Published: March 04, 2022 | Comments
AI may seem like magic, but it’s not the end-all-be-all of software. In fact, many of the features and user benefits you strive for can be achieved without it.
AI is a technology that enables machines to simulate human behavior and create software that can solve complex problems as we do. Machine learning is a subset of AI which allows a system to automatically learn from past and present data without having to be reprogrammed. Famous examples work off hundreds of millions of individual parameters, and require huge computational power (and costs) to train the system. Narrow-focused AI applications may provide more bang for the buck, however.
I recently read about a contact center using AI to analyze data surrounding an unexplainable increase in customer-address change errors. Bots studied worker patterns and made intelligent recommendations to prevent errors and increase efficiency.
I’m glad the problem was solved, but without knowing more detail, this sounds like a relatively easy fix without AI. With a standard issue like changing an address, you can create a simple decision tree process to:
- Guide agents step-by-step through a standard process
- Ensure the proper steps are always taken in the right order, and by everyone
- Reduce or eliminate training time
Artificial intelligence, as it stands today, cannot replace humans, nor is it a magic bullet. It’s a force multiplier, meaning it’s a factor that gives your staff and organization the ability to accomplish more than they could without it.
AI is part of the overall service relay race. We already rely on an entire team of people and tools to deliver support. That includes things like IVRs, CRMs, ticketing, knowledge management, chatbots, and, of course, staff. In this model, AI is a means to an end - a method of implementing features, but by no means the only one or best one for every use case.
So How Will AI Help Agents?
In the context of chatbots and knowledge management, here are two specific ways AI will help:
This can include:
- Deriving information from behavior by analyzing and learning from how users interact with knowledge, and then using those findings to solve specific problems.
- Exploiting unstructured data by better understanding, organizing, and being able to deliver it when needed
- Creating internal bots to actively monitor and analyze conversations in real time, and coach agents on how to proceed or with tips after the fact or while they are on the phone
Better Search with Profiling
There are two types of search profiling. First, there’s implicit profiling, which learns from individual user behavior. This includes how the user searches (including wording and phrasing), what the user searches for, and what the user actually chooses and finds useful. Collaborative profiling, on the other hand, operates similarly, but uses groups of similar users to additionally tailor results to best meet the user’s needs. An example is Amazon’s recommendations of “Users who bought this also bought that.”
How AI Won’t Help Agents
AI can do many amazing feats, but sometimes simple tasks still elude it. These include:
Consider this phrase: “A man went to a restaurant and ordered a steak. He left a large tip.” You and I both know what the man ate, but most AI will struggle with this because it involves inferring something (rather obvious) from statements that don’t directly state the answer.
Learning and Adapting
Humans learn throughout their entire lives, while AI is trained on static data sets to learn to perform specific tasks or solve problems within a certain framework. Once deployed, AI has typically reached its peak, and is not learning from new information. The real world, however, is a constant stream of new information, meaning AI models can get quickly left behind. While continuous learning is currently a hot topic and under development, it’s simply not there yet, especially on a commercial level.
Understand Cause & Effect
Machine learning works on correlation, identifying patterns that even humans can’t. Humans learn by creating causal models, like if I touch a hot stove, I’ll get burned. For AI, that cause-and-effect mode of learning is beyond its ability.
In customer service, when the reason your internet doesn’t work could come from a variety or combination of causes, AI will be about as useful as a solar-powered flashlight. This is another case where step-by-step troubleshooting via a decision tree would be faster and more effective.
AI is an exciting new development that can absolutely deliver real value for customer service. However, it’s currently best in a supporting role and not directly customer-facing, as many disappointing experiences with chatbots have shown us. Let agents provide the human touch and common sense they do best, and let machines handle the other routine, mindless tasks.