The chatbot canvas – for chatbot concepts
If you want to implement a chatbot, you need a good concept. The chatbot canvas helps you to develop a complete and structured concept for your next bot. If you use this canvas as the basis for your chatbot concept, you will prevent important steps from being forgotten.
With the canvas, you gradually go through all the important phases on the way to a successful chatbot concept.
Current challenges and chatbot use case Use case
Start with the challenges you, your employees and your customers face and then work out the use case for your chatbot.
The use case must generate measurable added value right from the start. This can be increased customer satisfaction, cost reductions, sales increases or similar.
If your use case does not create added value, it will be difficult to justify the time and cost of this project.
Furthermore, the use case should not be too broad so that you do not run the risk of getting bogged down. It is better to start with a manageable, limited use case, which nevertheless immediately creates added value.
For example, if you want to start with a chatbot to answer frequent questions, you could narrow down the type of questions, for example. Perhaps only questions about a product group that you offer.
Ideally, you can realize the first use case within 2-3 months and achieve initial results and learnings. Gaining experience is particularly important in the first use case. You need to learn how your target group reacts to the new medium and what you should do differently/better in further use cases.
Goals and target groups
Then define the goals and target groups for your chatbot.
The following goals are often set:
- Increase in customer satisfaction
- Increase in turnover
- Increase in leads
- Cost reductions
- Increase in employee satisfaction
- Strengthening the brand/image
Goals are very important if you want to justify the chatbot project internally.
However, goals are also essential in order to measure the success of your chatbot after implementation.
In addition to the objectives mentioned above, you already define specific KPIs and ways to measure these KPIs here. Some goals are very easy to measure. Goals such as satisfaction can usually only be measured through further surveys.
You can do an exercise on personas to define your target group.
The persona definition exercise helps you to get a concrete picture of the chatbot users. You can only design a successful chatbot if you know your chatbot target group exactly.
Important: Do not do the exercise for the entire target group of your company, but focus on the part of your target group for which you are also developing the chatbot. This may or may not be the same.
You should define the following persona characteristics:
- Age
- Family status
- Technologies used
- Information sources and times
- Profession
- Position
- Income
- Hobbies
- Vacation destinations
- Property (car, house)
You can also give the persona a picture and a name.
The needs of the individual personas can then be derived from this. The needs in turn result in solutions as to how a chatbot could solve this person’s needs or problems.
Chatbot persona and personality
Do you know that? You go into a store and the person who greets you in the store does not meet your expectations and seems unsympathetic. In this case, would you like to leave the store immediately?
It’s similar with a chatbot. If you’re chatting with a bot that doesn’t suit you, you’ll want to end the conversation immediately. However, this should not happen during communication between your chatbot and your users.
Use the persona definitions of your target groups made in advance to define your chatbot personality.
The more specifically you know your target group, the easier it is to define the chatbot personality.
In addition, the image of your company and how you want to present yourself to the target group also play an important role in the choice of chatbot personality. If your company has a rather reputable image, then the chatbot must also appear that way. If your company reacts in a fun and relaxed manner, then the bot can act in the same way.
Features
Once the goals and target groups are known, the competencies of the chatbot can be derived from this.
In technical jargon, this is called the features that the bot should contain.
Possible features are, for example:
- answering questions on a defined topic
- the creation of new contacts in CRM
- Capturing new leads in a third-party system
- arranging appointments
- ordering products
- the reservation of services.
It is also defined here whether the chatbot should understand free text or whether the user can only respond by using buttons. With the first, the user can simply enter their question or answer in the input field. With the second, the user simply chooses between buttons to answer. An NLP (Natural Language Processing) component is required for the first case, but not for the second. You should know that using NLP requires far more effort than if the chatbot only works with buttons. Furthermore, some users do not feel comfortable if the chatbot does not provide any suggestions for answers, i.e. buttons. You often don’t know how or in what way you should answer.
A distinction is also made here between rule-based chatbots and AI-based chatbots. If your use case is a rather simple case where the questions and answers can be easily predefined, then I recommend implementing the chatbot using buttons, i.e. using a rule-based chatbot. However, if your use case is more complex and the conversations are not always predictable or the questions and answers are even context-dependent, then you should consider free text and NLP, i.e. AI-based chatbots.
A simple use case refers to a marketing case, for example. A more complex use case is when it comes to customer service or customer inquiries.
Conversion tool
You can then decide on a suitable implementation tool based on the features.
Depending on the use case, you may need to use a tool with NLP or one without. You must also check where your chatbot must be hosted, depending on the data protection guidelines. There are even tools, such as aiaibot, that host all data in Switzerland.
In the case of very complex use cases, in-house development is usually recommended, then you are completely free in the implementation.
Testing
…
I am a guest author at morethandigital.info. You can find the entire article here.
