24/7 targets an integrated customer experience
Core speech technology
hosting still at Microsoft
Microsoft
acquired Tellme Networks, based in
Mountain View California, in 2007 for about $800 million. Tellme Networks
provided a customer service automation (Interactive Voice Response, IVR) cloud
service featuring speech recognition. Tellme was using speech technology from Nuance Communications and IBM for its IVR services before the
acquisition. After the acquisition, Microsoft began using Microsoft speech
technologies and the Tellme Networks speech cloud service across a number of
its products, including Windows Phone (the mobile phone operating system with
speech recognition), the Bing mobile search app (with a voice search option),
and Xbox Kinect (which uses network-based speech recognition for games and TV
control, SSN, February 2012, p. 22).
On February 7, Microsoft and 24/7 Inc. announced an agreement that includes
Microsoft merging its interactive self-service assets (clients, people, and
technologies) into 24/7. The agreement also includes an R&D partnership,
long-term IP licensing, and Microsoft taking an equity stake in 24/7. Microsoft
and 24/7 have also agreed to a shared technology road map as part of a
long-term IP licensing agreement that provides broad coverage under Microsoft’s
patent portfolio for applicable technologies.
24/7 is a privately held company. With
the Microsoft IVR assets, annual revenues are projected to exceed $250 million.
24/7 is a global firm with sales and design/delivery centers in Campbell, CA;
Alameda, CA; London; Sydney; India; the Philippines; and Latin America. 24/7’s
customer base includes Capital One, Optus, Lenovo, Sirius XM, Adobe, United, and Avis. The
company has changed its name to 24/7 Inc. after operating under the name 24/7
Customer for twelve years.
Voxify
(Interview, SSN, September 2011, p. 19) had been acting as one company
designing applications for Microsoft’s Tellme platform, as Tellme moved to
hosting exclusively. The company will continue this role as part of 24/7, which
acquired Voxify late last year, according to Kathy Juve, Chief Marketing
Officer, 24/7.
24/7 has historically focused on web
and mobile solutions and is well-positioned to integrate voice service with the
other modalities, notes Juve. The integration will, for example, help avoid the
frustration of consumers who reach an agent and have to repeat data already
entered on a web site. The integration also allows more personalized service,
Juve indicates, by integration with backend databases, so a customer calling in
about a cancelled flight, for example, can be greeted by a prompt such as “I
see your flight was cancelled—Would you like other options?”
The actual cloud-based hosting
operations running the speech recognition engines stay with Microsoft, as one
might expect, since they also have to support Bing, Kinect, and the other voice
services Microsoft offers. Microsoft also retains the Microsoft Tellme branding
as part of its Online Services division. Brooks Crichlow, director, product
marketing, at Microsoft, noted that, while 24/7 will use the Microsoft speech
services, 24/7 does host the IVR portion of the former Tellme service that is
driven, for example, by VoiceXML-defined interaction.
Microsoft will continue its research
efforts in speech technology, with a focus on natural, conversational speech as
part of the company’s Natural User Interfaces (NUI) efforts. As part of the
R&D partnership, 24/7 will use Microsoft Tellme speech and natural language
technologies to deliver natural user experiences in their customer service solution.
The Tellme technology also supports other products and technologies across
Microsoft such as Windows Phone, Bing, and Dynamics CRM (Microsoft’s Customer
Relationship Management software). There is also a plan to integrate 24/7
analytic solutions with Windows Phone, Bing, and Microsoft Dynamics CRM,
according to a blog by PV Kannan, CEO of 24/7.
Juve notes that today’s consumers
expect to interact with businesses across a range of channels beyond the phone,
such as web, text, and email. These digital interactions also occur across a
range of devices such as smartphones, tablets, PCs and televisions, and
integration of these points of contact is a desirable goal. One of 24/7’s specialties is data
analytics at cloud scale (“Big Data”) aimed at this goal for customer service.
The company focuses on large enterprises in travel, cable, financial,
technology, and retail, using commonalities in customer service encounters to
leverage experience with these vertical market segments.
24/7 uses Internet-scale data and
advanced analytics to help businesses predict customer preferences and deliver
proactive, efficient and intuitive consumer experiences. With this agreement,
Microsoft and 24/7 are combining technologies that span interactive
self-service across mobile, Web, and voice channels; Big Data analytics; and
speech and conversational interfaces to create a “next-generation” cloud
platform for customer service. The combined Predictive Experience (PX) platform
will manage more than 2.5 billion speech and online self-service interactions
annually, according to the companies. With this combination, 24/7 Inc. will be
positioned to generate total revenues of more than $250 million annually, the
companies estimate.
Zig Serafin, general manager, Online
Services Division at Microsoft, said, “From speech to touch to gestures,
consumers expect and demand more natural and intuitive ways to interact with
technology. This same demand will change how consumers interact with
businesses, and it creates an inflection point for how people will expect
businesses to provide customer service. “
Consumers today not only expect
businesses to deliver service in more natural and intuitive ways but also
expect these same businesses to be able to reach out and interact intelligently
with them across a wide variety of channels, including on mobile devices,
through social media, on the Web and even through the living room TV. This
consumer expectation has created a key need for businesses to gather, analyze
and gain insight by using Big Data to proactively anticipate and predict
customer needs.
