Chat bots are starting to appear everywhere. From making an insurance claim, to customer support, and even providing guided professional coaching sessions.
Some of these chat bots are friendly and help you get what you need arranged really quickly. Others feel like a poor-quality customer service cost reduction sitting between you and the person who can actually resolve your problem.
Over the next two years bots will become much more common place, and not just in the field of customer support. Forward looking Enterprises, SMEs, and smaller organisations will start offering bots to complement their existing online services,
If you’re looking at adding an AI chat bot to your service offering, here are my top tips you can follow to make sure your bot feels like an improvement to your service, and not a barrier to get past.
Tip 1 – Use Language Recognition
When people use a chat bot they want to have a conversational experience, they do not want to feel as if they have to learn a set of specific commands for a UNIX terminal!
It’s vital when building a chat bot that you use human language understanding. This isn’t just about understanding what someone types either, most bot should also be able to handle the spoken word as well as typed text.
Microsoft CEO Satya Nadella explains:
“It is about taking the power of human language and applying it more pervasively to our computing” – Satya Nadella
In a conversation, if you want to set a date for a parcel will be delivered you my say:
I want to arrange a delivery I’d like to change when you will drop off my scooter I won’t be in to when you are coming
Or any of the other natural ways we each express what we want in our day to day conversations. A good chat bot understand what is required and doesn’t need you to be trained to talk to it.
Likewise, if you were looking to get training in fire safety, you should be able to say:
I’d like to find a course about fire safety. Show me best practice on being safe with fires I want fire safety training
Or any alternative that would come up naturally in conversation. The AI behind the bot should know what you are looking for from what you say, however you express yourself.
Natural language processing is a big field of AI, and one of the most well-established. Services such as LUIS from Microsoft make it easier to provide “Conversation as a platform” and with proper AI training can make your bot feel natural, and helpful, to converse with.
Never build a bot that doesn’t understand natural language. Experience shows people will not use it, and if you don’t give them an alternative, people will not use your service!
Tip 2 – Don’t pretend to be a person
We all want our bots to be friendly, but little is more frustrating when using a bot than trying to convince “Susan” to let you speak to someone real while “she’s” asking you what you did at the weekend. A bot does not care or need to know what you did at the weekend, and you certainly don’t want to waste your time telling it.
Yes, give your bot a name, Alexia, Siri, Cortana, these give our AI bots identities. But in the same way as Amazon, Apple, and Microsoft, you should be proud to use a bot and proud of its identity.
This comes a lot more naturally if your bot exists to enhance your service, rather than save money by replacing quality people with technology.
Give your bot a name that suits its identity, and tell people up front they are dealing with a bot. If the bot provides the ability to have a person take over the interaction at some point, make it clear when this happens, and start using the person’s name instead of continuing with the bot’s identity.
I had a recent experience with major online retailer’s customer service bot that shows what happens when this goes wrong. I asked for some customer support and a chat window opened to have an online chat about my problem. I was never told I was speaking to a bot, but it took the bot a lot longer than it took me to realise it wasn’t able to help me with this particular query. In the end I reached the point in the conversation where the bot had worked it out and told me it was going to go “check on something for me”. I then had to sit and wait for over 10 minutes, mid-conversation, to get to the top of the queue and talk to a real person who then continued to chat with me under the bot’s name. I spent most of that wait wondering why this company thought its customers were too dumb to recognise the difference between a bot and a person, and why they thought their customers were too unimportant to help. Even after my problem was resolved, that’s what I remembered.
That experience would have been much better if the bot introduced itself as a bot to solve common queries and asked my query. If it could solve my query I’d have been left thinking how great the company was for using technology that helped me so quickly without having to wait in a queue at all. If the bot couldn’t solve my query and told me it would connect me to someone who can, I’d still be happy that the company knowing my query was unusual and after a little wait I’d get someone with the time to help sort it out for me.
Bots can be fantastic at solving common problem and allowing you to make best use of your staff, but if you don’t feel you can be proud of using a bot as part of your service and want to hide it, then your bot is probably not providing a good service to start with.
Tip 3 – Use visual prompts as an alternative to typing
Sometimes we make a mistake of thinking chat bots are all about typing (or speaking), but quite often if people are interacting with their phone, clicking is easier.
If you bot has three common tasks, it can do, such as “rearrange a delivery”, “find out where my parcel is” or “return faulty goods”, then when the bot introduces itself and asks how it can help, show those common options on screen for a user to click on.
Yes let the user type anything they want in natural language as an alternative to choosing an option, but don’t force me to type when I could get to where I want with just a few clicks.
Likewise if you want my location, let me click so you can work out my current location, don’t force me to type out a full address.
When you want to confirm something, it’s great that I can type “yes”, “yep”, “sure” or other positive words, but a button that I can tap with my thumb that says “Yes” without typing is better.
Bots should provide a friendly alternative to more traditional technology. But don’t force people to type when they don’t want to, or can’t.
Tip 4 – Go to where your users are – be multi-channel
When you design a bot, you’re excited to put it on your website and to start directing your customers to it. That’s fantastic, but the jobs not done yet.
Ask yourself where your customers will already be when they need your service, and make sure you’re bot is available there too.
Make your bot available on Facebook, make it available via speech in Cortana, Alexia, and elsewhere. Let me Skype the bot. Let me interact with it on Slack, let me email it or send it a quick SMS.
Don’t force me to go to you to converse, come to me for the conversation.
To achieve this, you should insist your bot is built on a maintained multi-channel bot framework.
You will also want to make sure the framework isn’t just text based, speech is increasingly as important in designing bots, if you pick a framework that is text only, you will have to redevelop your bot much sooner, throwing away all your investment.
Tip 5 – Keep the bot an optional service, don’t force people do use it
I understand you will be excited when your new bot is ready to launch, and want as many people to use it as possible, but if you want your bot to succeed its vital that you keep its use optional.
A bot that replaces a previously personal interaction or an established web page isn’t a bot that’s extending your service, it’s a bot that’s creating a barrier between your customer and what they want to do.
Some of us love bots, others hate them. Use the bot to help those who want a quick, friendly, and direct self-service experience wherever they are. Continue to use web pages for people who are more comfortable filling out information online and don’t want a suddenly jarring experience to a “conversational mindset”. And use any time freed up by your bots to empower your staff to work where they can provide the most personal experience for those who need it.
If you do all these things, your bot will extend your service, and your customers will be grateful to be working with a technically forward-thinking company. Let the bot get in the way, and people will be convinced you care more about saving money rather than improving your service.
There are plenty more things to think about when building an AI powered chat bot. But following these five tips gets you a long way down the road to a bot that you and your customers will love to work with.
If you want to talk about how a chat bot might extend your own online service, I’m always happy to give advice, tips, and if the times right get you started in developing this or your online service generally.