April 15, 2020

Rise of the (Chat)Bots

Get notified whenever a new podcast is released!

They’re low cost, relatively easy to implement and can ease the pressure on massively stretched teams in these difficult times. So why aren’t you using them?

We’ve recently had significant success implementing chatbots, with the majority of end users preferring the experience to talking to a call centre. The use cases have been broad, from complex implementations with significant back-end integration to highly personalised answers to much simpler affairs using configured lookup tables.

So how do you get started?

You already know what customers want to know;  you’ve been through the pain of putting your FAQs together. Your call centre staff have years of experience. Your ticketing system knows which types of call are received most. You can quickly narrow down to an achievable number of questions. It doesn’t need to be daunting

Your customers probably want to know other things too; don’t try to answer everything, you never will. Humans are still essential, chatbots are great for fairly complex interactions, but there’s always a point you’ll want a customer to talk to a person. Only take it so far, if it’s making the customers life harder rather than easier, you’ve got it wrong

Keep it simple; when starting out, consider a menu based approach. This is quicker to implement:
You can easily map existing content
Intent is easier to infer
End users will only see 3-4 options at each level (any more than this is frustrating)

Until you can’t; you’ll get to the point where the complexity of questions requires more sophisticated natural language processing (NLP). This allows intent to be trained and inferred from anything the user types in. The engine will continually learn and evolve at this point, and can appear to have a semblance of artificial intelligence. More content, data and functions can start being fed in at this point, and once over the initial complexity, becomes easier again

Branding; don’t bother too much, it matter less than you think in this context

Style & Personality: make it appropriate. Humour, avatars and emojis can all have their place, but if you’re delivering serious information (employee information about COVID-19 for example), keep it serious

Clarity of channel; users are used to chatbots and all sorts of automated communication, let them know when they’re talking to a bot or talking to a human. They won’t mind if they’re getting the answers they need

Explain why; when launching a chatbot a user should be told that this will save them time and get them quicker to the answer they need

A simple chatbot, based on Salesforce Einstein for example, can be up and running in a matter of days, particularly if you’re an existing Salesforce customer. If your requirement is related to the current pandemic, there’s also free support to do that from both Salesforce and BML.

Get in touch if you’d like to discuss getting started.