What makes SpeechCycle Automated Agents Unique?

SpeechCycle Automated Agents are capable of not only providing information, but also reaching out to a caller’s environment to reboot a modem or assist in troubleshooting in other equipment such routers or specific files on a PC. A key differentiator is the ability of SpeechCycle Automated Agents to identify and confirm problems to an extremely granular “micro-symptom” level. Callers thus feel that they are understood and are encouraged that progress is being made. As a result, average call durations or 8-10 minutes in speech automation (application dependent), and sometimes significantly longer are unheard of within virtually any other speech recognition solution.

What is Natural Language speech recognition?

This is NOT “press or say” technology. SpeechCycle utilizes a form of speech recognition technology called Natural Language Understanding to allow callers to express their issue in their own words. The foundation of Natural Language is understanding not necessarily what callers say, but rather what they mean. This is accomplished by matching utterances to what is known as a Statistical Language Model (SLM). The SLM is created by “training” recognition via capturing thousands of calls, transcribing them, and then assigning appropriate meaning. The result is accurately identify what caller’s are saying even if they express the same issue with a different set of words or phrases. For example, Natural Language speech recognition will understand something like “I turned on my TV and while I was talking to my wife the screen went from fuzzy, to white snow and back.”

What makes SpeechCycle’s Natural Language different?

• “Domain Experienced Data” – SpeechCycle Statistical Language Models (SLM’s) are trained on 100’s of thousands of utterances specifically for Complex Customer care Calls
• DART – SpeechCycle patent pending Dynamically Adaptive Response Technology elegantly recovers from lower speech recognition confidence scores to improve caller satisfaction in Natural Language applications.
• Symptom Identification and resolution – SpeechCycle Automated Agents are able to identify an extremely granular level of problems and issues, including some that are infrequently encountered. This capability is supported by SpeechCycle tools such as VKP and Annotation Manager.
• Transition to Directed Dialog – SpeechCycle applications will transition to guiding callers to say something specific if a dialog with more guidance helpd to drive to a solution more quickly

How does speech recognition handle callers with accents?

Speech recognition software has improved over the years by capturing millions of utterances from a huge population of callers, and then modifying what are know as “models” for utterance capture. As a result, speech recognition is now extremely accurate with an ability to recognize callers with accents and regional differences. However, its not perfect. Just like humans understanding, there are challenging similarities. Fore example “B” can sound like “3”. However, for the most part speech recognition to date is extremely accurate. Dialog recovery and disambiguation (did you say this or that), makes speech recognition today extremely conversational.

How can SpeechCycle Solutions improve caller satisfaction?

Call centers dealing with complex customer care calls require well-trained agents who are up to date on all technical aspects of a caller’s environment. Agent turnover, training schedules and a lack of agents available to answer calls can all negatively impact caller satisfaction. SpeechCycle is able to improve caller satisfaction due to:

• Immediacy – No waiting for an available agent
• Consistency – Every call answered with an identical knowledge base
• Dialog – Intelligent mix of Natural language and directed dialog
• Continuous Improvement – Content and speech frequently updated
• Context Sensitive Help – Dialog assistance for each point in a call
• Engagement – Interactive diagnostics to provide caller encouragement
• “Micro-symptoms” – Identify and confirm very specific problems and issues
• Call Resolution – Completely resolve a significant portion of calls

What degree of call resolution can be expected?

The specific application and level of integration to key billing, provisioning and diagnostics system all have a bearing on automation rates. A non-integratreted SpeechCycle solution relies more heavily on a Q&A process. Integrated solution can identify and verify callers, demonstrate to callers that the application knows the specific make and model of devices in their home, and automatically reset equipment at a caller’s site to help increase automation rates. Full automation rates will tend to range between 20% to 40% depending upon level of system integration. A significant percentage of calls call also be partially automated,

What is the value of partially automated calls?

Partial automation means that the LevelOne Automated Agent has engaged in some diagnostic steps such as checking modem lights, assistance with reboot etc. Escalation to an agent may be required because reasonable troubleshooting efforts have been exhausted, or simply to connect to an agent by caller request. In these cases a forward looking screen pop can be sent to an agent providing a “next step” suggestion along with some background about the call. Value includes potentially reducing agent handle time as well as customer satisfaction by avoiding repeating a diagnostic step.

How is performance verified?

SpeechCycle reports track a wide range of issue including full and partial call resolution rates, problem frequency, overall call volume, etc. Reports provide evidence of success as well as assisting in identifying issues facing a caller population.

How do these applications integrate with contact center agents?

SpeechCycle applications are designed to transfer a call to an agent when automated diagnostics are exhausted, or when requested by a caller. Additional integration will enable SpeechCycle application to send a forward looking screen pop to the agent desktop. Information provided includes a next step suggestion as well as detail on previously completed steps.

How are these applications deployed?

SpeechCycle LevelOne Automated Agents can be deployed as a hosted solution, or a premise-based managed service.