In the near future, AI assistants may become the norm. Despite the impressive technological advancements that have taken place recently, though, AI tools still have their limits. Insurance processes, in particular, sometimes require a human touch. Insurance companies are eager to embrace AI tools, but to ensure policyholder satisfaction, every AI agent should have a human counterpart.
The Limitations of Chatbots
AI agents (evolved chatbots) are a convenient way to provide 24/7 support, but they can also lead to customer frustration.
Gartner surveyed customers between December 2022 and February 2023 and found that:
- Only 8% had used a chatbot during their most recent customer service interaction.
- Only 25% of those who reported using a chatbot said they would use one again in the future.
The effectiveness of chatbots depended on the task at hand. They resolved 58% of return and cancellation issues but only 43% of account information requests and 40% of payment or transaction requests. When customers had a problem, chatbots were even less effective – resolving only 25% of complaints and 17% of billing disputes.
AI Is Improving, But It’s Still Limited
In November 2022, ChatGPT launched and AI tools took a huge leap forward. Although the AI-powered, large language model chatbot was publicly available during the Gartner survey, there’s a good chance that most companies had not yet adopted it. That has been changing. According to Gizmodo, companies already using ChatGPT include Slack, Coca-Cola, and Instacart.
This next generation of highly intuitive and responsive chatbots are often called AI agents. They’ve advanced dramatically and some people have a hard time discerning the difference between humans and AI agents. Nevertheless, there are still limits.
In fact, when asked “What are ChatGPT’s limits in customer service?” ChatGPT highlighted eight potential issues:
- A lack of specific knowledge
- Inability to perform actions
- Contextual understanding
- Sensitive information
- Bias and inappropriate content
- Ambiguous queries
- Limited problem-solving skills
- No emotional understanding
These limitations could spell trouble for insurers that want to rely on an AI agent for customer service. Although AI agents should be able to resolve clear questions and requests that require an answer they already know – such as when an office opens or where to find policy information – chatbots may struggle with more complex and nuanced questions.
The Key to Chatbot Success
A chatbot or automated phone system that is able to resolve simple customer requests quickly can be the ideal channel for customer support. However, when AI programs cannot solve an issue immediately, customers may become frustrated if they can’t easily switch to a live agent. Therefore, the key to chatbot success lies in providing easy access to human representatives.
Imagine the following scenarios.
- A policyholder has just experienced a car crash or other traumatic event and is trying to report a claim. The policyholder is upset and not speaking clearly. The chatbot does not understand what the policyholder needs, which results in the policyholder becoming increasingly frustrated.
- A prospect is interested in coverage but doesn’t know much about insurance or what he needs. His question is vague, leading to a chatbot reply that isn’t informative. He ultimately buys coverage from an insurance company that has a human representative.
When to Involve Human Representatives
Adding chatbots to your customer service can provide customers with a quick way to ask questions while reducing your company’s administrative burden, but this strategy can backfire if you’re not following best practices about when to involve human representatives.
- Have human representatives available 24/7. In some businesses, it may be appropriate to have human representatives available Monday to Friday, 9 a.m. to 5 p.m., and let chatbots take care of needs outside business hours. Insurance is different because claims are often urgent and can occur any time. For this reason, you should always have human representatives available.
- Make it easy to request a human representative. Your customers shouldn’t have to jump through hoops to speak to a human. Make the option readily available throughout the interaction.
- Train your chatbot to suggest a human representative at the first sign of trouble. If the chatbot can’t answer a question or complete a request or if the customer shows signs of being frustrated (such as by repeating the same question), the chatbot should immediately offer to transfer the customer to a live agent.
- Consider outsourcing your customer service. You don’t have to do everything in house. Outsourcing some or all of your customer service is a practical and affordable way to make sure your customers always have the option of speaking to a human.
Need support for your AI processes?
Teleconnected can be your AI agent’s human partner. Learn more.