Conversational Trends 2019 - Part 3

This change is important, because probably for the first time in HCI history, we are moving away from humans having to learn how to interact with computers towards computers learning how to interact with us.

Chatbots provide quite a different experience compared with web pages or mobile apps. They are (or feel) personalized, they are easily available across multiple channels and they allow us to focus on the task at hand by eliminating unnecessary UI elements. The journey feels more guided, which can also increase the comfort of the interaction.

Since bots are all about conversation, there are many rules and guidelines that we as humans have developed with time that make conversations a pleasant experience. Below are some of the ground rules we have developed at Hycom while working on conversational interfaces for our customers – I hope it will be helpful for you as well.

Narrow domain > wide domain

The less you try to offer to users first, the more you can focus on the script, personality and usability. Not only that – there will also be many less ways for users to go off the track. Make sure you state the purpose of the bot to set the expectations right in the beginning. Users will try to derail your bot with personal and political questions – the more it’s focused, the better the experience.

Start with a great script

Use all the design tools you would normally do when designing great user experience, but instead of mockups – prepare great scripts. Instead of icons and graphics – work on the tone of your copy. Create personas and study their typical use cases. Shape the flow of the conversation around them. Be direct and polite! Think about the inputs that will be required and try to minimize the amount of work for the user.

A great way to test a scenario is by interacting with another human being via SMS text messages – you will instantly spot the stages where they feel lost.


Chatbot interfaces tend to be quite modest compared with apps. Your main way of conveying great user experience will be copy of the script. Take some time and try to make it sound like human, while also conveying the brand values. If you wonder about the language you should use – take a look at your website, print brochures or ads - they probably already use the correct tone for your brand.

Data Entry

Users expect conversations to flow naturally. This means you have to put in an extra effort to think how to ask them for input. Try to eliminate multiple responses (think +10) to a sensible couple of defaults. If not possible – hide the less-commonly used ones under „Other…” button for example. For simple interactions, it’s easier for users to select “Yes” or “No” buttons instead having to type in the text itself. It also saves you from causing ambiguity in the input and makes it easier for everyone.

Try to use the values you already have. If you need location info – ask for the correct permission for example. Some messaging platforms will also provide certain details, like first or last name and email.

Conversational Search

Voice search can work great for e-commerce websites with thousands of SKU’s. Browsing through them using conversational interface is not just an option. Instead, provide tools to easily narrow down the category and main features of the product. Expect to receive feedback like “Show me a pair of comfortable home slippers” or “I’m looking for a colorful summer dress”.

CI’s lend themselves greatly to curated searches. Use attributes like “promoted” or “most popular” products to limit the amount of search results. Remember – it’s not about users being able to browse your whole inventory – it’s about providing them with a limited list of relevant results fast.

Personalization and History

When we talk to our friends, we expect them to remember certain important facts about us, like the fact that we have a wife and kids or go on holiday next month. We do the same with bots. Bot personality will be greatly reinforced if it remembers previous conversations.

We should remember what users mention their preferred size, delivery address or payment preferences. This way, common tasks like future purchases or returns become much simpler. We also increase the cumulative value of the bot – the relationship becomes more valuable for users, because it saves them time.

This also creates great opportunities to trigger purchase notifications – think of your kid’s birthday or anniversary.

Monitor fallback answers

Oftentimes your users will surprise you with their requests. Bot creators create a generic “fallback answers” for this occasion. It can be a randomized list of answers similar to “I’m sorry – can you rephrase your question?” Using tools like Dashbot or Chatbase, you can analyze the conversation log files to look for recurring users’ phrases that end with bots being clueless as what to do next. Try to address them using the fact that you can push the changes instantly.

Bots are very easy to build; Helpful bots are very hard to build

Nowadays we have plenty of platforms that allow us to create and publish bots in matter of hours.  They even provide ready-made scripts e.g. for generating leads on your webpage. Do not fall for such “instant” bots, as they will probably be missing most of the features we have discussed so far. You usually have one chance to build a meaningful relationship with your users. Take your time to provide a great user experience that’s on par with your brand values and will keep users coming back to you.

We’d love to help you!

Starting your conversation project should follow the agile method – build it, measure it, learn from it, rinse – repeat. Right now is a great time to start with Your chatbot. At Hycom we can help you make it happen at each of the described stages. From discussing business goals, through designing the customer experience up to implementing the bot along with custom integrations. We also work in iterative process so that we can learn and improve the solution with time.

  • Łukasz



    Digital Innovation Manager