Distrakt
+44 (0) 121 212 9737 / team@substrakt.com
© 2017 Substrakt Registered in England. Company no. 5916054

Thinking about chatbots

January 24, 2017

Chatbots dominated new tech last year but what are they and how can an organisation start to think about making use of them?

Anatomy of a Chatbot

Modern chatbots are enabling natural language, either by voice or text, in the home – allowing users to retrieve information or perform transactions often made accessible by apps and web APIs.

A typical conversation will attempt to convert a sentence into various entities and an intent,

“I want 2 tickets to see Rogue One tomorrow at 8pm”
Intent: Purchase a ticket
Entity: Rogue One (Film title) , 2 (quantity), Tomorrow at 20:00 (time)

A chatbot will aim to retrieve all the required information in a conversational manner before making the transaction.

“I want 2 tickets to see Rogue One tomorrow”
Intent: Purchase a ticket (requires a film title, quantity and a time)
Entity: Rogue One (Film title) , 2 (quantity), Tomorrow at ? (time)

As you can see the time object is incomplete.

-> “What time would like see it?”
“8pm”
Entity: Rogue One (Film title) , 2 (quantity), Tomorrow at 20:00 (time)

Future of Chatbots

Although we’re still in the early days of this technology with many teething problems, there is a large investment being made by most of the major technology companies. Google (Google Assistant), Apple (Siri), Amazon (Alexa), Facebook (Messenger) and Microsoft (Cortana) all have their own “assistant” and they are hoping to become the industry standard.

Not only are they creating their own assistants, they are also creating platforms in the hope that developers will support their assistant and become the de facto standard.

Although it’s unclear which platform (if any) will become the industry leader, the above tools mean that developers won’t need to create their own natural language processors to build their own chatbot, reducing the cost for organisations to add this new platform.

When touch screens started to feel more intuitive to touch they quickly become a necessary part in mobile phones and now there are a generation of children who are surprised when their TVs don’t respond to their touch. Natural language interactions has the potential to be just as intuitive to the next generation and become an expected medium.

With lower development costs, assistants in every personal device and a growing expectation; it will only be the quality of response and range of tasks available that will determine the rate of adoption.

Considerations when making a chatbot

Firstly it’s important to ask if chatbots are right for your organisation. Certainly at the moment your target audience may not have the devices available or are comfortable with this method of interaction yet but will that change? Are your audience early adopters?

Considering how you incorporate chatbots to your organisation’s digital offerings requires the same attention to detail as any other interaction your organisation offers including websites, social media, newsletters, leaflets and call centres.

Features and user experience lies at the heart of all of these and should do so with chatbots.

Features to offer

If you already have a digital presence some of the main functions that a chatbot needs to support should be easy to determine.

A few quick questions to ask

  • Main goals: What are the primary user flows/call to actions?
  • Search: If you have internal search, what do users search for?
  • Phone: When you receive calls, what are the main questions?

Quality interactions

In UX circles there is a mantra of “Don’t make me think”. In natural language this could be applied as “Don’t make me talk like a robot.”

Most tools to create chatbots will allow you to “train” the system based on example sentences. This training is the cornerstone of machine learning and will help the system learn how to map sentences into intents and entities. There are countless regional differences in sentence structure so the accuracy of the chatbot will depend on how well trained your system is.

Branding

Branding is also an essential part of any interaction and the same thought needs to go into a chatbot. These will be the same questions that get asked when deciding on content for your website.

  • Is it formal or informal?
  • Chatty or matter of fact?
  • Are sentences understandable to the average reading age of your audience?

Conclusion

Development platforms will quickly start to improve and with more devices being sold with chatbot assistants, it could soon become an expectation to offer services on these platforms.

Now is a perfect time to start creating a roadmap for what services you want to offer and how to create quality interactions for this specific platform.

Sam Knight