Conversational AI
January 23, 2020 • 5 minute read

3 Considerations When Implementing a Conversational AI Application

This blog was written with help from Sheetal Sekhar

Conversational AI is a subset of artificial intelligence (AI) that refers to building a voice or text assistant that can engage in human-like dialogue, capture context, and provide intelligent responses. A common example of where Conversational AI is used is in personal virtual assistants such as Apple Siri, Amazon Alexa, and Google Assistant. Conversational AI is also increasingly being used to automate and solve customer service inquiries. 

In the last few years, we have seen many Conversational AI vendors offering solutions and services to deliver enterprise virtual assistants, from DIY platforms to managed services vendors. Recent announcements by Google and Amazon presenting their capabilities for automated customer service solutions in the contact center has raised awareness to the growing field of Conversational AI. Gartner estimates that there are approximately 2000 vendors in the market worldwide, ranging from emerging to experienced vendors, offering a variety of AI-driven services. This can make it challenging to choose a path forward.

What business are you in?

Many leading solution providers offer DIY ‘open developer platform’ solutions, which seem simple and intuitive in the beginning but get complicated with time. DIY solutions and platforms can provide a lot of freedom: control over the prompts, ease of making changes, control of your data. But with that freedom also comes a lot of responsibility: hiring a skilled team, day-to-day management of the platform, and understanding how to build effective solutions that can engage with your consumers. 

Once you get started with these DIY platforms, many questions pop up. What happens if you get stuck during the development? Are there resources you can call for support? How many vendors do you need to work with to stitch together a Conversational AI solution that will deliver a satisfying experience for your customers? And what will it cost? According to Gartner, virtual assistant technologies will surpass IVR and message-based interactions in the coming years. Gartner goes on to say “Many enterprises mistakenly assume that they can build chatbots using the artificial intelligence (AI) services available from major cloud vendors. Reduce the risk of failure by sourcing your chatbots from external providers, unless your organization already has the right data science and machine learning assets.” Despite all the DIY tools available, many companies never get beyond testing to actual deployment. This is because designing engaging and truly conversational customer service applications is complicated and requires daily care and managing of data and interaction flows to remain functional and robust. 

Today, when customers are choosing brands based on customer experience, is it better to try and create something yourself or go with a partner who can take on the heavy lifting? In the end, you need to ask yourself “what business are you in?” A business that is focused on being the leader in selling its product/service? Or in the business of building Conversational AI applications? Sometimes it can be difficult to do both, so make sure you ask yourself that question. 

Voice is a Deciding Factor

In a study published by Oracle on the Impact of Emerging Technology on CX, 82% of respondents agreed that voice will be a deciding factor in customer retention. Yet, today many solutions, from DIY platforms to managed service providers, can only identify the caller and direct them to the right agent. Simpler tasks such as order status or power outage notification can be easy to automate using simple DIY solutions. But try placing a food order for pick up at a local restaurant or making a hotel reservation and you end up finding yourself confusing the system and being placed on hold for the next agent. Even when you think you are asking a simple question of the self-service system, you may hear “I’m sorry, I didn’t get that” more often than not. Is it because you are in a noisy environment? Or maybe your accent isn’t being understood? Delivering a truly conversational and effective self-service solution using voice is very challenging and requires the right expertise to succeed and retain customers.

Tuning and Training – join the pool?

An important element of an effective Conversational AI solution is the ongoing tuning and training. Some of the platforms on the market today encourage you to share your data with the larger pool of data from other enterprises using the same resources. These Conversational AI platforms present this as a positive feature that enables enterprises to tune and train models by benefiting from the Machine Learning powered by access to a broader set of data. But are you really ready to jump into this pool? Do you want all your efforts to fine tune your models to be leveraged by someone else? And what about security? Are you at greater risk of sensitive data being leaked?

 

Implementing a Conversational AI solution is not a straightforward process. It becomes the face of your business, and therefore should be approached with great detail and care. A good Conversational AI application can keep existing and create new customers, while a poorly designed one can lose a customer forever. Refer to our checklist to ensure your virtual assistant is designed with the right considerations in mind.

Want to learn more? Let’s talk.