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

Digital Engagement Blog

NLU Router Use Cases

The Sami people, who live in the northern tips of Scandinavia and Russia, use at least 180 words related to snow and ice. And as many as 1,000 words for reindeer.  

What does this have to do with NLU?  Read on... 

Sami Women (wikipedia) who look like they might be contemplating NLU use cases

Use Cases for an NLU Router

Before we dive into these use cases, you might want to bookmark these recent posts that provide additional background info and perspective leading up to the NLU use case discussion: 

  • This post about conversational interfaces discusses about good and bad use cases for conversational agents and speech IVRs.
  • This NLU Cheat Sheet post is a great backgrounder to boost your general understanding of natural language.

Ready?  Let's dig into the best use cases for an NLU Router:

The Brains:  The NLU Engine

The NLU router has a brain (that we like to call an NLU engine) that is used to extract information and meaning from the spoken conversation and determine caller intent. The NLU router relies on that brain's intelligence to manage the conversation, personalize the interaction and choose the best next action for routing.  It can carry on a dialog asking follow-up disambiguation questions if it needs to clarify intent. 

So a good NLU router does a lot more than “route.”  It uses context and intelligence to create an engaging experience. 

"Fine-Grained" NLU Router

We are going to talk about use cases where a fine-grained NLU router can drive strong business outcomes around CX and ROI. 

Fine-grained means that it can understand call reasons and intent down to a very fine level of detail.  The first use case below highlights what fine-grained means and what it can do.

To give you an example of fine-grained NLU, our NLU router can intially recognize over 250 distinct call reasons, and treat each one with a unique business rule for follow-up action.  By asking additional questions, our NLU router can drill down further to recognize over 400 distinct call reasons.

NLU routers are used in many places, including IVRs, Chatbot, Virtual Assistants (VA) and services like Amazon Lex. We find that the IVR space currently has the most advanced and fine-grained NLU routers deployed in an enterprise operational setting.

NLU understands meaning and predicts intent

Important!  Our point of view is that most speech IVR offerings (and many other systems) get NLU wrong.  NLU is only one piece of the pie.

NLU needs to be wrapped with a well-designed experience.

Having an NLU system understand 250 distinct meanings is only useful when you have deep user experience uniquely designed around each and every one of those meanings.

Five Uses Cases for an NLU Router

If you think you might be interested in an NLU router, take a closer look at the problem you want to solve.  These five categories of use cases are the strongest indicators of a good fit for an NLU router: 

  1. Wide variety of contact reasons
  2. Diversified agent skills
  3. Multiple automated call drivers
  4. High contact-per-customer rate per year
  5. Rich customer account data

 [1] Wide Variety of Contact Reasons

The Sami people, who live in the northern tips of Scandinavia and Russia, use at least 180 words related to snow and ice.  And these distinctions are borne not of excess, but necessity.  "These people need to know whether ice is fit to walk on or whether you will sink through it...It’s a matter of life or death."

But what about the contact center?  What are the risks of treating "billing" as one request and one task?

The more reasons there are for a customer to contact a company, the more difficult it can be for a caller to shoehorn their reason into a menu option, and the more difficult it is to design logical sub-menus that don’t frustrate callers.

Such a complex menu places a big burden on the customer…no wonder consumers hate IVRs.

Consider Billing as just one example.  Billing could mean -

  • ‘make a credit card payment’
  • ‘pay my overdue balance’
  • ‘check on status of a payment’
  • ‘make payment arrangements’
  • ‘set up autopay’
  • ‘pay from my checking account’
  • ‘I’m moving’
  • ‘find a payment center’
  • ‘schedule a vacation hold’
  • ‘I didn’t get a statement this month’

In a DTMF world or even a speech-enabled IVR world without a fine-grained router, all of the above would likely fall under a generic ‘Billing’ category.  It would be nearly impossible to determine the more granular call reason with any accuracy.

Many customers would choose ‘make a payment’ at a sub-menu.  Many customers would get frustrated because couldn't find an option listed that described their reason for calling.

The result would be 45-90 seconds of frustration as the customer wades through multiple menus, and still ends up routing to the wrong place, where they have to explain themselves again from scratch…no wonder consumers hate IVRs.

Using an NLU router, each of the 10 billing reasons route to a distinct route point (either automated self-service or agents).  The NLU router also greatly reduces the customer time from the beginning  of the call to the route point-- down from 45-90- seconds to 11 seconds in typical deployments.

A NLU router reduces customer time and effort in this use case.

 [2] Diversified Agent Skills

If contact center agents are trained only on a small subset of call drivers, it becomes crucial to ensure customers are routed to the correct queues.  Problems of being routed to the wrong agent queue include:

  • expensive follow-up internal transfers (as an example, a transfer can add another $5.00 to the call cost)
  • wasting the caller’s precious time, which is never good for CSAT scores

Agent-to-agent transfers are one of the biggest caller complaints, and an enormous waste of caller and agent time.

