, Botsify, Motion AI, Octane AI, etc.
This is how Snatchbot’s building environment looks like:
How to write chatbot dialogues?
Every platform allows you to create one or more conversational paths. These are all unique funnels that start with the welcoming message and lead to successful purchase (eCommerce) or to the interaction that provides the desired information (customer service).
By making the chatbot more involved, we are increasing the average time users spend on the page, which is a very important for its search engine rating.
Conversion is another important metric. When they are written by a good copywriter, chatbot dialogues can be as efficient as skillful sellers from the Middle Eastern bazaars. You know, the ones who always pull out the right product, a moment before you exit their shop. That is why writing skills and UX design knowledge are essential for bot building.
Apart from sales, the purpose of interactions is to keep the users inside the loop. The developer should always have both company’s and consumers’ goals in mind. That means that a thorough audience research should always precede the dialog writing.
The tone of the conversation depends on the bot’s purpose, but the general rule is to keep it friendly, exciting and informative, same as in regular customer service. This way, your chatbot will be able to provide pleasant and personalized experience to users, who will keep coming back.
Check out the two examples of welcome messages, and try to determine, the one that sounds more effective.
… or this one?
The first message that shows up after the user clicks the “Start Chat” button is often called a welcome message. Most bots use it to introduce themselves and ask for users’ name. We will use this interaction to collect users’ names and export them to our CRM list, but more about that later.
One of the biggest developer mistakes is to fully rely on users’ intentions and choices. Most users don’t know what to expect when clicking the chat button. Even if they want to ask a particular question, they need at least some level of guidance.
That is why chatbot needs to present its abilities at the very beginning of the conversation.
The sentence that explains that chatbot can answer inquiries about “the company, startup reviews and advertising,” narrows down the conversation and keeps users away from dead ends.
How do connections work?
Users’ response to the welcome message triggers the bots’ actions. In this case, most users’ will type their name. Still, some people might want to skip this formality and ask their question straight away. For them, we will need to define several directions in which the conversation can go.
We do that by adding keywords. One of the reasons why I have chosen the Snatchbot platform is the abundance of criteria you can select for defining the connections. Most other platforms only allow you to check whether users’ response contains the specified keyword.
On Snatchbot you can create a special connection if the users’ response starts or ends with a keyword, contains some part of it, or doesn’t contain it at all.
Each connection should contain all known variations a people can use for defining their intentions. This is how Snatchbot’s “if statement” connection looks like:
Simplify users’ choice with quick answers
Users often state their intentions in a very odd way. So, even if we add all synonyms and variations to a connection, the conversation might head in the wrong direction. To make sure users will take some of the desired conversational paths, we can use the quick answers feature.
These are the answers the bot proposes to the user, which are added as regular connections. When using quick answers, we can turn off the chat and narrow users’ choices to the displayed options. We can also leave it on and use the buttons as suggestions.
Adding dozens of quick answer buttons might confuse the user. Chatbots should imitate human interactions, and that is why I always create two layer connections. In the first layer the chatbot will ask a question, if it doesn’t recognize the user’s intention, another fallback interaction will pop up, with a quick answer choice.
What is a fallback?
Misspelled words, wrong contexts, and various other misunderstandings are common occurrences in every chatbot development process. Fortunately, we can decrease the volatility of these errors by adding the fallback connection to every interaction.
The fallback is the message that pops up when chatbot doesn’t understand the user’s response. It can be a standard error notification, but the conversation will sound much more natural if you create a custom fallback for each interaction.
Avoid admitting your mistake. Treating it as a confirmation of users’ choice sounds more natural, and it is an elegant way to hide your chatbot’s shortcomings. Here is one example where T-1000 is saving its reputation, by pretending to understand the user:
Fallbacks should be added to every interaction, and you can easily add them as any other connection:
How can we extract users’ data?
Data extraction is the critical feature for all the marketing professionals out there. Chatbots will soon send the newsletter forms into obscurity.
The attempt of RapidMiner, to replace all of their lead generation forms with Drift chatbot system, lead to creating more than 4,000 interactions and generating more than $1 million through their sales pipeline. This case study was also shared in the Brad Power’s article on Harvard Business Review.
Extracting emails has never been easier, especially with the use of third-party apps, like Zapier. Before we build this system, we need to determine the essential data our chatbots should acquire.
In the case of Killer Startups that would probably be the users’ name, website, and email address. T-1000 can ask users for this data in the first three interactions. By adding the webhook there, we can send the user’s response to the outside location.
Zapier is one of the most popular apps for creating third-party connections. It will also allow us to connect the webhook to some other app, by filling in the short and easy-to-understand form.
Integrations like this can bring tremendous benefits. By connecting T-1000’s interactions with a Google Sheets file, we can place users’ names, websites, and emails in a single CSV document. Later, we can upload it to Adroll, or some other CRM platform.
Zapier will only paste the contents of the POST requests it receives, so you need to do a little formatting if you (like me) want your CRM list to be in perfect order.
Email notification is another impressive feature. If you want to be alerted when users’ leave their data, you can also add it to the desired interaction, and the bot will send users’ response to your inbox in the form of an email message.
The email looks like this:
Shift to live chat
Although chatbots and A.I are quickly conquering the market, there are still consumers who always insist on live chat. The reason for this might be the dystopian influence the Terminator 2 movie had on millennials.
Anyway, that is why we will need to add the live chat option to our T-1000 bot. Luckily, Snatchbot and most other platforms allow humans to intervene and join the conversation.
This feature can be lifesaving in case chatbot is not able to recognize user’s intention. On most platforms, we can add the live chat interaction to the permanent menu, which is a list of commands that the user can call up throughout the whole conversation.
If we add the email notification to this interaction, we will receive an email, whenever someone asks for (human) help. Then we can log into the platform and answer the users’ inquiry.
Chatbots can present items in the same way as regular eCommerce product pages, which is much more efficient than location-based store’s windows and shelves. Interactions can contain videos, photos, product descriptions and catchy slogans.
We can also add the link to the product page, but that isn’t necessary since many platforms offer independent payment getaways, which means that users can choose, review and purchase their favorite items within the chat window.
Snatchbot offers a PayPal getaway that can be used by anybody who owns a PayPal business account. We can also add various other payment getaways with the use of plugins and third-party APIs.
Apart from the functionalities we explained in the previous paragraphs, chatbots can also:
Extract URLs, emails, phone numbers, dates, and addresses;
Translate users’ answers;
Add context to the conversation or its segments with a use of natural language procession;
Broadcast company memos and other valuable information;
It is tough to explain all the chatbot functionalities in one article. That is why I recommend you to do a little research on this topic before you start creating your solution. Artificial intelligence is a very trendy topic, and you will find plenty of great articles and tutorials, on Medium and publishing platforms like
The T – 1000 chatbot, is now ready to take over the Killer Startup’s customer service. Click on the link, talk to it and try to recognize all the functionalities we discussed here. We can add it to the website, Facebook Messenger, Skype, email, or Twilio, while some platforms also allow the deployment on Telegram, WeChat, and many other communication apps.
I hope that T – 1000 has shown you that chatbots and artificial intelligence, won’t take over the world and exploit the human race, but help us to improve our communication with consumers and automate tedious and repetitive tasks, so we can have.
About the Author – My name is Branislav Srdanovic. I am a copywriter, who enjoys building chatbots in automated platforms and code-based frameworks. My bots are live, chatty and incredibly human-like. In my spare time, I write articles about the benefits of A.I. and explore new industries my chatbots can target.