By Neelesh Kripalani, Chief Technology Officer, Clover Infotech
Customer experience is one of the biggest costs for business. It is also one of the most important levers of the business. A good customer service experience is likely to trigger positive ‘word-of-mouth’. On the flip side, an unpleasant customer experience can lead a brand to lose multiple customers at a time.
Businesses are aware of this and have been keen to introduce technology to empower the customer experience. The creation of chatbots has been a first step in this direction. Chatbots are a self-service system that can help customers to address mundane, redundant, and repetitive queries on their own. It helps businesses to reduce the number of human agents needed to manage their contact center.
So, what’s the next step?
Businesses are keen on building speech-enabled Interactive Voice Response systems. IVRs, as they are popularly known, will make a customer feel exactly like she/he is talking to a human agent. It can also be empowered with switchover capabilities to transfer the call to a human agent if required.
But Speech and AI-powered IVR may not be easy to develop. More importantly, it requires technology and skill to manage, host, and maintain IVR. However, with cloud gaining prominence, several service providers have created cloud versions or migrated their speech recognition technology, and natural language processing (NLP) capabilities to the cloud. This has marked a tectonic shift in consumption of self-service AI and NLP Powered IVRs. It is now become a SaaS based offering and businesses can now just pay and consume as much as they require rather than having to buy or build such technology and manage it in-house.
Another advantage of these Cloud-hosted IVRs is that they are exposed to a huge number of instances and scenarios which helps it to get trained and learn faster. They are trained to not only understand the words of the end-customer but also comprehend the intent and respond accordingly. With the evolution of seamless and fluid minimal code frameworks and platforms, even the smallest of businesses are deploying virtual agents and creating a very seamless customer experience.
When the capabilities of a chatbot, speech recognition powered IVR and AI/ML comes together, it creates a very robust Intelligent Virtual Assistant (IVA)
Is there a blueprint to develop an IVA?
While it might depend on your business and your objectives, it would be helpful to have a blueprint and consider the following while creating an IVA: –
- Its level of skill and expertise:
An Intelligent Virtual Assistant (IVA) could have very basic level of expertise or could have expert level of skills. For example, a basic IVA may just answer the phone and direct the caller to options as per his query or just place the caller in the queue. An advanced IVA can however understand the user’s language and intent and give a more meaningful response. It can also process a payment or successfully record a product warranty/ replacement request etc.
- Omni-channel capabilities:
Customers of today are used to interacting with brands across multiple touchpoints. They can call the brand directly or go through the messaging route with SMS, WhatsApp, or even chatbots that pop up as soon as a customer visits their website. IVAs can be built to ensure maximum coverage. The business can first understand the most popular channels and the frequency of interaction across each channel. Then, they can intelligently understand where to deploy the IVA to ensure a seamless and consistent omni-channel experience for their end-customers.
- Analyse and Automate Tasks:
The general rule with automation is that businesses must find rule-based and redundant tasks and train IVA systems to automate such processes or tasks. These processes or tasks could be categorized as most frequently asked queries, most impactful etc. to gauge which one is best suited from an IVA perspective. It will also be prudent to check which one can create the most positive impact with respect to freeing up the bandwidth of the engineers and employees of the company.
- The best Speech Recognition Technology:
The IVA must be powered with very relevant speech recognition tech. With the hyperscalars such as Google, Amazon etc. on the cloud, you could receive ready templates with different accents and tones. However, depending on its customer interaction and engagement objectives, the business must choose the most relevant technology. It is always better to do this irrespective of what the offering, or the invisible voice behind the IVA. The business must carefully evaluate the options and select the one that is best suited with respect to politeness and keenness to address the customer’s issue.
- Strong integration with Organization’s Application and Technology Layer
An IVA must seamlessly integrate and exchange data with the organization’s customer relationship systems (CRM) and enterprise applications. An advanced IVA can also work on pre-configured customer journeys and ensure that the apt messaging reaches the right customer.
- Train the AI System:
The biggest factor impacting the success of an IVA is its ability to comprehend customer’s intent. The only way for the IVA to get this is to make it smarter by getting it accustomed to multiple instances so that it can draw patterns and understand the intent correctly almost all the time. An IVA can initially be trained using data gathered from human agent interactions. The power of AI and ML can then be leveraged to ensure that it can understand speech and intent well and ensure an apt response accordingly.
- Success and Improvement Metrics:
IVAs can record interaction data and bucket it as per query, response, intent understanding etc. This enables a business to understand the performance of IVAs and gauge its success rate and areas of improvements. These inputs are critical from a business and customer understanding viewpoint. However, they can provide ample fodder to train the IVAs and make them smarter. The idea is to make the IVA smarter with every interaction and a periodic review of measurement metrics and course correction plan can help in attaining this objective