Voicebots are software systems which use AI, text to speech and
automated speech recognition to understand human voice utterances and respond
in a human-like manner to a wide variety of requests. They are typically
configured to integrate with back-end systems thereby increasing their range of
skills and eradicating or reducing human intervention.
Many organisations are heavily investing in digital transformation initiatives in 2020 and voicebots can deliver a great customer experience 24 hours a day for a fraction of the cost of a team of agents. AI technology has now matured and become easier to use so it is possible to create a bot and monitor it without needing an army of experienced software developers. With this in mind, we have listed below the main considerations that will make your implementation a success.
Work With The Right AI Platform Provider
A new voicebot, like a chatbot, cannot simply start answering questions. You need to load it with the data it needs to ask and answer questions about your business and then its needs to be trained to be able to use that data to respond to your customer’s questions in a natural, conversational way. This is where the natural language platform is crucial. Natural language processing platforms use AI and machine learning to model questions and answers and the different ways they are asked to work out intent. When you upload your data, the platform will analyse based on its own models and give suggestions on how questions and answers may be asked. The more data sets the platform provider has access to, the more accurately they will be able to work out intent, making it imperative to work with a provider who has the appropriate scale – For example, Google AI.
Choose The Best Speech Systems For Your Voicebot
A voicebot requires automatic speech recognition (ASR) and text to speech systems (TTS) to understand and reply respectively to spoken utterances. You should not compromise on these systems because the quality of the voicebot heavily depends on them but there are a wide variety of systems available with varying levels of quality. Your service users do not want to continually repeat themselves or listen to an annoying robotic voice. Choosing a low quality voicebot can result in quick failure and possible damage to your brand.
Monitor Continually, Real Time if Possible
Your voicebot interactions will generate vast amounts of utterance text, which is good news because text data is fast, easy and cheap to analyse. This data can be presented on a web application console, real time, so that you can see voicebot interactions as they happen. Analysis can be performed to distil the text utterances into colour coded sentiment flags or keyword markers giving a supervisor immediate, real time indicators on customer experience (CX). Real time CX analysis gives you instant client sentiment without having to bother your clients with feedback forms or surveys, if they are happy or annoyed it will be detected.
Monitoring in this way is especially important in the early stages of your voicebot implementation to understand how your customers are feeling about interacting with your bot and how your bot is performing.
Agent Handover
It is important to give the service user the option to transfer to a person, especially in the early days when your clients are less familiar with your bot. A voicebot can detect when it needs to handover to a human either from the service user's utterances, too many default answers (eg "I'm sorry I didn't understand") or from sentiment. It can also ask the service user if they want to be transferred.
Choose The Right Skills
A voicebot can handle up to 85% of general customer service requests and some are better suited to a voicebot, such as appointment or meeting time management, PCI DSS compliant payments and requests which require strict GDPR compliance. These can all be performed 24 X 7 by the voicebot which will actually improve the customer experience in the process or make it much more secure. As a contrast, configuring the voicebot to answer questions where the answer can be found as quickly and conveniently through a search is not as impactful.
Don't Forget Outbound
Voicebots can be used as an outbound calling agent and work very effectively in conjunction with an automated dialler. The dialler instantly connects the voicebot once the called party answers their phone and can perform a range of suitable tasks such as debt collection, appointment reminders, important notifications, surveys sales lead generation. Voicebots can be used to pre-qualify sales calls to reduce wasted agent time. The voicebot can ask a series of simple qualifying questions without tying up an agent.
Security
Whilst voicebots process sensitive personal data in a much more secure way than chatbots or humans they can still be hacked or exploited if the backend systems are not secure. The voicebot supplier should be using vetted developers, ideally full-time employees and should have demonstrable security processes in place such as PCI DSS or ISO27001. If any personal data is to be stored it needs to be encrypted and behind a firewall, Amazon AWS provides suitable storage in the cloud.
Crawl, Walk, Run..
By allowing your voicebot to perform a few tasks really well rather than numerous tasks badly it will gather momentum and reduce the time required to manage it. Monitoring will allow you to pick up and deal with user ability issues quickly. Once you feel it has nailed a particular set of tasks open up the range for your (now familiar) service users. This strategy will also keep the costs down and reduce risk.
Conclusion
Millions of businesses around the world already use interactive voice response (IVR) systems and queues for handling callers, but they ultimately just cause delay, provide a poor experience to get information and eventually route to a person. Voicebots offer a much better alternative with no queues and the ability to get much more information, immediately and more naturally compared to an IVR, whatever the time of day. We are moving into an era where CX is critical to success and AI bots can help you make necessary improvements and deliver vast amount of useful data.
