What Speech Technology Buyers Really Want: How to Meet the Needs of Enterprise Customers

Speech has entered a new era, one where its enterprise uses are virtually limitless. While chatbots are emerging for many consumer and customer service applications, the enterprise market has been largely hamstrung beyond those use cases. For acceptance to spread and for speech to reach new heights, the development infrastructure needs to improve significantly.

Speech has been a popular customer service option since interactive voice response (IVR) systems emerged decades ago. In many organizations, that area is where the technology is still largely found. “In most cases, companies use speech for customer support and marketing but little else,” says Deborah Dahl, principal at Conversational Technologies.

But other applications are possible. “Companies could rely on speech to enhance employee productivity,” says James A. Larson, vice president at consulting firm Larson Technical Services. For instance, he says, “A retail clerk could use their phone to check inventory if a customer wanted a pair of shoes in a particular size.”

Many employees work in areas where inputting information via a keyboard is challenging, and this makes them good choices for a voice-based revolution. Auto insurance inspections could take less time if agents could speak rather than type; and the inspection of bridges and highways and other complex field repair services, such as oil rigs, could benefit from voice interfaces, according to Dahl.

Speech Tech Puzzle Still Missing Key Pieces

While the possibilities are great, the use cases are limited to date because of speech app development shortcomings. Recently, vendors focused on improving the accuracy of natural language processing (NLP) systems. While not perfect, the systems now recognize many words and do a good enough job of accounting for idiosyncrasies, like a person’s accent, so they could be used for many applications.

What is now needed to take those steps is a more robust application development environment. Tying NLPs to enterprise applications is a complex, time-consuming, expensive, and often frustrating undertaking.

To make such connections, speech suppliers publish application programming interfaces (APIs) that enable third parties to connect their software to these systems. Given its fledgling nature, the NLP API ecosystem is immature and in some cases nonexistent. Consequently, the onus of putting the building blocks in place falls largely on corporations and their partners. Few businesses have been willing to take on the work because of its cost and complexity.

A lag time between when a new technology matures and when a software ecosystem develops is common. First, vendors build out base system functionality. Determining how to open their products up to outsiders is complex, time-consuming work taken on only after a product gains significant market traction.

To date, the focus of popular NLP solutions, like Amazon Inc.’s Alexa and Apple Inc.’s Siri, has been extending their ecosystems to the consumer market. “Voice skills are being developed to help consumers find information, like the local weather or directions,” says Sriram Chakravarthy, chief technology officer at Avaamo, a five-year-old conversational speech vendor that raised $25 million in venture funding from investors like Intel, Ericsson, and Wipro.

Speech Tech Is Not Yet Enterprise-Quality

Consumer-facing interfaces are much different from what enterprises require. “We looked at a few dozen Bluetooth headpieces and found only a couple that were suited for enterprise use,” says Bruce Rasa, CEO at AgVoice, a start-up building voice-enabled applications for the agriculture market. Without the hardware to support the company’s software, input is compromised.

In addition, connections to legacy systems are also largely missing. Enterprises want to leverage information stored in legacy systems such as SAP, Salesforce, and popular databases. Currently, the NLP APIs and toolkits offer little to no help there.

Compounding the problem is the emergence of new types of information. “In the agriculture business, increasingly consumers want to know if their food was grown in a safe, environmentally friendly way,” says Rasa. Food producers are starting to collect such information, and software has to be written that outlines how this information is stored in a database and used by other applications.