boost.ai’s Head of Voice lays out why voice tech sucks and how to fix it
At this year’s Voice22 summit in Arlington, Virginia, I had the pleasure of holding a keynote presentation on how to unlock true customer value with voice technology. I’d like to share some of my thoughts from that presentation and summarize the key steps that can be taken to ensure a successful voice project.
But first, let me start with a fairly controversial statement (followed by a quick qualifier)...
Voice tech sucks… if we don’t implement it properly, don’t have the right teams in place, and don’t educate enterprises on best practices around implementation.
The good news is that it doesn’t have to be this way, and it’s possible to implement voice tech so that we can make every customer, and agent, feel like a VIP.
Today, however, that VIP treatment just isn’t true when it comes to customer service. Poor service is the number one reason for customer churn and unhappy customers cost businesses over $500 trillion dollars each year - that’s more than double the value the global real estate market, lost due to poor customer service.
Over-the-phone service is part of every enterprise’s service strategy but waiting times are long, call drop-off is high, and both customers and agents have bad experiences - it’s an antiquated channel that’s begging for reinvention.
By implementing voice tech correctly, we can bring value to contact centers, both through voice bots and analytics. We can make products and services more easily available and drive up adoption rates.
There are many strategies to achieve this, with artificial intelligence – and specifically conversational AI – being the driving force and helping to solve key service challenges like,
- How do we ensure that customers don’t keep getting handed over between agents?
- If a customer has a simple question, how can we automate to ensure they get their answer with minimal friction?
- How can we take the user's information and find out more about them?
- How do we augment call center agents with information and tools to provide a better customer experience?
One of the areas where conversational AI can really help is in voice-based telephony. The elements for successful voice experiences in this channel include,
- Speech flexibility - allowing organizations to use the speech services that they prefer and avoiding a “crap in, crap out” scenario
- Strong NLU - means the ability to actually understand user intent. We’ve all experienced a bad chatbot, and adding voice on top of it will not make things any better
- User friendliness - aka the organization of data and information and content design
In developing a successful voice project there are several key areas that need to be considered.
Prepare for success
Before deploying a voice bot you need to ensure you’re starting with the right technology and that you understand what use cases you are attempting to solve.
Ensure that all stakeholders are aligned and then work together on scoping sessions so they understand what is expected and what the goals of the project are.
An example of this is FAQs. It’s often thought that they are not the optimal target for a voice bot, but this isn’t necessarily true. If a company’s FAQs are organized structurally, then it’s actually not the worst place to start. It can act as a soft launch and demonstrate what’s possible while aligning decision-makers on the potential of the technology.
Optimization is crucial
As soon as a project is underway, we must begin the process of optimization. You should never consider deployment before training the model and understanding (and hopefully gathering) data on the end users.
If you are deploying both voice and chat, you need to optimize for each channel specifically. The content can (and should) differ, whether you have a narrow set of use cases or a broad scope that covers varied support scenarios.
It’s through optimization and thorough testing that we elevate the end user to the VIP status that I mentioned earlier.
Going live doesn’t mean going home
We see the same scenario often - companies invest time and resources in building a voice or chatbot, push it live and then let it languish. This is the wrong approach.
Once a bot is live, there’s an incredible opportunity to gather real-world data from users and use it to optimize even further. Pre-deployment, the data gathered often presents an incomplete picture of how a voice bot can help serve customers. It’s only after the channel is live that you can start filling in the gaps as users ask for things you may not have scoped for and as their expectations increase now that they are interacting with the channel.
Educate, educate, educate
From project start to finish, education is crucial and often overlooked. Whether you’re working with AI trainers, engineers, the contact center, or marketing, everyone needs to be up-to-speed and aligned on how the project functions and fits within the organization.
Do business decision-makers get educated enough as well? More often than not, that isn’t the case but it is just as important as other facets of the business.
Test for the unexpected
I briefly mentioned testing earlier, but it really deserves its own focus. Before a project goes live, we need to make sure that it works and testing is how we avoid (as much as possible) any unexpected hiccups.
And don’t just test the happy path - test deep and broad, and those things you think user may not ask because I can guarantee you that they will ask it.
Remember that testing for voice and chat varies, and requires different prompts. For voice, you need to be aware of how it sounds and remember to omit things like gifs and emojis which don’t carry over from chat.
So remember, not all voice projects have to suck. By following the above guidelines we can make strides in ensuring that we create automated experiences that treat every customer as a VIP.