Create your own chatbot? Here’s how!

The die is cast — you want to build a chatbot and implement it into your organisation. The good news is that new chatbot platforms are making it easier and easier to develop a chatbot all by yourself. However, there are things that you’ll need to consider carefully before you start. What questions should you ask yourself before you start, and what’s the best way to start a new chatbot project? It’s all covered in this blog.

What issues do you want to solve with a chatbot?

Let’s start at the beginning. Before you start developing your chatbot, it’s important to have a clear goal in mind. This goal will help you through the entire process and you’ll need to test the success of your chatbot from time to time once it’s up and running.

What organisational issues are you planning to solve? For example, do you have high volumes on your messaging channels or live chat? Then of course you’d like to be able to answer all the questions that come in as quickly as possible, but the service employees are short-staffed. You’re looking for a solution to offer not only the best, but also the most efficient service. A chatbot can help answer frequently asked or simple questions, making it an overall great solution.

If you notice that your organisation is mainly being contacted by telephone or e-mail, then chatbots would seem the less obvious choice. However, you can handle part of the workload if you use channel management to send people to channels where chatbots can be integrated.

But there may also be another motivation for exploring the use of chatbots. For example, do you want to offer faster service and achieve better response times? Do you want to expand accessibility even outside of opening hours? Do you want to create efficiency for your service employees by letting chatbots do more of the preparatory work? Possibly all good reasons to start. But make sure you compare these ideas with your company’s data. Because only in this way can you truly predict whether it pays to use a chatbot. Read more about it in the blog ‘How data-driven chatbots will ensure success’ or request a feasibility report. With a feasibility report you gain insights into what a chatbot can bring to your organisation.

What kinds of questions are suitable for a chatbot?

To clarify whether a chatbot really offers the right solution, it’s best to start by analysing and categorising the types of questions that you receive per channel. Once you’ve mapped those questions out, you can then estimate whether or not a chatbot can provide added value. A chatbot can answer a lot of questions, but not every question is suitable.
Questions about, for example, opening times, return policies or where certain information can be found on the website can be handled very well by a chatbot. More complex or personal questions about, for example, a specific complaint or the application for a building permit are more difficult, or even impossible, to answer by a chatbot.

How do you start with building a chatbot?

Once you’ve analysed past customer questions and made a selection of questions the chatbot can answer, it’s time to start writing the chatbot dialogues. On the OBI Bots Platform, for example, you can put chatbot conversations directly into the dialogue designer of the interface. Here you add both the questions that the chatbot should recognise and the answers that the chatbot should provide.

You need to look at the conversations as real conversations between two people. Write the text exactly as it can be said and also give the text a personal touch. For example, with funny elements, provided this suits your organisation’s culture of course. Just make sure the conversation doesn’t get too human. It needs to be clear to the user that he or she is conversing with a chatbot — that keeps expectations clear.

You can give different options for each intent — the goal a customer has when he contacts the organisation. A customer can make something clear in different ways. If the intent is “apply for a new passport”, then you can also add these options, for example: “I want to apply for a new passport”, “I want a new passport” or “I need to renew my passport”. For the OBI Bots platform, the entire conversation can be made on one page.

How do you train and test your chatbot?

Once you’ve worked out the conversations, it‘s important to train your chatbot. This way you can be sure that the chatbot understands the intent, or intention, of the user, even if it is formulated differently than you have indicated as an example. Within the OBI Bots platform, this training is automatic when you press the “Train Bot” button. The bot then creates a new language model that helps recognise the questions, even if they differ from the examples you provided.

After this, it’s important to test your chatbot properly. Don’t just ask technical people to be your test panel. Also make sure that you ask people who have less technical knowledge and who are at the level of your target group in terms of knowledge. During the test phase is where you learn what is going well and where there is still room for improvement.

Is all the work done once you’ve gone live?

You have demonstrated the added value of a chatbot with data, built and tested the chatbot. After the release, the work is done and the chatbot can get started independently. Or not? Once you’ve gone live, there is still work to be done. Improving the bot is a continuous process.

Monitor the chatbot conversations and see how people react to the chatbot. If you see that people get stuck on the same conversations, you may want to adjust some words and see if conversations run smoother after that.

The OBI Bots Platform offers a “Test Center” for this in which you can view and evaluate the conversations of the chatbot. You can also label these conversations by intent so that the chatbot can be able to better recognise the intent of future conversations.

Can I really build a chatbot without any help?

Innovative chatbot platforms allow you to build a chatbot without having to code. So you don’t need a technical background and you don’t need an IT department to do it. Nevertheless, there are a number of reasons why it is good to set up a project team before you start with a chatbot.

One reason is that you’ll need to make sure you have sufficient knowledge of the target audience and the type of questions that target audience will ask. Are you building a chatbot for customer service purposes? In that case, involve a service employee who knows everything about this target group and who knows which questions are asked the most and what the answers to these questions are.

Start now (right at the beginning) with your very first chatbot!

Time for action! Would you like to know where to start this adventure? Reach out, I’m happy to help.