How to Get Started With Conversational AI For Customer Service

Conversational AI can improve the customer experience and reduce costs, but not every use case will return ROI or even prove effective. Understanding the limits of the technology, how people use it, and how it fits into your organization are crucial to achieving scalable value.

So how do you implement it, and what results can you realistically expect? That’s what Director of Practices and Solutions, Michael Ricks-Aherne, and AI-Enabled Customer Service Solution Owner, Jay Garvey, talked about at Project Voice when they presented 5 Things to Consider Before Implementing Conversational AI in Customer Service. Here are the five things and some takeaways.

Consider the Business

It might sound obvious, but it can be incredibly difficult to consider the business in totality because there are so many elements that come together:

Internal stakeholders
Timeline
Problem space
Complexity
Regulations

For conversational AI solutions in the workplace, the internal stakeholders are widespread. It will usually include the call center, security, sales, and marketing departments, as well as the IT teams and maybe people from other operations and product departments. And getting alignment from all of these people can be difficult, because while their ultimate goal is the same, they may see different paths - or have different requirements - to get there.

For instance, Michael told a story from one of our clients where we were working with a client that was ready to deploy a solution for their data and analytics - until their security team told them that technology wasn’t approved.

Even determining the problem space and what the specific goals are in this space can be difficult because different departments have different needs, and determining a timeline - and when different departments may be involved - can be difficult because it has to work for everyone.

Consider the Experience

You may be implementing a solution focused on cost and time efficiencies, but it’s important to still think about the experience. Jay noted that experience isn’t just the customer experience either - the employee experience is also vital.

You need to think of several elements about the people themselves and their surroundings. In particular, Jay suggested businesses consider:

Customers and call center employees
Demographics of the user base
The environment of the caller
How your customers speak and like to be spoken to
Conversational design

Jay talked about how he was about how his preference for communications is different from Michael’s, but it’s also different from his mom’s. Some people are going to call and just want to talk to a human, whereas others are going to be willing to interact with a bot. You need to be able to meet the needs of your consumers - and that’s different industry to industry, business to business. Jay emphasized that you should measure your customers’ preferences and not just make assumptions.

Consider the Technology

Mike spoke about an example where a customer replaced their existing solution with an off-the-shelf buy option. It met the needs they were looking for but also lost them some capabilities - including automated callbacks. This meant that agents were spending time manually calling customers, which cost them significant time. Long term, they needed that capability back, which meant the off-the-shelf solution wasn’t really suited to their needs.

That isn’t the only technology consideration you need to consider, though. The complete list Mike covered is as follows:

Long-term implications of the stack/product
Transcription and intent-matching technology
Reliability/availability/scalability of services
Understanding the test/promotion/deployment process
Integrations to enterprise systems

The technology needs to fit in with your current stack long term, and it needs longevity as a solution. Contact center solutions are large investments, and getting something that you need to rip out and replace quickly is only going to lead to more long-term pain and, as Mike’s example illustrated, often short-term pain.

Consider the Metrics

Do you know how your current solution is performing? As Jay pointed out, it’s important to have a baseline of metrics so that you know what’s going well, what’s not going well, and what the results of a new solution are. The things Jay specifically mentioned as important relating to metrics are:

Determining who in the organization will consume the metrics
Establishing a baseline
Determining what success looks like
Choosing an approach with monitoring in mind
Considering the customer journey and supporting systems

A new customer service solution is generally focused on finding cost and time efficiencies, and improving the customer experience. Measuring against those two goals with various metrics is therefore vital to understanding if the solution is performing and how it can be further improved. Some important things we encourage you to consider are if there are any friction points customers are encountering, if they’re abandoning your IVA or IVR in certain places, if they’re put in the correct agent queue, and what system performance looks like.