The internet is full of predictions about chatbot trends, but here are some things I don’t think are ready for prime-time yet.

1. Website and Apps Replaced by Chatbots

Look at any chatbot trends list and you’ll see predictions of companies replacing their website with a chatbot. While I don’t doubt we will see one or two experiments with this during the year, it’s unlikely to be widespread. There are things that websites are much better than chatbots for — like browsing lots of information. As a result, there isn’t a reason to get rid of it just because you added a chatbot. We also need to consider that most chatbots are currently either part of an app or website, or use one for user settings (consider Alexa or Google Home).

For now, we will see companies adding a chatbot to their websites rather than replacing them with one. You can design experiences like this over at Botsociety.

2. Big companies investing in specialized Chatbot Design roles

Lots of startup companies have been focussing solely on building bots. These companies are truly embracing the necessity of conversational design in user experience.

Large companies are simply adding chatbots to their existing (and often long-standing) websites and applications. This means that they are much slower to see that chatbot design needs a new way of thinking about design. As 2018 progresses, we will see companies with good chatbot UX starting to take business from those who lack it. Building bots are important, but there’s more to the equation.

By the start of 2019, I would expect large companies to have started to realize that investment in specialisms and tooling for proper conversational user experience is necessary. This will start to create whole new classes of jobs.

3. True Sentiment Analysis

Designing your conversation prior to building it will make your chatbot fly. Check out Botsociety.

In the reporting about chatbot trends, there is a lot of hype about computers being able to determine your emotion. They do this using your language, the tone of your voice — even facial expressions. There are many services that offer that ability. However, they are variable at best, and chatbot conversation tends to be short sentences with little to no emotion. After all, you can’t tell a lot from a user saying “order me a pizza”. By the time a user’s responses are clear enough to detect frustration it’s time to pass your user over to a real human!

What we will see later in the year is the chatbot getting better at guessing your emotion using a variety of ‘hacks’. One of the key ones will be using the context of the chatbot itself. If your chatbot handles making insurance claims you can assume that the person is likely upset. If you have built a chatbot that tells jokes, you can assume that the user will not need the same caring tone as the insurance claims chatbot.

Using context to adjust the conversation to the correct mood will become common practice this year. Detecting and using the users’ actual emotions in real-time will take a bit longer.

4. Chatbots being able to write their own conversation

When chatbots make up their own conversation it’s called Natural Language Generation. It works by allowing the computer to take in a large amount of relevant data and then using it to construct similar things. Although some advances have been made in this area, the outputs are still a little odd.

An example of this is Microsoft’s social chatbot Zo. It’s designed to learn from users, and use what it has learned to form new conversation responses. You’ll see from the screenshot below that the conversation can go from cute to surreal very quickly.

Most chatbot creators are not going to be comfortable setting their chatbot free to say whatever it wants. At least for 2018, almost all chatbot conversation will still be carefully crafted by humans.

5. Chatbots having their own Social Norms

Conversation is a very social activity. We have all sorts of unwritten rules about how we should talk and what we should say in different situations.

Should we say please and thank you to a bot? Do we expect them to understand us, or do we adjust our conversation to simpler, shorter phrases like we might for a toddler? Is it ok to use in a shared space? Most people would happily sit in a coffee shop and have a conversation with a colleague about getting something organized. But, few people would feel comfortable speaking to a chatbot in the same space, even if the conversation was the same and the responses not any louder than another human speaking.

Design chatbots of your own over at

Chatbots promise a way of creating a truly “human-like” interaction with technology. As a result, we are still interacting with them in a human-like way. This is still a very new way of interacting with technology, and some of the capabilities Apple, Google, and Amazon add into their virtual assistants will be key in shaping this behavior. Consider that Alexa’s new follow up mode will start to encourage saying ‘thank you’. This year we will see conversational interfaces appearing everywhere, next year we will start to learn what that means for social norms.

There are lots of exciting advancements being made in Conversational Artificial Intelligence right now. This year will be very interesting.

What do you think? Are there any more chatbot trends you think we will or won’t see this year? I’d love to hear from you in the comments below!

Also published on Medium.


Gillian Armstrong

Conversational AI Specialist