Steps to Implementing an Intelligent Virtual Agent

The previous articles in this series gave an overview of what an Intelligent Virtual Agent (IVA) is and what one can do and outlined the common barriers to deploying an IVA.  This third and final article will define the steps a company needs to go through as it starts to implement this cutting edge technology.

The first step is to discover and define the business problems the IVA will be solving.  These may be defined already, but most companies lack a deep understanding of an IVA’s exceptional capabilities and thus would benefit from a several-week consultation by a leading IVA company.  Under a non-disclosure agreement, their experts can look at all of your current operations and identify which would benefit from the IVA solution they offer.

Once the exact scope is defined and the agreement is finalized, the next steps are to determine how to best structure the IVA – from behavioral, to integration, to knowledge – so it will deliver the desired experience to the end user.  Since each IVA can possess a “personality,” it is important to spend some time on this aspect, as your IVA will become part of your brand just as Alex is part of United Airlines’ brand. 

The next step is to design how your IVA will interact with your current website, back end systems and current contact center.  Since the IVA can “understand” natural language (be it speech or typed words), and it can understand user intent, you may determine that some interactions must be immediately escalated to a live agent.  Nowhere is this more important than in the medical arena. 

But, since the Intelligent Virtual Agent can handle a vast majority of your inquiries, I would suggest you give it leeway to obtain the maximum amount of benefit. 

The next step is to build a test model.  This allows the experts at your company to go through a few iterations to ensure the IVA meets your needs and expectations.  It also ensures that it is performing accurately and is providing the exact same response that your best live agents would provide to your customer.  Along with this step, reporting metrics are defined and built within the IVA.  Remember that an IVA is literally the biggest customer listening device ever devised.  You will benefit greatly from its ability to capture vast amounts of data concerning your clients and turn it into useful, actionable information that can be used to continuously improve your contact-center operations.

Once delivered, there is a continuous cycle of improvement.  Evaluation of the IVA chat sessions will be done by your quality improvement experts.  Analysis of the early interactions results in improvements, additions and tuning of the language model to deliver a better user experience.  Reporting can be virtually instantaneous.  An example for an airline-industry early adopter comes to mind.

When the influenza pandemic struck a few years ago, numerous changes needed to be implemented to the airlines’ IVA literally overnight in order to answer an avalanche of questions flooding the call center from people concerned about traveling during this health crisis.  The exact nature of these inquiries was captured by the system, and scripted answers were provided to the IVA, so the flood quickly subsided and the live agents could go on to handle more complex and financially rewarding services.

Finally, strategic enhancement must be mentioned.  Implementing an IVA is like buying a new tool.  You quickly wonder how you lived without it in the past and now find new uses that you could not have even imagined earlier.

For most companies that have purchased an IVA, its availability will rapidly lead to proposals for additions to the domain of knowledge the IVA can access, redefining of additional areas where the IVA can be deployed, and continuous improvement of the personalization, integration and ultimate user experience. 

Senior management will rapidly look the IVA system as a critical measure of the vital signs of a health care organization and become reliant on its data to make management decisions.