Kannan explained, “By bringing
Microsoft’s interactive self-service technology together with 24/7 Inc.’s
predictive consumer experience technology into one unified cloud platform, we
will deliver solutions and services that truly enable businesses to
differentiate through customer service.” Tellme is already supporting large
customers such as Avis and United Airlines.
Microsoft’s Serafin gave an example
of the kind of interaction the partnership would enable. Suppose your bank
notifies you that your credit card has been blocked due to unusual activity. In
the near-future when you get one of these notifications, your bank will know
how to send you the notification in a way that is best suited for where you
happen to be (for instance, on your smartphone if you’re on the go), and the
notification will not only already include the critical details you need to
evaluate whether there actually was an unauthorized charge, but will also
anticipate the actions you most likely will want to take immediately. You’ll be
able to reply directly to the notification on your phone and use your voice to
highlight the charges that are bogus and then take action, such as cancel your card
and have a new one issued. Serafin said that this type of service “will literally
create new industry leaders” and is “the future of customer service.”
Kannan, in a blog, characterized the
current state of customer service as “hopeless online self-service, frustrating
IVR systems, and half-baked mobile service apps.” The partnership will change
things, he said, in “three key ways”:
- A new design
framework summarized as “Anticipate-Simplify-Learn” for customer experience.
Kannan summarized the goal for the customer: “Anticipate what I want and how I
want to be served, make it quick and learn from my feedback, even when I don’t
expressly say anything.”
- A “Big
Data” platform that can be easily implemented on top of existing customer
service systems, so that they do not have to replace what exists today. The
cloud-based solution facilitates this approach.
- Data
Science is used to build a dynamic mathematical model that works in
real-time to observe customer interaction and feedback, learn what is important
to the user, personalize the entire interaction, and then integrate that
learning back into the situation, all in real-time.
In a long discussion with Bill Meisel,
editor of Speech Strategy News (an
engineering PhD whose technical background is in data analysis and
computer-based pattern recognition), Ravi Vijayaraghavan, vice president and
global head of analytics and data sciences, 24/7, and Patrick Nguyen, CTO, 24/7
(formerly with Voxify), discussed the company’s core technology and the basic
principles that will allow extending 24/7 technology to voice-based systems and
integrating those systems with other customer service modalities. 24/7 has
previously focused on text-based customer service interactions, such as chat.
They have used a data-driven approach to discover patterns in customer interactions
in specific vertical markets such as financial and retail segments.
Vijayaraghavan indicated that the
company did an oft-updated off-line analysis of many interactions (historically
from such sources as chat lines or web sites, with occasional surveys to
understand customer reactions at the end of an interaction). The analyses were
tied to customer databases when available that had information on the customer
and previous interactions. Empirical analysis (e.g., clustering and
classification methods—supervised and unsupervised methods), for example,
determined the best way to characterize the customer objective in the call
(e.g., re-booking a cancelled flight, what Vijayaraghavan called “intent
models”) or whether the call was successful in meeting customer objectives. The
methodology detected and summarized characteristics of a call and caller that
could be used to detect the category of caller objective. (The need for
summarizing “features” of a caller or interaction is necessary to make the most
efficient use of the limited data available.) Nguyen noted that most of this
work could be leveraged with voice interactions. For example, knowing the
categories into which one wanted to sort a call in a particular vertical market
segment could accelerate its use in the voice space. Information from a
previous chat or web interaction could thus be incorporated in the voice
interaction using this categorization.
The data-driven approach can be used
in many ways to improve the customer experience. For example, it can detect
“dissatisfied” versus “satisfied” customers or “sale” versus “non-sale” in
specific interactions and to address specific problems in an automated system
or in the information delivered to agents. The system uses “machine learning”
in that the repeated off-line analyses with new data create a continually
improving system. Nguyen noted that access to data across multiple channels and
from Customer Relationship Management databases leveraged this technology
beyond just one customer interaction modality. The technology can also in some
cases anticipate the likely category for a customer call (e.g., if they were
just notified of a flight cancellation) and be proactive in shortening the
customer interaction.
While there are still some legacy IVR
applications hosted by Tellme that use Nuance speech recognition engines, they
will all be migrated to Microsoft speech technology, according to Nguyen. Tony
Laurentsen, vice president, professional services, Nuance, indicated in an
interview that the version of Nuance software used by Tellme is an older
version that doesn’t have the accuracy of Nuance’s current versions. Nuance has
a competitive customer service hosting operation, which Laurentsen believes
handles more customer service interactions (about 2 billion calls annually
using speech recognition) than Microsoft Tellme did.
Microsoft launched its speech recognition research in 1993,
licensing the source code for Carnegie
Mellon University’s Sphinx speech recognition. Later, Microsoft bought Entropic Software, which had speech
recognition based on the HTK toolkit from the UK’s Cambridge University. The company has continued to invest in speech
and natural language technology in Microsoft Research. The move to 24/7, with
its web-based skills, allows the technology to be more effectively delivered as
a customer service solution that meets some of the oft-stated goals of serving
customers better in a multi-modal customer service environment.
