AskSid.ai Blogs on NLP, Machine learning, Cognitive or insight mining

My Top 10 learnings from Designing Conversational Systems and Chatbots

 

”In the old days, we didn’t write the way we talk because there was no mechanism to reproduce the speed of conversation. But texting and instant messaging changed that—and a revolution has begun.”

 

The truth is that conversational systems and AI have shaken the marketing and technology worlds because they brought the focus back to the customer. Adoption of these systems may be difficult but, in most cases, it is necessary.

Why? … Because, through a conversational system, we get a unique opportunity to design a highly personalised user experience which is available in very few other user-facing systems.

Designing such a system involves automation, personalisation, and ideation of the experience and expertise of the domain to make the system powerful.

Below are my top 10 learnings which must be kept in mind while designing a conversational system:

1. Decide the Journey of the chatbot:

The journey of the bot is key to a successful conversation – It must be intelligent, intuitive, flexible and easy for the user. The message structure, sentences, questions, and compliments must be planned in the flow with a deep understanding of the user problem that the bot is supposed to solve. Plan both guided as well as free flows throughout the journey. Differentiate your conversation experience by designing it to fit your customer’s journey.

2. Remembering the User Context:

This is actually one of the most important aspects of conversational design. Try not to lose user’s context throughout the conversation and be able to refer to it at any point later on – the user will fall in love instantly if you can deliver on this one. The underlying platform architecture and deep knowledge of the industry domain are the two critical success factors that decides how easy or how difficult this can be during implementation.

3. Align to the User:

Bot should be smart enough to align itself to the user’s style. This is key to personalisation. For example, if user sends emojis, the bot should respond with emojis; If user is using exclamation a lot, the bot should also use exclamations. This is a bit difficult if one attempts it at the beginning itself but with time it can be achieved.

4. Give a Personality:

As Vittorio Banfi from Botsoceity puts it – “The personality of your chatbot is like the colour of your website. You can NOT pick a colour”. A common problem in giving personality to the chatbot is that a company always has more than one customer avatars. Does that mean you will have two completely different chatbots within one company? It would be a better idea to ask the users who they are or where they came from and thereby adjusting their experience accordingly. The truth is, like every customer and all human beings, chatbot too needs a distinct personality.

5. Learn and Adjust:

There is no perfect conversation in this world because the “perfection” varies with perception. There is always scope to learn from misunderstandings and adjust the conversational system. The key to a good conversation is how your bot learns and fixes it’s mistakes. Train your bot how to self learn and adjust by building it into various dialogs and providing a dependable knowledge base on which it can fall back on. Using ‘disambiguation’ to train the bot can help a lot in building a successful conversational system.

6. Design with Scale in mind:

Make sure the code base, the conversation flow, messages, and personalisation are built for the future. Think about how the same system can be used by another set of users and different languages in various markets.

7. Switch to human:

Whilst AI is still in its infancy, the chatbots will undoubtedly have problems. Thus, having a fallback to real people can help and delight you and your customers at the same time.

8. UI smooth as butter:

The face of your bot is the UI. Any experience you design for your system is heavily dependent on the user interface it is presented on. The UI for your bot must be clean, fast, interactive and easy to work with. Ease of use may be invisible, but its absence surely isn’t.

9. Fail fast:

In a conversational system, failing fast is very important. Failing in this system will not be the technical errors or bugs but any unexpected flows which leads to an abrupt end that must be adjusted instantly.

10. Design for multiple channels:

keep in mind that messaging platforms are proliferating at a rapid pace. You have Facebook messenger, Skype, web that supports bot technology today and surely Whatsapp, WeChat and many more will follow soon. Your conversation design should be such that when you switch to new channels, you don’t need to rewrite the code or redesign the conversation experience.

In summary, Want your users to fall in love with your bot? Make your bot fall in love with your users.
Curious to know more about how we do things at AskSid.ai? Reach out to me at curiousmuch@asksid.ai and lets start a conversation…

Author: Puneet Joshi (puneet.joshi@asksid.ai)

BOOK
A
DEMO