Step 3 of 5: AI Self-Service Without Compromise – AI Without Data is Like a Race Car Without Fuel
<<Step 2 | Step 4>>
This is the third of a five-part blog series that outlines the Five Best Practices for AI Self-Service Without Compromise. Use this guide to automate your contact center and Customer Experience (CX) with AI self-service in voice, chat, and text.
Best Practice #3: AI Without Data is Like a Race Car Without Fuel
Artificial Intelligence thrives on data, and virtual agents powered by Conversational AI are no different. Most might assume this to mean “conversation data” (call recordings and chat transcripts) to build out language acoustic models and intent possibilities in pre-production efforts before going “live.” To be clear, the focus of this blog is access to customer data – the same data repositories exposed to your live agents. The more data you expose to AI-powered virtual agents, the better they will be able to emulate live agents, serve customers, contain/deflect calls and chats, and save money for your organization.
For the most effortless and human-centric CX, virtual agents need access to your customer data for a host of reasons. Your live agents require access to customer data for the purposes of reading and recording data to support their linear processes. While AI-powered virtual agents can also read and record data and exercise cognitive capabilities to follow linear (or near linear) processes, they extend the value of customer data for other purposes that include:
- Personalization – mapping caller ID to an account and greet by name
- Prediction – algorithms to predict why someone might be calling
- Guardrails – defining the lanes and outcomes within call types or chats to determine what should stay with a virtual agent and what should be handled by a human
There are many different kinds of data that help virtual agents do their jobs better. There are the obvious examples, like customers’ personal information, which virtual agents use to authenticate customers before helping them self-serve or transferring them to the contact center. Phone number is a very powerful data point because virtual agents can match it to an account and offer personalized experiences. For example, when customers call Legal & General to make payments on their term life policies, the virtual agent matches the caller’s phone number to an account, greets the customer by name, and only requires one additional data point before accepting payment on the phone. Without a caller ID match, the virtual agent must ask the customer for two pieces of information, which adds time and effort to the call.
While phone number sounds like a straightforward data point to use, it can sometimes be much harder in practice. Be sure to confirm with your data and/or IT team that phone number can be used to access customer accounts and – importantly – that you maintain accurate and up-to-date phone numbers for your customers. If not, you’ll simply want to choose other unique identifiers that verify customers in an effortless way. Best practice verification is to use one piece of information that the caller knows and one piece of information the caller needs to look up. The more pieces of information that are provided to virtual customer assistant, the better!
Data exchange to read and record data is a critical piece of AI automation implementation, and is generally setup via web services, APIs, or FTP file transfer.
Virtual Agents Offer Predictive Personalization
Integration to customers’ personal information is table-stakes for a good CX. But how nice would it be if every time a customer called in to your contact center, you knew why that customer was calling? What if you could review and compare all the data points from recent activity for any given customer and make a prediction about the reason for call? This is the true promise of AI-powered customer service: to analyze and act on large amounts of data in a split second. Conversational AI can recognize a customer, review and analyze recent activity, and predict the reason for the call. Even if it’s not why the customer is calling, he/she is more likely to use and trust the automated customer service for additional self-service.
A great example is in the healthcare space, specifically around Explanation of Benefits documents. Customers often have questions about these documents because they can be confusing. If the virtual agent matches caller ID to an account and sees that an EOB recently went out to him/her, it can offer a predictive greeting such as: “Hi Jeffrey, are you calling about the recent Explanation of Benefits we sent?”
Moreover, this predictive action is something your live agents can’t do without being given the proper time and instruction. Meanwhile, AI-powered virtual agents analyze vast amounts of data in nanoseconds to outperform live agents in terms of predictive personalization, which means less friction for the customer to get the job done.
More & Better Data Leads to Effortless CX
Ultimately, data access impacts customer effort – how easy it is for customers to accomplish their goal. Studies have shown that customer effort is a stronger predictor of loyalty and increased spending than any other metric (CSAT, NPS). By equipping your call center AI automation with customer data, you can make an enormous impact on reducing the effort it takes to interact with your organization.
J&B Medical Supply must authenticate every caller according to HIPAA regulations, which means agents must capture three pieces of information (of a possible eight) about the patient in order to provide service. This process was taking agents three minutes on average – a painful amount of time to simply verify identity. Then, J&B implemented a call center virtual assistant that integrated with patient data, so they can match phone number to an account and quickly move callers through the HIPAA authentication process because they only need to capture two additional pieces of info. AI self-service has helped them cut authentication times in half, to just over 80 seconds per caller. J&B has realized ROI both through agent savings and customer loyalty due to this initiative to reduce effort.
Clearly J&B Medical Supply was able to give data access to the virtual agent in order to perform authentications. If data is missing from a customer record, there are “business rules” (discussed in Best Practice #2) that guide the virtual agent to either move on to the next piece of info or transfer to a live agent. Creating the “guardrails” is dependent on customer data, so without access to that data, you cannot identify or build the proper handling rules for virtual agents. There are dozens of organizations that have data in disparate places or don’t prioritize keeping phone numbers up to date, which limits their ability to create the necessary “guardrails” for automation.
This is why it is imperative that, as part of your virtual agent implementation project, you also elevate your data initiatives. Whether you need to consolidate data, clean it up, collect more, or create easier access to it, these projects are crucial to helping virtual agents provide a CX that rivals or exceeds live agents.
This was the case for another medical supplier. They began using virtual agents to automate only the supply reorders that went through private insurance, but not those that went through Medicare. This was because the data needed to automate Medicare calls was not readily available at the time of implementation. However, once the organization was able to show success and ROI with AI automation, they were able to secure the resources to expose the data needed to automate Medicare calls as well.
Step #4, “Avoid Silos for an Omnichannel CX,” explores how to take state-of-the-art design and data integrations across every customer channel, as well as the reasons a unified platform are essential in today’s contact center environment.