Designing for Chatbots

Oli Taylor
Oli Taylor 1st February 2019

As chatbots become prevalent and their AI matures, we’ve started exploring use cases for a number of our clients looking for ways to improve efficiency and service experiences.

Chatbots may not be appropriate for all organisations/services, so as we move beyond the initial “must-use-new-tech” phase, there are a number of design considerations, to identify whether a chatbot is likely to be genuinely useful to the specific service. These principles include:

1. Understand when you don’t need a bot; it shouldn’t be a replacement for forms or functions that are already performing. Like any new piece of technology, it can be tempting to adopt it for the sake of being seen to use it.

"Where can a chatbot be genuinely useful within your service?"

Holistically think about where you can create a better experience or make your own processes more efficient. If the chatbot doesn’t answer user needs, it will be detrimental to your service and create more work for your customer-facing teams.

2. Remove repetition; chatbots are good for repetitive tasks; if a service desk is answering hundreds of requests from users every day that can be defined into themes or a common set of problems, a chatbot may be ideal to save customers’ time and reduce operational efforts by the business. This could be queries relating to opening times, stock levels or for a specific piece of information on a website.

"Where users have unique needs and require complex processes chatbots aren’t so useful."

3. Don’t pretend to be human; it’s not live chat. There is no benefit of dressing up a chatbot to appear to be a human customer service agent. This is sure to frustrate users, particularly if they get to a point where the chatbot can no longer keep up with the conversation and answers become generic or even worse, irrelevant. Being transparent with customers from the outset manages that expectation and is less likely to create a negative experience with a brand, should they not be able to get the support they are looking for.

4. Use structured input; if you can surface options within the chat, you can help guide conversations leading to a better customer experience.

5. Keep everything native; keep users on the platform that they initiated the chat on. If they’re on your website, don’t jump them to Facebook Messenger. There are traditional conversion issues (such as URL hopping) but also expectations of users to manage.

6. Design for failure; you can’t plan for all scenarios, but you can plan for your customers to find those that you missed. Like any product or service experience, think about the onward journey so that customers don’t hit a dead end.

"If the chatbot doesn’t have the answer, how can your customer find the answer?

7. Don’t forget traditional conversion principles; think about both the chat and onward journeys. Firstly, the chat needs to be succinct, too many steps and you’ll lose the customer. Secondly, consider how you can support your conversion rate; if you’re designing to help customers find a product; consider how much of the product information you surface and the process for that customer to add the product to their basket.

8. Learn and improve the conversation; most chatbot providers provider analytical data to show what your customers are saying and where they drop off in the process. Use this to optimise the conversation and improve your service and conversion rates.


Ollie Bailey

Ready to start your next project?

Get in touch with Oliver to find out how we can help.

Get in Touch