Skip to main content

For customer experience, AI must be substance not form

In a recent blog post, Gartner analyst Michael Maoz observed that the mere presence of AI in a CRM system or on a website doesn’t guarantee customers will find help easily. Not if the AI is less intelligent than its designers imagine it is. Or, even worse, if the AI is “artificial, superficial, condescending, or insulting”.

When AI fails to respond appropriately to the customer’s need for information, issuing unhelpful replies or sending her on a circular tour from website, to call centre, and back to the website, its presence becomes more of a hindrance than a help. It’s too often assumed that AI per se will benefit customers, without focusing on the effort required to complete an interaction.

Of course, not all AI is created equal. Some implementations of AI, like the unhelpful chatbot Maoz encountered, pay more attention to their outward form than to the quality of the user experience. They stake a lot on simply being AI, assuming the wonder of a machine mimicking intelligence is sufficient, without fully addressing the complexity of real user requirements. The efficiency of knowledge retrieval becomes secondary. The main emphasis is on scripted interactions, which fail users if their requirements do not match prescribed usage patterns.

In more user-friendly AI design, efficiency is primary. The outward form of the AI is secondary, or even invisible; users can complete a transaction without knowing that the site is enhanced with AI. The goal is to serve the information the user needs with as little effort as possible. It’s not to create a program that performs intelligence for an audience.

At Transversal, we embed our AI behind the scenes. As far as our clients’ customers see, they’re not interacting with an “intelligence” but simply performing a search, following a link or reading a knowledge article. The intelligence that retrieves, predicts and actively pushes the information they need is hidden in the background.

We choose this model of interaction for several reasons:

  • It lets the customer get on with their primary task, without distracting them with a gimmick.
  • It keeps the range of possibilities open. Users can search, browse lists of articles, or read automatically suggested articles. Chatbots, by contrast, force interaction through a narrow channel that often returns only one piece of content at a time.
  • It lets the editors of the underlying knowledgebase concentrate on maintaining quality content, rather than having to spend time scripting conversational responses.

AI might be the new must-have for customer service. But AI doesn’t have to be a chatbot, or a digitized voice, or anything that appears to have stepped out of a sci-fi film. The intelligence in “artificial intelligence” is what matters. Intelligence lies in providing the right information, in the right place and at the right time, with minimal effort by the user. That intelligence can work to deduce needs and serve appropriate help whether customers are aware of its presence or not. It doesn’t have to be “artificial, superficial, condescending, or insulting”. Unless it’s programmed that way.

“An example of how BMW Group is using AI to enhance the exchange between customers and staff in our Customer Interaction Centre, is by relevant information being proactively served up to our product experts who are web chatting.  Our knowledge tool reads the customer’s chat question and provides information the expert can use or share with the customer. This is really useful in an age where there is simply too much information to expect our experts to retain through training.”

Suzanne Gray, responsible for Group Customer Strategy at BMW Financial Services and BMW Group UK

If you would like to see cognitive knowledge at work, you can arrange a free demonstration here.