Human beings are conversational at our core. We enjoy the associated thoughtfulness and unpredictability. While today’s bots can’t achieve a human level of conversation—of not only thoughtfulness and unpredictability, but intuition— that doesn’t mean they must be completely robotic. Here are a few steps you can take to make your bot more conversational and, as a result, more attuned to your customers and their needs.
1. Tailor the tone of your bot to your audience.
When you have a diverse user base, you can use data you’ve already collected to categorize customers into tonal buckets. Create different bot personas for different user categories, and consider these questions to help shape the associated content.
- Who are they? Are they old or young, tech savvy or not— are they professionals? A human agent would use different language to interact with a 70 year old retiree who needs help navigating a food delivery app than they would with a 20 year old who is complaining about an order that is 5 minutes late.
- Why do they usually turn to your bot? What’s their situation or problem? What are they looking to get from the “conversation?” This will likely depend on the products or services you provide. Are you a bank? Most people take their money pretty seriously, so you wouldn’t want to get too cheeky with your bot’s responses to, say, a conversation about fraud alert; for this, a direct, professional tone is probably the best choice. But if they are just looking for help making an online deposit, a more casual but instructive tone would be better suited.
Remember, though: while the tone can vary depending on the bucket, consistency is key when it comes to your brand’s voice.
2. Personalize conversations with existing customer data.
Also based on the user data collected, you can include personal touches to your bot’s conversations. For example, if you’ve already collected the customer’s name, instead of saying, “Hi! I’m [bot]…,” the bot can say, “Hi [name]! I’m [bot]…”
And to again use a food delivery app as an example— you probably also know the customer’s order history. So you could have the bot go so far to say “Hi John, I see that you’re a fan of pad thai. Would you like to reorder or get something new today?”
Just be sure not to cross into the realm of creepy with this tactic. Customers may not have realized that they have provided this data already. It’s a good idea to account for this. For example, if the bot wants to send a follow-up to the associated email address, instead of saying, “I’m sending this to [email address],” it might be a better practice to confirm the data with the user first—“I have this email on file ([email address]). Is that correct?”
3. Prepare for when the bot does not compute.
Part of being a good conversationalist is being able to pivot when you don’t understand what is going on. No matter how many contingencies you’ve prepared your bot for, you haven’t planned for all of them. There are going to be times when the bot fails to understand or is unequipped to answer what the user is asking.
When your bot is struggling to meet your customer’s needs, it still needs to be able to guide the customer to a solution. You can prepare for these instances by:
- Being honest and having a standard message that communicates the bot’s limits ie. “I’m sorry. I don’t know that.”
- Informing users what the bot understands and asking follow up questions to properly categorize the issue ie. “I’m having trouble understanding. Is this issue related to [x,y, or z]?”
- Offering suggestions that imply what the bot is equipped to help with ie. “I can’t help with that but if you’d like to discuss [x,y, or z] please let me know!”
- Providing an escape hatch to start over, edit info already provided, or hand off to a human agent—rather than pushing the user through a fixed decision tree when it’s leading to a bad experience.
By making your bots more conversational, you can improve the customer experience so that it feels a little more human and a little more natural— also allowing for a more pleasant self-service experience than a traditional knowledge base search.