Why natural language search is the way forward

By Jessica Vella, Associate Data Scientist – Advanced Analytics and Paul Sherlock, Associate Partner, Offering Lead – Cognitive Care A/NZ

Every week, hundreds of young couples take their first step on an exciting, scary, nerve-wracking journey – one of the most important of their lives. They decide to buy a house.

If you’ve been on this journey yourself, you know how daunting it can be. You may comb through mountains of research, loads of policies and complicated jargon. You may head to five different company’s websites, trawling through links and documents. And when you finally find a search bar, it may give you 7,000 results – none of which answer your question!

This experience is enough to scare off even the most committed homebuyers. The task is likely to be assigned to the too-hard basket. Does the experience have to be this painful?

A more natural way to search

As a data scientist at IBM and someone who has been through it, I would say no. It has to be easier, whether you are the prospective buyer or that buyer is your customer! Imagine a world where instead of getting thousands of irrelevant results, you could ask your question as if you were talking directly with an expert. You might ask, “What is your home loan comparison rate?” Wouldn’t you love a single line answer that told you exactly that? No more clicking 6 pages deep on a website to find your answer! No more trawling to the 10^th page of Google search results! Instead, you receive exactly the information you need.

The great news is that this can be the reality for websites with the power of natural language understanding, the branch of artificial intelligence dedicated to understanding humans. And the power can extend beyond websites too – with a “just ask” capability embedded in social messaging channels, mobile apps and even in tools used by staff!

Search engines need to catch up

Today you’ll find a search engine embedded in almost every website you come across. For many customers, this is the first way they try to get product questions answered. They try to self serve, to “Google it.”

However, for many organisations, these search engines have become part of the furniture – placed into position and then largely ignored. And so they return results with very little relevance, not answering the customer question at all. Few businesses take the time to optimise and improve their search engines, missing the chance to learn more about their customers.

The way customers search is changing

If they did, they might realise something important: The way customers interact with technology is changing. Not only are customers more likely to do things for themselves online, but people are getting used to interacting with computers in natural language. This is a result of the popularity of virtual assistants such as Siri, Alexa and Google Assistant. In fact, almost 70% of voice requests to Google Assistant are expressed in natural language. This is how customers want to interact and be served today.

Unfortunately, most of the search engines that businesses use on their websites aren’t built for these types of queries. They leave customers with no results or, at best, a link to irrelevant material. How frustrated would you be if the response to “My internet is down” ended up being an article about how “Our company is doubling down on network upgrades”?  These old-school search engines can serve up to 7,000 search results, or none at all! And this leads to a frustrated customer, who’s more likely to give up on the search and switch to a competitor with a smoother experience.

Create a better way to search
Given 70% of customers use self-service channels to find answers to their questions, businesses need to make it easier for customers to search for information they want, in language they actually use. Think about when you’re on your mobile phone, trying to find a quick answer and scrolling through search results on the tiny screen. Think about your customer, as they have a similar experience.

This experience could be replaced with a context-driven, natural language approach. Natural language search works by allowing interaction in the language that humans speak every day. Customers can ask, ‘How do I close my credit card account?’, instead of guessing keyword combinations like ‘card account close’. The artificial intelligence understands not just the keywords, but the intent of the customer. Instead of ranking results based on their relevance to the keywords in the query, the intended action is understood, and results are returned that help the action to be completed. For example, if we’re talking credit cards: the search can differentiate between wanting to close a card account, making a payment and simply wanting information about the monthly fees.

By developing the tools even further, the search can maintain context – understanding who customers are and what other enquiries they’ve made. It can give the customer a comprehensive and accurate answer after searching knowledge articles, policy documents, FAQs and other resources. It can return different results for you and for your mum, based on your specific requirements and personalisation.

Optimise outcomes for customers and businesses

Our IBM Services team is dedicated to creating elegant and differentiated experiences for your customers and your staff. We transform the customer experience journey by making self-service channels more readily available and more easily accessible. In the past year we have worked with Telstra, developing both customer and employee experiences, as well as with large insurers and federal government departments. While we at IBM proudly focus on the experience first and the technology later, we are also heavily invested in and believe in the power of machine learning in customer care. Our own tools have been named as leading the way in natural language understanding.

Our Cognitive Customer Care team works closely with customer service representatives to enable customers to interact with brands on their own terms and at their own leisure. It creates better search outcomes and improves customer satisfaction. It can also help reduce the cost and complexity of serving customers through assisted channels. 

Search is only the beginning of Cognitive Customer Care

Transforming legacy website search functions is a valuable first step and great entry into the Cognitive Customer Care space. A comprehensive natural language search can be stood up in a few hours, then continuing to learn and grow thanks to machine learning. But the tech isn’t the only thing that can learn and grow from advanced search – there is an opportunity for businesses to better understand what their customers are looking for by analysing the search terms.

For example, if your accounting firm gets 20 times more enquiries around tax time, you can use that information to prioritise resources. You can even publish knowledge that can proactively answer the most pressing questions for the next financial year. Value is delivered not only through the improvement of customer service through better result relevancy, but as an ongoing source of information direct from the customer. Knowing how your customers communicate can be an extraordinary asset – one that can fundamentally change the way in which you personalise their experience.

Gartner research states that over 40% of today’s live enquiries could be resolved in self-service channels. You can not only meet the customer when, where and how they want, but also reduce the cost and complexity of serving customers.

No longer will that young couple have to trawl through policy level banking documents with a fine tooth comb. No more fighting with complicated website navigation on a phone screen. No more abandoned journeys from disheartened searchers.

Want to discuss how IBMcan help you improve your search engine and make the most of your resources?
Let’s connect @ https://www.linkedin.com/in/paulsherlock/ and https://www.linkedin.com/in/jessica-vella
or start a natural language conversation with IBM Services – Cognitive Customer Care: cccare@au1.ibm.com

---