The State of Intelligent Virtual Assistants
The desire for companies to automate rudimentary tasks is pervasive today, so it’s no surprise that intelligent virtual assistants (IVAs) have been gaining interest. In many respects, they represent the latest wave of speech-enabled automation solutions.
To show just how pervasive IVA technology is becoming, research firm Global Market Insights estimated the global market for such solutions at more than $1 billion last year, but it expects the market to grow to more than $11.5 billion by 2024, at a compound annual rate of 37 percent.
Fueling this sharp increase is growing consumer demand for online self-service, self-reliance, and rapid query resolution, while at the same time helping companies enhance operational efficiency and reduce costs.
IVAs will allow companies to reduce call wait times and resolution times and soften call transfers in cases where escalations are needed; they respond to text or speech queries through smartphones and apps. Of the two, speech recognition is expected to witness higher growth.
Global Market Insights expects the financial services, automotive, IT and telecom, retail, healthcare, and education segments to see the greatest adoption, but it notes that small and midsize businesses are increasing adoption due to the minimized need for full-time customer service employees.
The research firm identified Apple, Anboto, Artificial Solutions, Clara Labs, eGain Communications, Existor, IBM, Next IT, Nuance Communications, and Speaktoit as key vendors in the IVA market, though many other players exist.
The Year in Review
IVAs have the potential to automate a wide swath of routine business tasks. Contact centers are one area that has relied on speech systems for many years and continues to be an area of emphasis among IVA suppliers, according to Ian Jacobs, principal analyst at Forrester Research.
Take Salesforce.com, for example. It announced its Einstein Voice Assistant in November 2018, and that first iteration was built for sales personnel, who continue to be a focal point of system development. This past November, Salesforce added Einstein Call Coach, allowing managers to use conversational artificial intelligence to glean insights and trends. Natural language processing identifies keywords in sales call transcripts; alerts managers to trends, like a spike in competitor mentions; and then presents best practices.
The vendor also extended its Einstein Voice Assistant so third parties can customize it for other use cases, like updating fields, creating tasks, or reading out predictions. By building their own or deploying others’ Einstein Voice Skills, businesses can create custom voice apps supported by the Salesforce Customer 360 Platform.
Nuance’s Project Pathfinder, which was unveiled a year ago, is a machine learning solution designed to increase IVA business intelligence. The project’s goal is to increase appropriate responses in vertical markets, like telecommunications and healthcare.
Virtual assistants need to be trained to understand industry word usage, and to date, that task has been largely manual, time-consuming, lengthy, and prone to human error. “With conversational design being so heavily manual, we applied our deep learning expertise to build tools to automate the build of enterprise IVAs through existing data streams like contact center call and chat logs,” explains Tony Lorentzen, senior vice president of omnichannel solutions at Nuance.
Other vertical use cases emerged in 2019. Kasisto launched KAI, an IVA for commercial banking, in October. The solution enables employees to access information like payment activity, account balances, and cash exposure, and it automates such routine tasks as reporting, reviews, approvals, and reconciliations
In July, Interactions moved into the food service sector with its Guest Experience Platform (GXP), an IVA developed with Canada’s Pizza Pizza. The solution is designed to streamline phone, digital, and drive-through ordering; improve customer care; reduce operational costs; and boost revenue. Its Kitchen AI feature automates preparation and cooking steps via intelligent voice input to free up employees’ hands so they fill orders more efficiently.
A Look Ahead
While IVAs have enormous appeal, many early iterations have not been effective. “AI is no longer the silver bullet cure-all,” Lorentzen notes.
These systems fall short for a number of reasons. The first wave of solutions was rules-based, offered consumers a very limited set of options, and lacked sophistication. They could answer simple yes or no questions or quickly pull out data, like current account balances, but answers that depended on a number of factors often failed.
One challenge is that humans can give different answers to the same question but still have the same basic meaning. IVAs typically cannot make such connections, so their effectiveness diminishes when conversations veer in directions beyond their basic programming.
Because of the missteps, suppliers pivoted in delivering their marketing messages: “Vendors have been running away from the term ‘chatbot’ because of the negative image that these systems now have,” Forrester’s Jacobs says.
IVAs have become the preferred term, and Jim Freeze, chief marketing officer at Interactions, makes a distinction between the two. “Chatbots can only answer simple questions because they do not have connections via APIs to business back-end systems,” he says. “They need access to information in a billing system, for instance, to answer a customer’s question.”
Customization is another area of increasing importance. Building one solution for all corporations can miss the mark because every business is unique, according to Michael Southworth, general manager of intelligent self-service at Verint Systems.
In April, Verint began packaging services and business process analysis solutions to increase IVA effectiveness. The Verint AI Blueprint is a conversation analysis system that identifies potential IVA use cases. It leverages machine learning to analyze, classify, and label structured and unstructured conversational data and then determines whether the business could automate processes. If the AI Blueprint identifies a potential business need, the system returns customized recommendations. It identifies which use cases will generate business value, develops a range of measurable key performance indicators (KPIs), and generates a project road map.
Replacing manual interfaces to business processes with IVAs can automate workflow, reduce mistakes, improve customer satisfaction, and drive revenue. The pluses have driven corporation interest in the solutions. To date, the reality has been less than the promise. To improve system performance, suppliers are emphasizing vertical use cases and tailoring their solutions to each client, steps that they hope will move the market in a positive direction in 2020.
Paul Korzeniowski is a freelance writer who specializes in technology issues. He has been covering speech technology issues for more than two decades, is based in Sudbury, Mass., and can be reached at paulkorzen@aol.com or on Twitter @PaulKorzeniowski.