Speech Platform Architecture

Article Featured Image

Whether your interactive voice response (IVR) system is premises-based or outsourced, contemplate the consequences of ignoring its architecture. Consider the angry backlash of consumers when Web sites are unavailable or functionality is defective: Blogs blaze with condemnation of high-profile organizations’ inability to serve the public. When speech-enabled platforms (Web sites in their own right) disappoint, the same frustration occurs, but communication is usually quiet—until it hits your customer retention, revenue, or profit measures. 

Given consumer backlash about touch-tone routing menus, organizations are moving toward placing a speech platform in front of their automatic call distributors (ACDs). The IVR is rapidly moving from a behind-the-switch, primarily self-service appliance to an in-front-of-the-switch, flexible gateway to callers.  

Consider this outlandish analogy for perspective: You’ve convinced your CEO that to provide the best service for your customers, your organization needs to raise the salaries of its customer-facing representatives to $100,000 a year. (Speech applications don’t come cheap!) However, you have not given any thought to your telephony, so you end up buying an ACD off eBay, and your carrier is a start-up from a third-world nation. Now that your customers cannot reliably speak with your fantastic and well-paid representatives, do you think your CEO will see a problem with this picture?

Just Like the Web

If your IVR is premises-based, then consider supplementing your IVR and speech recognition vendor’s architectural design with your organization’s Web architects. Web and VoiceXML IVR architectures are extremely similar; your Web architects will easily grasp the similar concepts. Outsourcing your speech platform brings a strong team of architects who constantly receive real-time performance feedback and regularly engage with speech technology vendors for architectural improvements.

The main architecture points to consider are resiliency, scaling, integration, operational efficiency, and cost.

For resiliency, design must first consider availability and survivability within a single site, and then extend the same to multiple sites. Usually a load-balanced design across multiple sites is best, but if your sites are extremely hardened, a primary/backup configuration could be a reasonable design. Nevertheless, understand that with primary/backup, a site outage can cause you to shut off inbound calls for a minimum of five minutes and up to 30 minutes based on your voice carrier management situation.  

Design for scaling should consider future growth as well as spikes for seasonality or out-of-norm operations. You should consider the ability to add platform components when required, as well as ensuring that your initial design addresses performance by including load balancing across Web application servers and content caching by external caching devices.

The platform should be capable of integrating with other customer support elements of computer-telephony integration, ACD, and agent-side applications to enhance data integrity. Moreover, your platform should certainly support current VoiceXML standards (currently version 2.1). Vendors are still selling hybrid solutions, and it is not always clear whose platform is truly standards-compliant. Hopefully you’ve already enabled your back-office applications for access via the Web. If not, then your IVR project timetable should include the amount of work needed to provide Web-enabled services to integrate to your speech applications.

Operational efficiency refers to both the cost and time required to maintain your platform and the applications residing on it. The design should provide flexibility to make changes without affecting caller experience and without requiring an army of staff to carry out the changes.  

Moreover, a good design should be able to address the previous four points in a cost-effective manner.

A good investment in speech platform architecture will present a tremendous return on investment while providing the means to protect your organization’s reputation through the reliable delivery of well-designed speech applications.

Kevin Brown is an architect at EDS, an HP company, where he specializes in speech platform design. He has more than 15 years of experience in designing and delivering speech-enabled solutions. He can be reached at kevin.c.brown@hp.com.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues