Face the future of customer service

As the costs of customer help services, such as call centres, for online customers continue to rise, many businesses are now looking for new and low cost systems to assist their online customers. Consequently, many businesses with online channels have introduced virtual assistants to their websites. These digital assistants are based on intelligent knowledge based systems that can help customers by answering frequently asked questions typed into a simple query box interface.

Many of these virtual assistants are also ‘personalised’ to make them appear more approachable and sometimes given a name and a picture like Ikea’s cartoon helper Anna or BT’s Emma.

We recently conducted research into how customers react to virtual assistants, and found that even when people knew the character was just a bunch of bytes, there was a high level of emotional engagement. Our test subjects would refer to the assistant as ‘him’ or ‘her’ (rather than ‘it’), and one said that it added a ‘human touch’ to the page.

Using a photograph for the virtual assistant helps to build trust with the service. When test participants were asked to comment on 11 different pictures of virtual assistants, the feedback was surprisingly personal. ‘She could tidy her hair up!’ said one, and another said ‘she looks helpful and friendly but too young and giggly’.

The industry tends to prefer female brunettes, and our research showed why: people don’t like their virtual agents to be too extreme in hair colour, age, dress or anything else. Fake smiles are out, but genuine smiles are favoured. The most popular picture was one that looked helpful (not too young), trustworthy, and professional (but not too formally dressed).

Here are some tips for using virtual assistants effectively:

• Make sure users know it’s not a live chat with a real person. They will expect much smarter responses from your staff than from a knowledge base, and will be frustrated if ‘staff’ appear unhelpful.
• Enable the agent to respond to all types of user input e.g. full questions as well as single keywords.
• Users appreciate a list of ‘related questions’ options, which can help them to surf to the answer they need.
• Position the assistant where users can find it: above the fold, and wherever users might seek help.
• Ensure that if the virtual assistant does not seem to have the ‘answer’ that other means to find help are available, such as a prominent telephone number. Our research suggested that users would only try two or three queries using the virtual assistant before seeking an answer to their questions elsewhere on the site.
• Give the agent a name that is easy to remember. Our research suggested that popular names such as Laura, Kate and Sarah were preferred. However, names like Polly and Penny were considered old fashioned, while names like Sophie and Olivia sounded ‘upper-class’ (and so less approachable).