Technology vs value: a practical guide to using AI in customer service

In recent years, one of the most popular ways to utilize the power of technology and AI in customer service has been to create Artificial Agents in the shape of bots. These bots are designed to handle customer inquiries across multiple channels, such as phone, email, chat, and online banking systems.

However, Artificial Agents have earned a mixed reputation because they either lacked quality or required extensive manual work and maintenance to make them effective. But, with the recent developments in generative AI, Artificial Agents have become much smarter and no longer act merely as routers to existing knowledge articles or FAQs. Instead, they can now routinely filter and summarize relevant information and deliver it in a way that is easy for customers to understand.

‍

Deciding on the right approach

Despite the recent technological advancements, many organizations are still not getting things right, as they lack a clear strategy for how they want to use AI to improve customer experiences.

The main reason this happens is that they tend to take a technology-first approach where they rush in, driven by technology hype, FOMO, and a desire not to lose their competitive edge. Consequently, they implement generative AI features without really thinking about what they and their customers really need.

Using this type of approach, most companies will train the technology based on specific static data sources – for instance, from knowledge databases or FAQs. However, this approach carries a significant risk of frustrating customers by not providing them with what they need and depriving them of the human-to-human interaction that is crucial in certain service interactions. 

To avoid that scenario and ensure your conversations are always high quality, we recommend that you take a step back and develop a strategy based on insights from your customer interaction data.

‍

Taking a strategy driven by insights from your data

Organizations that choose an approach based on their data start by understanding the reasons behind customer questions.

They do that by first analyzing all their calls so that they can group them into specific categories and then decide which ones they would prefer their human agents to handle.

 Here is a topography of calls coming into a typical organization:

20-35 % of all calls are made up of repeat calls, retention calls, sales calls, and calls that offer upselling opportunities. Commercially, these types of conversations should be answered by a human agent.

Now, the remaining 65-80% can be time-consuming and difficult to prioritize from an automation perspective. However, grouping them into categories depending on their complexity levels (low, medium, or high) can help with that.

Capturi defines low-complexity calls as conversations that are notably shorter than average calls and require no system integrations to be able to assist the customer. These calls usually account for 5-10% of all conversations.

In addition, Capturi defines high-complexity calls as conversations that are significantly longer than average and where agents need to access multiple systems to be able to provide an accurate response. Typically, 50-70% of all conversations would be grouped into this category.

The remaining 3-5% fall into the medium-complexity category.

After categorizing their conversations, the organizations can create a plan to determine which calls should be handled by Artificial Agents and which should be handled by human agents.

‍

Getting started

Always start with your low-complexity calls, as these are ideally handled by Artificial Agents, given that they do not require any system integrations and are generally easy to resolve. Concrete examples of conversations in this category include calls where customers have experienced log-in issues, have expressed billing concerns, or want to update key information.

As you progress and are successful in using Artificial Agents to handle your low-complexity calls, you can integrate your Artificial Agents with more and more systems. Such integrations will enable them to also handle medium-complexity calls – and in some cases, even select high complexity calls.

‍

What to look out for

It is important to remember that customer inquiries will evolve over time and that your Artificial Agents' approach should do the same.

Therefore, you should ensure that your Artificial Agents learn from your human agents so they can adjust their approach and develop new skills accordingly. That will help maintain a consistent brand tone, prevent misunderstandings, and ensure that customers receive the same level of understanding, empathy, and positivity they are accustomed to.