Artificial Intelligence - Content is King

All the digital transformation initiatives I have been involved in at their centre has been encouragement for users to digitally self-serve through a possible range of media types, for example text, voice, imagery or video. Reviewing the results there is little consistency across similar industries, products and services suggesting no consensus on the best solution for similar use cases, users and desired outcomes.

• Is there too much emphasis on how the data is served?

• Would this be better directed towards the quality of the content available and appropriateness for the audience?

• If users understand the content the media should matter less?

Analysis time spent understanding the target user group and their receptiveness to different media types will define the solution but will be useless if the content is not engaging or fails to deliver the answer the customer needs.

Having worked closely with content teams developing answers for machine learning solutions there are a few best practice themes that have appeared:

• Start building your content by talking to the existing contact teams and ask them for the most common and repetitive questions they get asked. What are the boring jobs or requests they dislike?

• Understand a user’s primary reason for asking a question and make sure you create content that answers it, don’t drift off into additional unrequested content but provide users with the opportunity to see that if they wish.

• Focus on visual display and how large blocks of text can overwhelm, break up messages into digestible sections to improve appearance and readability. Customer insight also suggests they don’t like scrolling, make answers short and visible at first glance.

• Position yourself as the customer, recognise and avoid jargon and complex wording use focus group insight as it’s not always obvious.

• Customers rarely have just one-question make related items available where they might want to continue their research. Identify where a series of questions will get the customer to the right response rather than just serve up what looks like the best option. When creating these conversational exchanges map out flows first, make them simple as possible and validate them carefully.

• Understand that customers ask questions in different ways, make use of capabilities that understand the intent of a question and learn to provide the most relevant answer. Don’t create multiple responses to the same intent it will risk customer confusion and complaint.

• Champion the customer in the face of content approvers; don’t let the experience get diluted. Its better not to have content at all rather than content by committee and confusion.

• Don’t try and answer every possible question before going live, get a critical mass based on the complexity of your product and service. Remember 80% of the questions asked are answered by 20% of the answers stored. Consider what would be enough content to get a customer to self serve again; this could be 20-100 answers initially dependent on the complexity of the product.

• Once live you will soon find out what customers are really asking and develop more content to meet this true insight. Review customer responses frequently and look for patterns don’t react to individual feedback unless it is highlighting a howler!