The Risks Posed by Biases in Voice and AI
Artificial intelligence has been integrated into numerous areas of business, including effective sales management, and customer interaction. There are still some risks associated with the use of AI as a new and evolving concept, but the many benefits of AI systems retain their worth for customer profiling and analysis.
AI can be used to detect speech characteristics, such as tone of voice, mood, and attitude. When customers call a sales team, AI can analyze their speaking styles and connect them with the appropriate call handlers. For example, an individual identified as friendly and talkative from previous interactions can be transferred to a phone operator who is most successful with these types of customers. This also works on the opposite end of the spectrum, with someone who has a history of negative interactions on the customer service line being connected to a customer services agent with more experience handling difficult callers. When customers are transferred to call handlers who better understand their needs and the styles of interaction that create the best connections, the company is more likely to make a successful sale or create a positive customer experience. Forbes estimates that voice AI will be handling 20 percent of all customer service queries by 2022.
Voice recognition AI typically involves conversational AI, natural language processing (NLP),, and automatic speech recognition (ASR). Combining these elements results in the effective analysis of records, the enablement of machine learning systems, and the breakdown of speech characteristics for a better understanding of customer needs. AI will use machine learning (using algorithms that can learn from recorded history and improve its function) to ensure that data collected by NLP and ASR is used to its full potential.
How is AI used for sales enablement?
Sales enablement refers to the multitude of techniques to increase sales and revenue. This includes both technological and human-based resources, coaching and training, and your customer-facing content. AI is enhancing various aspects of sales enablement, particularly through automation and machine learning, which improves the detection and management of customer details and customer profiling. Through the analysis and recording of customer interactions, AI systems can provide businesses with more comprehensive information concerning their customers' attitudes toward their brands, services, and advertising. They can also measure how positively or negatively customers are reacting to sales-based calls, providing the opportunity to adapt current approaches as necessary to ensure that engagement is maintained.
But, as with any new and developing system, it is important to be aware of the potential risks of using AI profiling. This method of customer analysis and monitoring is certainly not flawless, and the technology is still evolving and updating at a rapid speed. Businesses are much more likely to incur issues with this technology if customer data is not correctly updated. If, for example, there are multiple users of a single device and this is not included in business records, profiles might be compiled based on more than one person. This results in erroneous data being recorded and used by the AI systems.
There have also been various issues detected concerning attempts to link physiological traits to emotions. In these cases, results have shown cultural biases. Racial profiling has also been detected as an issue in new attempts to use voice recognition to determine physical features, such as height, weight, gender, race, or health.>
Speech recognition functions that coordinate with AI systems to record and analyze data also have flaws. The software used for speech recognition has been shown in a recent study by Harvard Business Review to have significant race and gender biases, performing more inaccurately for women and non-white individuals. These biases are not intentionally included in the coding of software, but they are nevertheless problematic. The difficulties that AI faces when recognizing and identifying the nuances of diverse vocal features mean that incorrect or skewed data can be recorded, resulting, in incorrect profiling.
It is also important to be aware of consent issues when recording phone calls for AI purposes. There are many legal angles to consider, but in the United Kingdom and most U.S. states, there needs to be one-party consent in all recordings, meaning that at least one person needs to be aware the call is being recorded.
While the rapid uptake in voice recording for businesses might seem intimidating, it is a change that aims to improve the quality of interaction for both businesses and customers. Businesses will be better equipped in adapting to customer needs, and customers will receive better-targeted advertising and contact. However, the increased use of voice recording means businesses will need to remain alert to the legalities surrounding consent to record. They must ensure employees have explicitly consented to the use of recording software. Usually, this will involve training to help employees understand their roles and signed forms to clearly confirm their consent. Customers must be alerted to being recorded, and the reasons for recording should be outlined. This can be conveyed through an automated message prior to the call being connected.
Nigel Cannings is founder and chief technology officer of Intelligent Voice, a developer of proactive compliance and technology solutions for various forms of media.