A rapidly growing number of businesses across the world already rely on mobile messaging to deliver customer interactions like this:
Business: Your delivery will arrive between 1pm and 3pm tomorrow. Text yes to confirm or no to rearrange.
Customer: NO
Business: Here are some alternative time slots.
But the marriage of natural language processing (NLP) and mobile messaging is about to increase the conversational possibilities.
By plugging NLP (a type of AI) into your messaging platform, you and your customers have far more freedom to communicate in a natural way. A conversation can pan out like this instead:
Business: Your delivery will arrive between 1pm and 3pm tomorrow. Is that ok?
Customer: No, I’m going to be out
Business: OK. Would you like to rearrange?
Customer: I’m in tomorrow between 10am and 2pm
Business: That’s fine. I have arranged a delivery between 10am and 2pm. Is that OK?
Customer: Yes
Business: Is there anything else I can help you with?
Customer: No
But how do these NLP and machine learning algorithms actually work? And how can you teach them to become even more effective for your business?
Intentions, information snippets and conversation flows
In simple terms, there are three types of information your AI needs to learn: Intentions, information snippets and conversation flows. Let’s look at each of these information types.
1) Your customers’ intentions
Before getting started with NLP it’s important to understand the different scenarios in which you’ll be communicating with customers, as well as the likely intentions of those customers.
For example, will they be speaking to you about deliveries of packages, your latest offers, or getting a device you sold them repaired? You probably already know the types of interactions you’re likely to have with the customers.
You only need a few examples to get started. The machine learning algorithm can then learn to understand the examples you enter, as well as other phrases that mean the same thing.
For example:
What’s your opening time?
When do you open?
What are your hours?
Or:
I’d like to repair my phone?
My phone is broken.
How do I fix my phone?
If you provide examples of the various ways a customer might communicate each intention, the AI can learn to respond.
You might have anything from a few, to thousands, of intention examples the AI can be trained to handle. As people use your service, you can gradually incorporate what they say as training examples. So the more usage you get, the smarter your AI becomes.
2) Snippets of information
As well as working out a user’s intention, your AI will be pre-programmed to look for snippets of information such as numbers, dates and times, amounts (including units), and geography.
You can train the AI to learn other words or phrases it will need to know to handle customer conversations. For example, if you run a mobile phone commerce site, you can add a list of phones you sell, and a list of services you provide.
How does messaging AI extract information snippets?
Let’s look at an example:
Customer: I bought a Samsung mobile phone from your store yesterday. But today I realized I don’t know how to use it. Can I book an appointment to see one of your consumer tech team tomorrow?
Business: Would 1PM be ok for an appointment to see one of our tech advisors?
Customer: Yes
Business: You’re booked in. Can I help with anything else?
Customer: No
Business: See you tomorrow
The user’s question mentions several important snippets of information: “Samsung mobile phone”, “consumer tech team” and three dates –“yesterday, today and tomorrow”. The AI is able to pick out the correct day and type of appointment required. It works out where the snippets are likely to appear in the text, identifies the snippets, then ignores any distractions.
3) Conversation flows
There’s far more to a conversation than making a single statement and getting a simple reply. Conversation is a process involving two speakers who negotiate meaning and understanding. So imagine a situation where the user only gives partial information:
Customer: I’d like to book an appointment at 2pm on Friday.
In this scenario, the AI has the date and the time but still needs knowledge about the type of service required. We can help the AI solve situations like this by instructing it to prompt for certain types of missing information. The customer can even go back and change things they’ve already said. Here’s a good example:
Customer: I want an appointment.
Business: What type of service do you need?
Customer: My phone’s screen has cracked
Business: What date are you available?
Customer: This Friday
Business: What time do you prefer?
Customer: Actually could we do PM on Saturday?
Business: How about 2pm this Saturday?
Customer: Great
Business: You’re booked in.
Making the AI your own
There are many ways you’ll be able to guide and control the flow of a conversation between a customer and an AI. You’ll even be able to give the AI a tone that suits your business.
Hopefully, you now have a greater understanding of how straightforward it can be to leverage NLP and machine learning to get closer to customers and enjoy empathetic interactions that they will deeply appreciate.
If you need any more help or advice, get in touch. Our account teams will be happy to help.
Or you can download our ebook: A strategic guide to AI-powered mobile messaging.
This post is number three in our NLP in texting series. Check out the other two:
Post 1 – How to use NLP for empathetic customer service messaging
Post 2 – SMS and AI: Goodbye to the ‘Don’t call us, we’ll call you’ era