When did you last get to speak to a human customer service agent at the first ring? Whether it is your bank, travel agency, hospital or appliance company, you are subjected to automated voices, or chatbots that are at times so frustrating.

Chatbots are nothing but AI programmes. The idea is to use them to answer in the normal conversational way when the customers raise frequently asked questions. It avoids directing people to information like Google does. Several companies like Uber, Pizza Express and some airlines have deployed chatbots effectively to handle customer queries and order booking.

The global chatbot market is slated to grow 8-fold in the next decade to $43 billion and Indian market is already growing at 24 per cent y-o-y. Clearly, companies see them as potentially more useful in addressing customer issues and cost-efficient than automated telephone systems. But can they really replace humans in terms of handing specific and difficult queries? More importantly, will they have empathy, one of the key parameters to excel in customer service?

Use of AI offers some tangible benefits such as lower costs, scale and solving mundane issues. It will also eliminate many trivial issues easily. But imagining AI to take care of customer service in entirety is akin to wishing away your worries to the Almighty. AI algorithms will have a tough journey in terms of languages and its nuances to understand and then suggest solutions. Translation and recognition will be another challenge.

Diversity issue

The complications of AI-based chatbots will be even more in India where the diversity in culture, language and dialects is huge. The implementation pf AI cannot be a one-size-fits-all model. It will face unique challenges and may result in worst-case scenarios that impact customer satisfaction and business reputation. What could be some of the scenarios?

First is potentially the failure to understand local languages and dialects. India’s diversity of languages and dialects poses a big problem. AI systems will struggle to understand and process this linguistic diversity, leading to miscommunication and frustration among customers. Airtel had an AI-driven chatbot to handle customer queries. However, it struggled with the variety of Indian English and Hindi accents, delivering incorrect responses or failing to understand the query entirely. Obviously customers were irritated, resulting in a surge of complaints. Airtel must not have invested more into it, which perhaps explains why their chatbot’s linguistic capability is not visible any more.

Second, AI algorithms have a limitation to grasp cultural nuances and sensitivities, which can result in responses that seem inappropriate or offensive to local customers. During a major sale scheme, Flipkart’s chatbot was programmed to upsell products aggressively, but failed to recognise cultural sensitivities around pushy sales tactics, particularly in conservative middle class societies. The company saw this approach backfiring because it caused dissatisfaction among customers who felt pressured and uncomfortable. There was also a lot of social media backlash.

A third case is tech glitches during high demand periods. As with most things technological, AI systems can experience technical failures, especially during peak periods such as festivals or sales events, leading to significant service disruptions. During one of their Diwali sale, Amazon India’s AI customer service system experienced a massive outage due to overwhelming traffic. The system could not handle the surge in queries, and resulted in prolonged downtimes and delayed responses. Many customers were left without assistance during the crucial shopping period, and this led to widespread dissatisfaction and negative reviews, not to mention customer attrition.

Fourth is data privacy and security concerns. India may not as yet has stringent data privacy regulations, but a breach can lead to severe consequences for businesses in the future. AI systems that collect and process customer data can sometimes be vulnerable to security breaches, and will cause trust issues. One of the leading banks had faced a data breach where the AI system handling customer queries was compromised. Sensitive customer information was leaked, raising significant privacy concerns and eroding customer trust.

Fifth, an over-reliance on automation can lead to poor human support. In an attempt to cut costs, some companies over-rely on AI, reducing their human support staff to a minimum. Disaster strikes when AI fails to resolve issues and there are no human agents to take over to resolve them. Vodafone India reduced its customer support staff, and relied heavily on AI to handle queries. When the AI system failed to resolve complex billing issues, customers found it difficult to reach human agents for assistance. Result: Prolonged resolution times and widespread customer frustration, ultimately damaging the already-damaged reputation further.

Inadequate training data that leads to poor AI performance is the sixth scenario. AI requires extensive training data to function effectively. Our customer service scenarios are highly varied or inadequate, and this biased training data will be problematic. ICICI Bank’s AI chatbot had faced issues due to this. The bot failed to handle specific queries related to local banking practices and frustrated customers.

A last scenario is ineffective handling of emotional and sensitive issues. AI lacks the empathy required to handle emotionally charged or sensitive customer interactions. Tata-Sky’s AI system failed to deal with sensitive issues like service disconnections and billing errors with empathy or understanding expected by such customers.

Some suggestions

(i) Instead of blindly joining the AI bandwagon, consider the following:

(ii) Train the AI systems to understand local languages, dialects, and cultural nuances to communicate effectively with customers.

(iii) Implement a hybrid approach of AI efficiency with human empathy so that complex issues are moved to human agents quickly.

(iv) Invest in strong data protection measures to safeguard customer information.

(v) Ensure your AI is scalable and reliable, especially during high-demand periods.

(vi) Regularly update and improve AI systems based on customer feedback and evolving market needs.

The writer is a Fortune-500 advisor, start-up investor and co-founder of the non-profit Medici Institute for Innovation