The Art and Science of Automated Speech Recogition
June 28, 2012
June 28, 2012
We have all heard the commercials, ‘If a company cares about their customers, why do they force them to ‘talk’ to a machine’? or ‘Call us and speak to a live person because we care’. It’s a great marketing pitch but do you really need to speak to someone to activate a prepaid credit card or check the status of your order? In many cases the answer is no. These common tasks, when handled by agents, greatly undermine their ability to service more pressing issues from other callers and diminish their capability and efficiency. In short, transactions like these can be handled faster in an automated manner as opposed to consulting an agent anyway.
Contrary to the popular belief that automated systems act as a barrier between caller and an agent, these serve as a bridge between an anxious caller and the appropriate agent. A well designed system will help clients achieve many goals including but not limited to increased customer satisfaction, reduced call duration, efficient call routing, lower agent burn out and above all higher customer loyalty. This brings us to the million dollar question, what does it take to build a system that will achieve all the above goals? How can one enhance customer interaction with a system that sounds natural while benefiting from the efficiencies of an automated design? The answer is a properly designed speech recognition solution.
Automated speech recognition systems have made significant strides in the past few years and with the introduction of cloud computing, they are evolving into a major force to reckon with. A speech solution is not an panacea however, when designed and implemented with right set of principles over a well-defined time line, it can truly yield significant benefits. It all begins with asking the right question(s), “Is speech really necessary?”, “What is the ultimate goal?”, “What are the key metrics?”, “What is the time frame?”, last but not the least “How do we define success?”. There are five major steps involved in designing a good speech solution (a) Business analysis – a detailed study of the existing system or a requirement that starts with identifying the top call drivers, marking tasks for automation and evaluating call patterns (b) Dialog design – creating a framework for the caller interaction (c) Grammar design – creating a ‘dictionary’ that will allow the machine to properly interpret the words or phrases spoken by the caller (d) Application design – creating a flow detailing the caller experience (e) Launch – determining how the system should be introduced to the end users. In the case of speech systems, the post launch period is one of the critical phases too. This involves call recording, flow analysis and fine tuning. it is an ongoing process and imperative to keep the system relevant and refine the solution to achieve maximum results.
A well implemented speech solution can provide a seamless experience for the caller and ease the transition from an automated system to an agent, where necessary. Remember, a call starts when the phone is answered and ends when the customer is served. Everything in between must be understood in order to provide the best customer experience. Speech is but one of the tools at our disposal to build that relationship with the customer and provide excellence in customer service.