Due to the variability of human nature, companies are always looking to reduce the potential of human error occurring. Errors are like blocks: they can cause a whole system to momentarily stop working efficiently or inadvertently fail. It is no surprise that lowering the chances of committing any type of error and improving efficiency is of paramount importance in all sectors.

In this article we will explore how not only humans, but also AI systems, can be flawed and how together they both build up a solid system in preventing and lowering the chances of error in customer service.

Human errors in Customer Service

In regards to customer experience, human errors can be caused by different factors that are based on the variability of human nature. For example, each agent will react differently to stress, fatigue, and repetitive tasks. Their reaction would also depend on time restrictions, unfamiliarity with the task, inexperience, information overload, and in certain cases over manning, meaning that the task to be completed needs the help of more agents.

According to NOPSEMA (National Offshore Petroleum Safety and Environmental Management Authority), these can be classified as three different types of errors:

  1. Skill-based
  2. Rule-based
  3. Knowledge-based errors

Skill-based errors are based on the inappropriate execution of a task. These errors usually occur when tasks are repetitive, and due to an overload of work and fatigue, the agent’s attention span is lacking, resulting in the failure to implement an action.

Rule-based and knowledge-based errors can be attributed to the inability of an agent to adhere to standardised practices and protocols.

A cognitive overload of information also influences the capability to perform the tasks correctly in a timely manner.

According to the International Director’s website, 61% of human errors are skill-based and are actually easier to detect than other errors due to the nature of the error.

Design AI to avoid errors

The proliferation of AI into mainstream adoption and application, has seen the usage of such technology picked up by customer service departments, to aid in the reduction of errors and to improve overall agent performance.

However, it must be taken into consideration that, when it comes to AI, the main issues are caused by system design. AI is developed through data which is generated by humans. The inability to determine where such errors were made causes the AI to carry on executing the same mistakes in any future operation.

This action causes a phenomenon known in computer science as GIGO, (garbage in, garbage out), Human Garbage In, Artificial Garbage Out: flawed input data will only produce flawed/garbage output data, although in this case, it would be human garbage in, artificial garbage out.  According to Nick Ismail, 20% of time spent developing AI is used to perform tests to fix possible issues that the AI might cause in the performance of tasks.

But what differentiates human errors from AI errors? Predictability. AI errors are more predictable than human errors.

AI-design is very important and it is up to humans to verify the correct functioning of the AI. One thing which we discussed in the first article of this series on AI, is that AI has no Fluid Intelligence, meaning it has no ability to make abstract decisions (yet). It provides important data, however, it is up to humans to make the right decisions also based on abstract thinking.

This proves how AI and humans can cooperate to succeed: AI and humans can help to cancel each other’s mistakes.

How to help each other

Since skill-based errors can be caused by an overload of work which is likely to increase fatigue and distraction, AI in this case can easily help with repetitive tasks, giving more time to the agents to focus their problem-solving skills on tasks which AI cannot perform. One advantage of using AI in customer service is that it actually helps in reducing all types of errors and not just skill-based errors.

AI can also reduce rule-based errors thanks to a rule-based system: human-crafted rules are followed in the correct order, performed within a specific time and agents are freed from the stress to remember all the steps. For example, when agents receive many inquiries, AI is able to individuate which agent should receive the task, reducing the additional cognitive overload of information that would be required for agents to understand, assess and designate who should perform a particular task, and the time spent on assigning it to the right person. In addition, AI is always collecting important data which can help humans to understand which actions might be causing issues and which strategy would fix it.

Conclusion

In conclusion, human errors can certainly be reduced through the use of AI, and AI errors can only be detected and adjusted by humans. Therefore, in order to succeed, both humans and AI must coexist and help each other to reduce the probability of errors. It is an equilibrium between the two which can bring companies to be more efficient and carry out excellent customer service.