Getting value out of Voice AI one stage at a?time

Improving the customer experience through automation

To improve the customer experience your product provides, you need to improve the way you do sales, onboarding, and customer support today. Often, this is using live agents. If your business operates a call center or has customer service agents who address customer issues, examine whether you could be leveraging Voice AI to (a) improve the quality of the service you provide, (b) operate more efficiently, and/or (c) gain insight into customer pain points. Then, consider expanding beyond Voice AI into Conversational AI (conversations can happen via text message or chatbots, in addition to over the phone).

Voice/Conversational AI is a proven technology at this point, and has provided tangible business benefits quickly in a wide variety of industries. In many cases, it improves customer experience while also lowering cost. Voice AI projects are also relatively easy to execute since the technology is available as end-to-end platforms or frameworks.?

Getting Ready

Even though Voice AI is available in end-to-end platforms, implementation is not turnkey. Once you have chosen a vendor, you will need to work with the vendor to integrate their system with your IVR, messaging systems, and your backend systems. You will, therefore, need to ensure that:

• You can invoke your backend services through APIs. If your backend consists of legacy applications running on mainframes, you can provide an API wrapper around them. So, it’s usually technologically feasible, but in practice, there are often throughput and latency issues.
• Because you are exposing your backend software for reading and activation, you have to budget engineering time to ensure that service contracts are being met, and authentication and PII are being correctly handled.
• Your data platform provides ready access to your company’s knowledge base and can store and process call logs. If you have not built an enterprise data warehouse or data lake, this may be the impetus you need.
• You have “conversational designers” who can identify and map out conversations and workflows specific to your industry. This is usually a new skillset that marries business knowledge with some technical knowhow — it can be helpful to have a few people in your company trained to do this.
• You have stakeholder agreement to integrate with your current IVR or to rip and replace it with a cloud-hosted solution. Often, this will involve getting agreement among the various business units that employ a call center.

Then create a project plan. The time it will take to implement Voice AI depends on the number of intents (customer tasks) and backend APIs and external systems that need to integrated for that intent. These days, a small team can build and deploy 3–5 intents in a two-week sprint. Vendors should be able to estimate the number of intents and integrations based on a couple weeks’ worth of call center logs, and scope out the project.

Staged Implementation

I recommend that you build Conversational AI capability in stages. This will allow you to reap the benefits of investing in Conversational AI even as you build out downstream capabilities.

Approach the Conversational AI project as consisting of six stages:

1. ?Call deflection. A large percent of the calls into your call center may be simple informational queries (“Are you open today?”) and can be answered relatively quickly using Voice AI. You might need to build an API that provides hours of service, but integration into your IVR might be as simple as adding a prompt in your welcome menu.
2. Conversational design. This is where you map the most frequent user journeys in terms of the questions, answers, the data required for the API endpoints, etc. To find the most frequent and highest value (e.g., highest conversion or transaction value) user journeys, and the words that customers use, analyze call center logs. A small fraction of most such workflows will end up at a point where a human agent is required. Make sure to build the capability to convey the information collected by the bot to the human agent so that the user doesn’t have to repeat themselves.
3. Omnichannel. Once you have a few journeys mapped, it can be helpful to expand beyond voice. Most Voice AI frameworks also natively support text, web, social media, and Slack. There is a lot of ROI in turning on those capabilities. Doing so is much easier if you build the ability to identify intent through text classification and generate first draft responses using Generative AI.
4. Call analytics. Before you can drive efficiencies in your call center (beyond the obvious ones such as call deflection and conversion-maximization), you will need to define your Key Performance Indicators (KPIs). Analyze your call logs and interview your staff to determine leading indicators (such as “Asking whether they want fries increases order amounts by 7%”). You can use these KPIs sliced by agent to find agent best practices, and sliced by workload to find use cases that need better conversational design.
5. Agent assistance. Ensure that the Voice AI is listening to the call and automatically pulling up relevant documents and forms. Also, once you have agent- and workload- best practices defined, you can build real-time coaching of agents.
6. Product insights. Use call logs to identify trending issues and poorly handled requests. Use these build new conversational flows or improve the coaching. Ideally, this ends up with a flywheel effect of increasing revenue/profit and lowering customer handling time.

Note that the stages are all optional — -make sure that the stage makes sense for your business. For example, one company determined that all calls would be taken by a human (to provide white-glove service), and so there was no call deflection in their system! They were purely focused on call analytics, agent assistance, and product insights. In another case, the company had very few call center agents, and so the ROI wasn’t there for call center analytics or agent assistance.

Summary

Voice AI is a proven technology -- it can improve the customer experience while lowering cost and is relatively easy to execute since the technology is available as end-to-end platforms or frameworks. However, it is not turnkey and requires integration with existing systems to derive the full benefit. This article provided a checklist to ensure that you are ready to implement Voice/Conversational AI and a staged approach to ensure that you are gaining value as quickly as possible.