7 critical principles you should keep in mind when designing a conversational experience


By Dan Leshem, CEO, Co-Founder of Plantt

3 mins read

Conversational commerce is gaining popularity: We see more and more companies offer their products and services on messaging platforms like WhatsApp in all business areas and across all industries.

This is not surprising: From the consumer’s perspective, interacting with the business in a conversational way can be much more accessible than other online platforms. We already use our messaging apps all day, and we expect businesses to interact with us the same way. Not to mention that’s definitely the way to reach millenials.

However, even more than customers love conversations, they hate chatbots, and can spot those conversations from miles away. Those silly robots who just do nothing but ask you the same question, repeatedly, and say “Sorry, I didn’t get that!”, or if you’re lucky, you’ll get some generic answer, which is far from the experience they expected, and didn’t really help them get anything done.

To avoid that, brands should carefully design the conversational experience and consider user needs, their business and technology limitations, making sure the conversation is not just about answering the customer, it’s about helping the get things done.

Following I listed some things to “keep-in-mind” when designing a conversational experience for your brand:

  • Learn from what’s working

When designing the experience, always start with dialog samples (forget flowcharts). If you already manage customer communication on chat platforms - analyze previous conversations: look for users intents, what their needs, what are the integration points or context-shifting.
If you don’t have previous communication, try to mock some. My suggestion: open a chat group with your colleague and ask them to pretend to be a customer trying to do something.


  • Verify understanding

In real-world communication, even when texting, we often signal the other participant we understood its meaning, and we often verify we understand correctly.
This should be applied to conversational AI: Always let your users know that you understand what they are asking.
For example - if your client asks for “one large pineapple pizza”, you should confirm by answering “1 Large pizza, with pineapple toppings. Got it. Anything else?”
This way, you make your users feel understood, making them comfortable with automated conversation despite technological limitations.

In some-cases, where you have low confidence when trying to get the user's intent - it’s better to get explicit confirmation from the user.

  • Show empathy

This relates to the previous one. The least thing users want to hear when they complain about their delivery being late is “Awesome!”. Better if you’d say something like “I’m sorry to hear your delivery is late. Let me check that for you.” If possible, use sentiment analysis to adjust your tone to the situation.

  • Guide your users to making the right choices for them

If you want to use conversational AI to automate sales, your conversational assistant should guide your customers to find what they’re looking for. In general, conversational platforms are not a good medium for discovery: customers would find it difficult to wander around your store.

In real-life, if you’d walk into a store, one of the store’s assistants would approach you and guide you through the store. This should be similar in conversational AI.
Be sure to ask the user for their preferences, consider previous purchase, etc. It’s important to know what they are looking for.

At last, don’t just throw a catalog/menu on them. If you have too many choices, offer them one or two. If they are not satisfied with these options, offer them more options. Much like in real-life.


  • Get things done

Your customers are there because they want to get something done: buy something, check-up in their account, etc - not to get a general FAQ answered. A good conversational experience must be integrated with other business tools to get meaningful data points and execute business tasks.

  • Always have a fallback

Automation should never compromise on customer experience. Decide on a fallback strategy: If you’re unclear about the user intent, it’s better to confirm. If you don’t understand the users, don’t try to force them too much - better to handoff the conversation to one of your agents.

The frustration of old chatbots and their limitations is why we built plantt.io, which learns from customers interactions and create an conversational experience that is automated, while not robotic.

More from tyntec

Blog - tyntec’s Latest WhatsApp Developments
Customer Service
What’s New in 2020: tyntec’s Latest WhatsApp Developments
4 min read