With a menu-driven system, misroute rates are typically about 25-30% of total calls.  So if you’re taking one million calls a year, misroutes could be costing you $1.25 million.

An NLU router is a game changer in an environment with diversified agent skills. Our experience is that misroute rates can be cut by 50% or more.   

An NLU router reduces misroutes, agent transfers, total call length and customer effort.

Bonus benefit: With the granular view of contact reasons a fine-grained NLU router provides, it becomes possible to examine patterns in contact reasons and missed opportunities for automation. Better routers drive better analytics so you can achieve better performance through continuous improvement. 

[3] Multiple Automated Call Drivers

Many enterprises have IVRs that live in their own private Idaho.

These Idaho IVRs automate very specific call drivers.

Idaho IVRs are useful and needed. But they are usually disparate and unconnected and that reduces their effectiveness.

Usually Idaho IVR gets very little call volume because:

  • their upstream IVR’s have no way to determine granular call reasons, or
  • they are buried so deep in a menu hierarchy that callers never hear the choices

In contrast, a fine-grained NLU router can recognize granular call reasons up front.  It can then route immediately to an Idaho IVR (one that automates a very specific call driver) to maximize effectiveness and get better performance from an existing IVR asset.  

Some good examples of Idaho IVRs that our customers have implemented are:

  • Order a replacement remote control
  • Get a hard copy of Terms & Conditions
  • Look up a local store that processes payments

 Let’s look at “order a replacement remote control” using a DTMF IVR.

To access the IVR to place an order for a remote control in a menu-driven system, a caller would have to choose:

  1. Tech Support (from 5-item menu)
  2. TV (3-item menu)
  3. No (I’m not calling about an outage: Yes/No question)
  4. More choices (after hearing a 4-item menu)
  5. Remote control (6-item menu)
  6. Order a replacement (3-item menu)

Then they would route to an agent to finish the order.

Compare this to an NLU router, when a caller says “my remote control doesn’t work.”

The NLU router can immediately understand the caller's intent, perform some basic troubleshooting, and then submit an order for a replacement if the troubleshooting doesn’t work. Quick, easy and no agent involved.

A NLU router improves operational efficiencies, reduces call length, reduces customer effort and increases self-service rates.

[Personal note on Idaho:  My team shot a 2 on the 14th hole at Coeur d’Alene golf course]

NLU Engine Understands Intent

[4] High Contact-Per-Customer Rates

The more often a customer calls, the more we learn about them, and the more we can provide customized handling based on contact history.

Then we can provide a context-rich, personalized and even predictive interaction.

When callers call in multiple times, we can build historical customer profiles and an NLU router can tailor and personalize the caller experience.

The interaction can be more sympathetic and predictive.  The NLU router can offer likely reasons for the call upfront to help callers get resolution faster.

For example, if a caller calls in at the same time of the month every month to make a payment, we can recognize that pattern and provide a fast-tracked experience: “Are you calling to make a payment today?”

A NLU router can also incorporate past selection in payment amount and payment method, and route immediately to the optimal automated call flow.

A NLU router, in a frequent contact-per-customer environment, can improve customer experience by personalizing the interaction and reducing customer effort.

 [5] Rich Customer Account Data

Access to back-end data allows an NLU router to provide customized handling, reducing handle time and increasing customer satisfaction and trust in the system.

Some examples of back-end data that can be leveraged for customized handling:

  • Payment date due
  • Make a payment
  • Pre-populate forms
  • Appointment date/time
  • Equipment type
  • Subscribed services

As an example, a cable company wanted to upsell a boxing match as a pay-per-view event. Their strategy was to do an outbound call to inform all customers of the match and allow them to order the fight.  However, this would have caused a 20 second increase in the handle time of each inbound call.

We proposed an alternative: provide targeted messages to a segment of their customer base.

With an NLU router, with personalization and rich customer data we could send a messaging only to those customers who subscribed to TV service.  (Those that did not subscribe to TV could not order the fight anyway).  The result was an enormous savings in handle time, caller frustration and loss of trust in the system.

An NLU router in a data-rich environment can really sing:  it  can cut handling time, reduce caller frustration and drive sales.

NLU Router:  Put It On Your Short List

Our experience is that an NLU router is a great fit when any of these five use cases are present:

  1. Wide variety of contact reasons
  2. Diversified agent skills
  3. Multiple automated call drivers
  4. High contact-per-customer rate per year
  5. Rich customer account data

These use cases show how an NLU router can drive out complexity, improve customer experience and increase operational efficiencies.  Evaluate your own business against these use cases to better understand if fine-grained NLU router will deliver business benefits in your environment.  If so, you'd better make it a consideration on your strategic priorities list, expecially if CX and cost savings are important drivers for your business.  


The next time you think about an NLU router, think about 180 words for snow.  Think about how many words for "billing" your customers use.  And consider whether your current systems sufficiently understand all those words to act upon them effectively.

Forrester Report

Forrester Report on AI

Free download of Forrester's The Top Emerging Technologies In Artificial Intelligence

Get Your Copy

Topics: IVR conversational IVR NLU