Natural language processing, or NLP, is a service which can understand the way we communicate in writing, speaking, and even in abstract forms such as art — and then derive insights from it. The business applications of NLP are much wider than simple chatbots (though those have risen to understandable prominence in recent years).
In this article, we aim to provide technology decision makers with a comprehensive guide to choosing natural language processing applications for their business.
Common NLP business applications
NLP helps businesses to make sense of the massive amounts of data they have at their fingertips, and make data-driven recommendations on it, far beyond what a human analysis could do.
Common forms include:
- chat bots – which are customer-facing text-based responders
- call center bots – interacting in a voice channel, also customer-facing
- same bot/assistant – used in internal helpdesks for resetting passwords, etc.
Other applications are found in large companies which need to scan massive quantities of documents to ingest draw conclusions from.
“NLP is still in a very early stage of maturity in the business world, so we tend to see it mostly in areas where the process workflows aren’t that complex,” says Greg Johnsen, CEO, LifeLink.
They’re also used to gauge customer sentiment during a new marketing campaign or product line roll out.
“It can be responding accurately to inquiries, finding patterns in documents, or gauging the sentiment of language,” says Stephen Blum, Founder & CTO at PubNub.
“Sentiment analysis is valuable to determine the emotion of language, whether positive, neutral, or negative, so NLP and sentiment analysis is immensely valuable to brands, especially those in the public eye.”
By hooking sentiment analysis into social media or other online forums, businesses can analyze how consumers are feeling about their brand at massive scale. This includes being intelligently able to process tweets, Facebook mentions, forum entries, and gauge the overall sentiment towards their brand overall, product launches, and news.
Changes during the pandemic
While we’re all a bit numb to learning about the COVID-19 pandemic’s broad effects on the business world, it bears mentioning here.
Johnsen says the big change has been on the adoption front, which has been driven by the pandemic.
“The urgency of the moment forced a major shift across the business IT landscape as they raced to find better, safer ways to interact with millions of customers remotely. In healthcare, conversational AI was still largely on the fringes in January 2020. By June, thousands of hospitals were using chatbots to screen millions of people for COVID-19 symptoms because it was the only way to handle the huge spikes in consumer service demand.”
By the summer, several large hospitals had implemented digital assistants to virtualize the pre-visit intake and waiting room experience for patients so they could reopen safely and get back to providing routine care.
“Now the healthcare sector is expanding this capability for the long run, as part of their long-term strategy. By utilizing innovative chatbot solutions that can harness the combined power of NLP and workflow automation, this will ultimately be greater for patients, clinical accuracy and the bottom line. Conversational technology is here to stay.”
The education industry is another one that has recently seen a massive shift to online equivalents. Remote and online learning is more important than ever, and natural language processing apps can help digital learning rise to a new level, expanding the reach of a learning course to an international audience.
Choosing NLP: IT department challenges
Quinn Agen, Vice President of business development at Omilia, lists a few of the major challenges IT departments face in choosing natural language processing apps for businesses:
Getting to a short list of vendors.
“There’s a lot of NLP offerings out there, but not all of them are created the same,” he says.
“It depends on your goals: there are differing levels of complexity and it depends on where you want to deploy it. If you want to deploy on your website for customer interaction and streamlining the web experience, you’ll need a vendor that has an NLP able to integrate into your website and supplement the NLP experience with rich visuals.”
This is a vendor-specific market, so you need to look for your company in your proposed vendor’s offerings, he says.
“The key to getting a functional system in a call center is speech recognition, so look for that in a vendor for call center solutions. They NEED to provide speech-to-text functionality.”
Track record: does the vendor have one?
“I don’t think anyone would want to be the Guinea pig. Referenceable case study where you can envision yourself in the success story.”
Johnsen adds that most IT departments have made big investments in their systems of record, so the trick is finding ways to augment those investments with new, emerging technologies that improve the consumer experience.
“If you have a CRM system, a conversational digital assistant can serve as an embedded, engagement layer above the CRM. Mobile is also a vital part of any IT strategy, but IT departments really need to come to grips with the notion that investing in more apps may not be the answer long term.”
Friction, both internally with employees and externally with customers, must be a major consideration. If the technology requires a user manual, downloads, passwords and a learning curve, that’s a path IT departments can now avoid, thanks to this new generation of technology.
Alison Thaung Smith, chief scientist at Booz Allen Hamilton, says the biggest challenge is that there is no “one size fits all” approach.
“Available software vary in ease of use and sophistication – an ideal solution for one use case may be dramatically different for another use case,” Smith says. “Having a skilled team that can thoughtfully assess the needs and requirements is paramount to choosing the right solution.”
The key factors any IT department should consider include:
- project scope
- workforce skills and expertise
- organizational priorities
Finding the right solution for you
If the technology is to be effective in helping consumers handle processes that go beyond “turn on the lights” or “play the Beatles,” then it’s important to look at domain-specific solutions that are purpose-built for specific industry verticals and the specific solution workflows within those industry verticals, Johnsen says.
In healthcare, for example, workflow packages would target high-value areas such as patient scheduling, referrals, reminders, adherence and intake processes.
“In these workflows, context is everything. Consumers don’t understand how the business works, so the provider organizations have to direct and orchestrate for them. That means the NLP portion of the solution is just a slice — the tip of the iceberg.
“Underneath the waterline is a deep-stack system capable not only of integrating to the systems of record of the service organization, but also capable of managing the controlling logic and state flow of multiple conversations over the course of a process journey. Healthcare engagements are journeys, not single-session conversations, so the systems need to do more than simply respond to a consumer’s questions. Other industry verticals face similar dynamics.”
Justifying the costs
What is a reasonable package depends on the company, but a good rule of thumb is to calculate the cost it would take a human to analyze the data you want your NLP solution to ultimately take over, and make sure your return on investment is worth it, says David Ciccarelli, CEO and Founder of Voices.
Johnsen says another important measure is to see how other similar companies are using the technology.
“This isn’t a new concept, but slick marketing materials and demos are not enough. Is the technology deployed? Are other companies talking about their success publicly? Are they willing to take a reference call? If the answer is no, that’s a warning sign.”
Stephen Blum, Founder & CTO at PubNub, says a lower-accuracy AI can be less expensive than a high-precision NLP, but even then either choice will save thousands of human hours. It may be that the business is able to easily tolerate lower accuracy to net the hours saved by human employees.
Agen says IT should keep in mind that the costs of labor and having someone on the phone with a customer typically works out to roughly 90 cents per minute, per employee on the phone.
“That could cost anywhere from $5 to $20 dollars per phone call. A company that may have 5 million or 50 million calls a year — even at only 1 million calls, if you can automate an additional 20% of calls, multiplied by the call volume, the result is clear: companies can save a lot using NLP to handle at least some of those calls.”