<img height="1" width="1" alt="" style="display:none;" src="//www.bizographics.com/collect/?pid=5286&amp;fmt=gif">

Virtually Speaking

The A.I. Buyer’s Checklist: Is Your Content Ready for an IVA?

Posted by Casey Gossard on Sep 23, 2016 12:58:15 PM


It’s one of the most common questions we get when talking to a new customer: is my content ready for an Intelligent Virtual Assistant (IVA)? It’s a smart question – savvy IT leaders know that an IVA is only as good as its training.

Since most businesses haven’t been capturing and organizing content, call logs and transcripts with the intention to train an IVA, they presume their content isn’t clean enough for IVA training. But you would be surprised how far we can get with what you already have. In fact, most companies have everything they need to create a robust natural language model for an IVA -- they just need to find it.

For most IVA implementations, we need 3 key pieces of information that you already have to get started.

1 – Quantitative information. This is typically a breakdown by volume of topics that customers ask about. We want to know what your customers ask you about. We can get this from content as simple as a list of common search terms, web page traffic, or more sophisticated analytics from your knowledge base.

2 – Qualitative information. This is all about determining how customers will ask their questions. Some companies may not track information in this way, so we’ve developed tools that use machine learning to comb through transcripts and call logs and figure it out.

3 – Business value opportunity. Finally, we need to find an area where an IVA can truly add value. Almost every enterprise has an area where IVAs can help, but we don’t “pass go” until we’ve identified a specific area where the IVA can truly save money, make money, improve the customer experience, or in some other way lift your business.

Most customers have the ability to compile this information, but they may not know exactly how to get started, or what their output should look like. Next IT has content analysts involved in every implementation for precisely this reason. Our content analysts use our tools to capture examples of the language that your customers use and map that against the most common topics and intents in your customer interactions. These tools are unique because they can work with unstructured data, eliminating much of the manual labor required in compiling and cleaning data.

But what if I have absolutely no content or customer data?

If that data doesn’t exist, we either set up listening posts to gather the data, or we design a phased rollout. In a phased rollout, we are able to start with the core domain expertise that our IVAs have developed over the past 15 years. That’s why experience matters with this technology -- you aren’t starting from a blank slate.

Relatively quickly, we expand upon the initial phase based on real user data captured from daily customer interactions. In fact, along with consistent nurturing over time, we find that the first few days after an IVA goes live will produce a very large and useful dataset for training and tuning purposes.

Whether your content is a mess or simply non-existent, you can still get started with an IVA. We have the people and the tools to make that possible. What’s more, our IVAs are designed to continuously learn and improve through a combination of machine learning, multiple algorithms, natural language processing, analysts and linguists. It’s an essential component of our solutions, and it’s at the heart of what truly separates intelligent virtual assistants from chatbots.

Even if you have pristine content to train an IVA, the fact remains that customer needs are perpetually in flux and product offerings constantly change. If your IVA isn’t designed to continuously improve, then the customer experience is going to suffer regardless of the content you started with.

An IVA is not only going to adapt to the changing nature of your business but also generate insights into your customers and business that you never had before. This is all made possible by the rich, dialogue based interactions our IVAs have with your customers – the data from those interactions is priceless.

Data for The Big Picture

Even though we know exactly what content and data we need to get started, we also urge our customers to understand the bigger point about IVAs: you can generate business value from this technology in multiple ways, for a very long period of time.

We encourage them to view IVAs less like a purchase and more like a long-term investment. Most of the technology that IT buys has a finite purpose and a predetermined upgrade path. IVAs grow and evolve with your business and your customers bit by bit, day by day.

Is your content ready for an IVA? It probably never will be, because your customer needs are changing by the minute. Instead of catching your content up to an IVA, you should focus on deploying your IVA as quickly as possible with the content you already have. Our tools make that possible.

Once we go live, you might just have your hands full keeping up with the insights that the IVA delivers to your business.

Topics: Business