Five9 Removing Barriers to Useful AI

Five9 made three AI announcements this week: the general availability of its virtual agent, release of its agent assist offering, and a framework for recording called VoiceStream. These represent the expansion of its previously announced Five9 Genius that are intended to address several pitfalls related to how contact centers implement AI.

 

AI technologies make for great contact center demos, but implementations are much harder. There’s a lot of practical barriers to obtaining a measurable ROI and driving specific outcomes when it comes to AI in the contact center. There’s also the issue that every vendor wants to deliver differentiated value with their products, and that’s hard when most are using the same underlying technologies.

 

We frequently hear how organizations are using AI, so the idea of implementing it in a contact center seems reasonable, but implementing AI as a solution is very different than enabling it as an effective tool. We got exposed to that with Google’s announcement of contact center AI (CCAI) and its general availability, which came almost two years later.

 

Google CCAI provides a set of tools that can enable virtual agents or agent assistant modules in contact centers. But tools don’t address outcomes or implementation; that’s up to each partner provider (and customer). Five9 has taken Google’s CCAI (and other AI engines) and extended it with its intellectual property and other partnerships, resulting with a ready to go, differentiated offer.

 

The first announcement is the general availability of Five9 Virtual Agent. The concept is familiar, a scalable bot framework that can intelligently resolve or route calls on its own. What’s new here is Five9’s integration with Inference and its own intellectual property. Inference provides a Visual Designer (Inference Studio) that simplifies setup and implementation. A rich API platform makes it easier to integrate the Five9 Virtual Agent to other services. Under the hood here is the Whendu integration platform that Five9 acquired last Fall.

 

The result is a significant reduction in the time to implement. Five9 cited the Colorado Department of Labor and Employment’s successful implementation with a COVID-19 use case, resulting in 87% of the calls being handled without a live agent. The new Virtual Agent works with Google CCAI, as well as AI services from AWS and IBM, and supports both tree (IVR) and fluid interactions.

 

Also announced was the general availability of Five9 Agent Assist. Again, a familiar concept with a new twist. The key feature here is a feature Five9 describes as assisted summarization, which uses AI technologies to create summaries that the agent then approves. Assisted summaries offer an improvement over dispositions, which are too brief, and manual notes are too time-sensitive.

 

Five9 Agent Assist listens to agent interactions and creates a summary of each call. It’s presented in the agent desktop, a web app that simplifies implementation. The feature has three components: a transcription, a summary (built as the call progresses), and a submit to CRM key that supports most CRMs.

 

A summary is more significant than it might first appear. Clearly, it reduces agent processing time, but more importantly, it also addresses AI training. The agent makes corrections to the summaries which addresses both mistakes and ongoing training. AI always makes mistakes, so there needs to be mechanisms in place to fix them. Mistakes are visibly apparent in transcription, but all AI use cases are about percentages. The summary flips training from a chore to a benefit. Reliable call notes are critical to understanding what occurs in a call center.

 

With both applications, Five9 has created differentiated and deployable AI solutions. The company has also announced a new set of tools and capabilities intended to simplify and accelerate deployment. Currently, these tools are limited to Five9 Professional Services with the intent to productize them for its partners.

 

Lastly, Five9 announced VoiceStream, a new framework for recording that is coming soon. It’s an intriguing evolution to how contact centers can record calls.

 

Recording is hardly new, but it hasn’t evolved as quickly as CCaaS. Recordings have historically been associated with post call use cases, and it’s been a struggle to adapt port mirroring or even SIPREC to real-time applications, such as sentiment and speech biometrics.

 

VoiceStream is a new framework optimized for real-time, cloud-to-cloud applications. VoiceSteam is effectively a call recording platform service (CRaaS?). It allows Five9 partners to tap into the Five9 call flow via API. VoiceStream provides support for real-time applications such as AI, speech analytics, voice biometrics, sentiment analysis, and more.

 

VoiceStream offers Five9 customers a point-and-click framework to integrate applications. It provides Five9 partners with a set of APIs for requesting real-time audio, text, or intents. Text is particularly interesting as speech-to-text conversions are often the first thing that many partners do, so now they can receive the spoken conversation as text.

 

Five9 VoiceSteam could very well become the way we approach real-time media streaming with other providers. The APIs are open and free, but the back-end VoiceStream engine is Five9 proprietary. The service is architected for multiprotocol support (SIP, GRPC, and SIPREC) and includes several developer-friendly attributes, including RESTful APIs and WebHooks. It’s also expected to meet rigorous enterprise security requirements with full encryption, OAuth, and PIC compliance.

 

The contact center industry has been crawling toward useful and differentiated AI use cases. I believe all three of these announcements are important milestones to increasing and expanding successful implementations of AI at contact centers.