AI and CX: How to Choose the Best AI for Your Business
Everyone’s talking about AI at the moment. How it’s going to replace, impact, or revolutionize different job roles. You’ve probably played around with ChatGPT or DALL-E and been impressed with some of the output — but also realized the limitations pretty quickly. The important point about AI is — it’s only as good as its training data. However, it's crucial to understand that not all AI training models are created equal. Broadly speaking, there are two primary types of training models: standard AI models and those based on customizable intents.
The question then arises: how can you leverage these models to get AI working effectively for your specific business?
It’s all about how the AI can interpret intent
The issue with standard AI models — even those pre-trained on a handful of industry issues, is that you can’t teach them anything new and unique about your business — your products, your USPs, your policies, etc. They don’t allow you to customize anything — so whatever they can do out of the box is all you’re going to get.
If AI can’t understand the intent of the message, it can’t provide a personalized solution.
It’s kind of like getting a temp in to cover for an experienced CX agent. Sure they can answer the phone and handle basic stuff like payments or WISMO — but they don’t understand the details of your product or service, they can’t give really specific solutions, and they often have to transfer the call to someone else in the end.
What if you could train your AI in the same way you can train your agents?
Well, that’s an option. The key phrase here is “customizable intents”.
AI solutions with customizable intents allow you to tailor how they interpret and respond to customer interactions — and the benefits are significant, as we’ll see.
Why does customization matter so much?
Well, why does agent training matter? It’s pretty much the same answer. You need your AI to be adapted to the types of queries or issues that your customers commonly have — and to be able to suggest solutions based on your existing processes or policies.
To do this, you need to tailor your AI models in advance by defining common customer intents and indicating the correct processes to resolve them. You need to program the AI to be able to understand customer issues at a granular level — not just a surface interpretation. And the roots of each issue will be unique to your business.
Side-by-side comparison: standard AI vs customizable intents
Here’s a brief overview of the main differences between the two types of AI.
We can see that the AI using customizable intents is more flexible, more adaptable, and more tailored to industry-specific needs — and that translates into better performance in niche scenarios. Now let’s see how the two solutions perform in a real-world scenario.
How standard AI compares to customizable intents when dealing with real tickets.
Let’s take a look at a example of a customer support ticket that might be received by a medical insurer.
“Can you confirm if my plan covers the prescribed medication Prozac? I have been out of work for 10 days and haven’t gotten a response. Does my insurance cover this?”
The standard AI model interprets the ticket as follows:
Intent: Insurance Coverage
Based on this, the AI might suggest the following general response regarding insurance coverage.
"Thank you for reaching out. We're sorry to hear about the issues you're experiencing. Please contact our benefits department for information about medication coverage."
As the AI is not customized to the needs of a mental health provider and doesn’t interpret the meaning of “Prozac” or “sick leave”, it results in a poor customer experience.
AI with Customizable Intents:
The AI with customizable intents interprets the ticket as follows
Intents: Sick Leave Coverage, Mental Health Medication, Urgent Request
Thanks to customized intents based on the provider’s niche, the AI understands that “being out of work for 10 days” means taking sick leave in this specific context, that the ticket relates specifically to mental health medication, and that the fact that someone with mental health issues has been waiting for a response for 10 days makes it an urgent issue.
Based on this custom taxonomy for the company, in this case the AI can route this customer directly to a mental health professional, providing a far better customer experience in what may be a high-risk scenario for the patient.
So, what this comparison shows is that the AI with customizable intents outperforms the standard or pre-trained AI on two levels.
- It understands more context, and can therefore more accurately categorize and respond to the ticket.
- It is able to direct the customer to the relevant next step, rather than requiring them to initiate contact to resolve the issue.
And most importantly, the customer experience is improved — more personalized communication and a clear path to resolution.
If you’re a CX operator interested in how an AI model using customizable intents could transform customer support in your organization, feel free to book a slot to speak to one of our CX experts. We’re not going to give you a hard sell — just some honest insights into how this technology could work in your specific niche.