Close
Written by
Matthew Weaver

Matthew Weaver

10 tips for delivering an engaging chatbot experience

Preparing your business for digital transformation.

People are adopting voice and text communication more than ever before. Predictions state that, by 2020, 50% of all searches will be conducted by voice. By this time, Gartner believes that the average person will have more conversations with bots than with their spouse. Whilst SMS messaging is slowly declining, the use of messaging applications is on the rise. With 50% of adults between 18 and 24 feeling text conversations are just as meaningful as a phone call.

Tech-savvy businesses will benefit from these trends by automating many of their customer interactions. Chatbot applications provide a platform for providing this interaction. They can reduce costs, easily scale to meet demand and provide round the clock availability.

People are increasingly happy to engage with chatbot services. Equally, frustrations can occur quickly if these services are poorly constructed. This article provides 10 tips for creating an engaging chatbot experience.

These suggestions may not all apply to your own chatbot challenges. However, thinking about what’s possible will definitely help to promote innovation and creativity. In turn, this will generate interest, and contributions, from a wider audience.

two chatbots

Act like a human rather than an application

A chatbot should mimic a customer agent rather than an IT system. After all, its aim is to represent a person – not an application. The user experience should be more like a search engine than a prescribed application workflow. Interaction is informal and conversational – rather than relying on specific commands and instructions. It should be clear which functions the service is able to perform – and those it cannot. In this way, if the service cannot address a request, it’s possible to delegate to a human advisor.

Think about the real objective

A chatbot is essentially a communication device. It provides a way for users to interact with business systems and services. The ability to integrate with other systems is essential. Without this integration, your chatbot’s capabilities will be severely restricted. Your chatbot does not have to support every conceivable activity. For the activities that it does support, a single entry point makes it easier for users to perform key tasks. It can also reduce the time it takes to become familiar with back office systems. Your bot should politely inform users when it doesn’t understand something. When this happens, make sure your users can easily learn what is, and isn’t, possible.

Create an event-driven experience

The user should lead the conversation, working towards a specific objective. The chatbot determines the user’s intent and the information it needs to fulfil the request. The bot can then identify critical data and, where it is missing, prompt the user to provide it.

Context is one of the toughest challenges to solve. Within a given thread of conversation, your chatbot should remember and learn from previous user interactions. The ability to understand context makes a conversation more natural and far less prescriptive.

Give your bot a personality

We can probably all recall a lifeless conversation with a customer agent or representative. Or indeed, the monotony of an automated call that strictly follows a predefined course. ‘Please press 1 for yes, 2 for no, or 3 to repeat the question’. I get frustrated just thinking about these experiences.

Your chatbot is aspiring to replace elements of a person to person conversation. Giving your bot a personality will help to keep users engaged. Just be careful not to go too far, avoid sarcasm and smart one-liners that get very dull, very quickly.

Engage in idle gossip now and again

A typical conversation will often start and/or end with a little small talk. ‘How are you today?’, ‘The weather has been awful, hasn’t it?’, ‘Hope the rest of your day goes well’. These pleasantries can, of course, be avoided when communicating with a chatbot. We have a long way to go before bots can think for themselves.

Still, an understanding of common statements will help to humanise your chatbot. Sometimes people measure a chatbot’s ability by how well it understands ‘general chit-chat’. Testing how ‘human’ the chatbot is. If nothing else, it amuses users to test your bot’s general knowledge from time to time.

Don’t treat it simply as an IT challenge

Building an engaging chatbot solution requires contributions from many people. Don’t treat it purely as an IT function. Domain experts and stakeholders will be able to describe what they need and why. UI and UX specialists will help to create a compelling and consistent chatbot experience. End users will explain what works for them and, equally as important, what doesn’t. For B2C solutions, marketing experts will ensure your message and your services are well understood and higly visible.

As with any other digital service, seek advice from anyone that will contribute and/or consume the solution you are creating.

Make it personal

People love to be recognised – your chatbot users are no different. Consider ways to record what users have previously asked for. Storing key topics and requests can create a timeline of activity. This builds a more personalised experience, people feel their concerns are being remembered as well as being addressed.

Cognitive services can extract key topics from natural text. Knowing your users’ intents and interests helps you to understand their challenges better. Sentiment analysis can infer if someone is using positive or negative language. This allows you to adjust your bot responses accordingly. Perhaps passing your user to a human representative if necessary.

Share regular updates

Delivering a chatbot with limited functionality is fine. In this way, you can collect early feedback and adjust accordingly. Strive to add new features quickly and often. A simple newsletter can inform people that your bot has learned something new. Users are generally eager to try out new capabilities. With a disciplined release process, small changes can make a big impression.

Auditing interactions will inform you what users are asking for and how well your chatbot is helping them. You can use this information to influence your chatbot wish list. Address key user requests before they get fed up of asking!

Keep an eye on the future

Language processing and cognitive services are evolving at a rapid rate. Competition is fierce in the marketplace amongst technology providers. Keep an eye on new developments that may improve your chatbot capabilities. Additionally, the channels that support chatbot integration are increasing all the time. If broad exposure is important for you, then this should be of interest.

Be aware of vendor lock-in when choosing a chatbot framework. This may be unavoidable to some extent. Still, a well-designed architecture may help to reduce your dependencies on any particular service. This may be especially important when the framework you didn’t choose releases a killer feature.

Measure and report success

Measuring how well your chatbot is working is essential. It will show what your users are asking for, existing features that aren’t popular, and how well your bot deals with user requests. Additionally, you will learn common user vocabulary that your chatbot does not understand.

You can extend your bot to address common shortfalls. It may help to support user suggestions from the bot itself. In this way, users can register ideas and recommendations directly with the chatbot. You can then choose to share common feature requests with your entire user base.

Share this post on

Tags

One thought in Comments

Leave a Reply

Required fields are marked *

Read next

Don’t ask the robot to volunteer

Mundane tasks are what they are supposed to free us from. Working faster and cheaper, giving more space to human creativity and unleashing the potential that we, people, have. To some folks, it sounds creepy though. Artificial intelligence fulfilling a colleague’s duties? And what if bots replace us also in the typically interpersonal matters? Will you ever be hugged […]

Read more