Some edited excerpts:
How can generative AI like ChatGPT help businesses deliver better customer experiences?
Consumers want fast and helpful service, and in India, 85% of them expect to engage with someone immediately after contacting a brand. This is true for consumers across the world and globally, 67% customers predict generative AI will soon play a crucial role in customer service. This rises to 78% among those who have already used it. The conversational capabilities and easy accessibility of generative AI gives consumers answers quickly, with the right amount of detail and context when trained well.
Nearly 70% of customers across the world expect most companies to use generative AI to improve their experiences and they also believe companies that do are ahead of their competitors. This is because generative AI has the potential to improve self-service drastically by providing a conversational reply with information a customer needs instead of chatbots pushing out a list of results from a company’s FAQ page.
The biggest opportunity lies in using AI to eliminate much of the manual workload that can be low value and time consuming, thereby improving agent and admin productivity. Imagine an agent instantly getting an accurate summary of a customer’s interaction history and suggesting the ideal reply that allows them to solve customer issues quickly, saving so much time from reading pages of text or spending significant time looking for information.
It can also enhance the self-help functions and address knowledge gaps by creating missing support articles and reply templates boosting productivity.
ChatGPT is powerful, but we’re seeing companies getting caught up in the hype and often not being thoughtful in its application. ChatGPT’s strengths lie in finding general responses that are highly conversational and easy to understand, but it often lacks the context to answer questions specific to your business – this can be problematic, especially when customers want correct information. That’s why companies looking to implement generative AI in their CX ecosystem must ensure it’s trained with CX-specific data.
How can large language models help in detecting customer sentiment and intent and what are its uses for personalising customer experiences?
Poor customer service interactions can significantly impact customer retention and loyalty. In the APAC region, 71% of consumers said that a bad interaction with a business can actually ruin their day. And nearly half feel that brands don’t care about their wellbeing. Customers want to be understood. And when businesses don’t actively track customer sentiment and intent, it can be challenging to anticipate and fix issues before they can become bigger problems for customers.
Understanding customers’ emotions and offering the right solutions proactively, will lead to happier customers, and agents who are better prepared and more confident. That’s why at Zendesk, we created our own proprietary models, specifically trained with CX data to understand CX. Large language models like ChatGPT lay a strong foundation with their highly conversational capacity, and can make such a big difference when layered with our CX-trained models.
AI can identify at-risk customers (ones in danger of churning) by conducting customer sentiment analysis to gauge intent and tone. When paired with automated routing and AI-powered workflows, the insights gained can ensure the most experienced support agents handle the tougher interactions.
For example, when a customer raises an issue about a product, CX solutions that have a good understanding of what the customer wants and how the customer is feeling can automatically route the issue to the right customer service agent, especially if specialised knowledge is needed. These solutions can also provide a summary of the customer that includes purchase history, past interactions, and the existing issue, thereby offering more personalized assistance. This helps agents save time and also resolve issues quickly.
Can generative AI like ChatGPT be used to ensure AI-powered customer conversations are as smooth as human interactions? Can it handle complex customer questions?
Businesses that intend to adopt generative AI must keep in mind that AI can make mistakes. ChatGPT must be fed with accurate context and domain specific knowledge if it has to build the capacity to provide the customer with appropriate resolution. But generative AI is not always 100% accurate in problem-solving skills and empathy that only human agents can possess.
Companies should bear in mind that the majority of consumers want generative AI to improve interactions, not replace agents. In fact, 81% of consumers say having access to a human agent is critical in maintaining their trust with a business when they have trouble with AI customer service.
AI learns over time, and during the initial stages of training an AI model, it’s crucial to involve humans to ensure that the business is offering the best possible customer experience. This is where intent and sentiment detection can be of profound use – it can help decide which queries can be resolved fully automated and which should be escalated to a human agent immediately. Businesses should avoid blanket implementation of AI and automation.
Ultimately, it is up to the businesses to set the right threshold for how complex a question generative AI can handle before handing it over to an agent. Because there will inevitably be customer queries that require a human and AI doesn’t have the necessary data to handle it.
There is always the risk of technologies like ChatGPT being inaccurate in the responses it curates. What should businesses keep in mind while integrating ChatGPT into CX functions?
At Zendesk, we believe that AI will drive each and every customer touchpoint in the next five years. Even so, we are still at the initial stages and must stay grounded in the knowledge that LLMs today still have some limitations that may actually detract from a customer’s experience. To avoid this, companies must understand where generative AI is ready to shine and where it isn’t—yet.
While it’s important to move quickly, we must also be thoughtful in the way we implement LLMs like ChatGPT. One of the biggest challenges in adopting AI solutions is to build trust in the technology.
Businesses looking to integrate LLMs into their CX ecosystem should first consider the following aspects:
● Privacy and security: It’s imperative to keep customers’ data secure, especially since 93% of Indian customers are concerned about data privacy. While choosing the right solution, businesses must identify if the solution is built with the best security and compliance standards, and if there are checks and balances to evaluate data usage to ensure it is used responsibly.
● Transparency: Not all AI predictions are equal in terms of quality. Sometimes AI models have high confidence while performing a task, while in other scenarios the confidence is low. That’s why it is important to know whether the CX solutions provide information on the quality of AI predictions so they can know exactly what they are working with and act accordingly.
● Accuracy: While ChatGPT is extremely conversational, it doesn’t always provide accurate information. Businesses need to ensure that the AI models have been properly fine tuned to understand the business context and those models should have boundaries in place to avoid inaccurate replies. This does mean some use cases that are still risky today should be approached with much care, and in other cases, passing on the information provided by confidence scores to administrators or agents, so workflows can take the level of accuracy into account.
● Human oversight: While working with new AI models, it’s always better to have a balanced human-automation strategy and ensure humans are kept in the loop for a better experience for both the agent and customer.
What steps can businesses take so that the customer experience won’t be impacted when implementing large language models into customer service?
AI as a whole is a game-changer when it comes to transforming businesses, but merely implementing LLMs like ChatGPT for customer conversations does not always result in positive outcomes. Businesses need to identify areas in which the solution can cause a lot of pain to customers, take the right measures to safeguard and test the integration and ensure that the LLMs are trustworthy. This can mitigate issues that may negatively impact CX.
Before investing in generative AI, it is crucial to have a solid human-automation strategy in place. It’s important that humans guide AI on what is true and false, and feed the right information about the business. While access to AI must be democratised, it should also be customisable to scale based on the company’s requirement. It is also important to know that ChatGPT makes mistakes and while integrating the API into CX solutions to make AI sound more human-like, it is important to ensure the solution is trustworthy.
Customers care about how efficient businesses are in solving their problems. And in India, 81% of customers will forgive a company for its mistake after receiving excellent customer service. With the right CX strategies, LLMs can help provide accurate responses and know when to escalate to an agent, especially when combined with intent and sentiment analysis, thereby resulting in positive growth and a seamless, efficient customer experience.