How Sharpen Benefits from Deepgram
The Landscape
Better Customer Service Starts With Better Tools for the Agent
Sharpen is disrupting the traditional contact center with a cloud-based, agent-centric platform that simplifies customer service interactions and serves over 200 global customers. As a result of COVID-19 and the urgent shift to a remote workforce, better customer experiences are more critical than ever.
To boost customer interactions, Sharpen first focuses on the agents. “We believe that a better agent experience leads to a better customer experience, and that starts with providing agents with the tools and resources they need in one place,” said Adam Settle, Vice President of Product at Sharpen. “If an agent is frustrated due to navigating many screens, then that behavior may negatively affect the customer. But if the agent has the right tools at their fingertips and uses technology to take care of their customers, it will make a big difference.”
The Challenge
Getting the Value Out of ASR
One of the most crucial tools in the arsenal of contact center solutions is automatic speech recognition (ASR). Sharpen supervisors, quality assurance and operations team members alike rely on ASR to identify coaching and training opportunities for agents, and maintain compliance with their customers. Prior to Deepgram, Sharpen used a solution that relied on older speech-to-text methods (tri-gram model) and offered less accurate transcriptions as a result. When one of Sharpen’s biggest customers raised concerns about the ASR’s transcription quality, Sharpen set out to find a replacement tool that would meet its needs.
The Solution
Increasing Transcription Quality and Speed, Thanks to Deepgram
Deciding to try another vendor, Sharpen started its journey with Deepgram by feeding hours of existing calls into Deepgram’s general audio model, and immediately noticed a significant improvement in transcription quality. Deepgram’s ability to auto-learn and facilitate the creation of custom voice recognition models made the decision a no-brainer for Sharpen. “When we compared the result to our previous solution, the difference was night and day,” said Settle. “The trained model feature was especially a game changer for our customers, as they could generate even more accurate transcriptions based on the type of data or keywords that mattered most to them.”
Unlike other ASR solutions on the market, Deepgram completely rebuilt the speech processing stack and is the only speech recognition company powered by end-to-end deep learning. Deepgram’s deep learning approach constantly enables the transcriptions to continuously improve, delivering even greater accuracy and understanding over time.
We first considered creating an ASR solution in-house, but there’s a significant difference between dabbling in transcriptions and having a reliable speech recognition product. Ultimately, open source options are expensive to successfully pull off, and building an ASR tool from scratch would add another feature that we would need to constantly support and improve, and spend valuable development cycles on.
The Results
Improved Accuracy and Usability
Better Accuracy: Working with Deepgram, Sharpen can provide highly trained and tailored transcriptions to its customers with greater than 90% accuracy levels. By pushing its models to perform under complex, real-life conditions with background noise, multiple speakers and diverse accents, Sharpen’s customers achieve vastly improved accuracy rates without compromising on the speed of the transcription. “We initially considered using a speech solution from a legacy, Big Tech company and realized that we would need to pay eight times the price for it, without getting a solution that was eight times better,” said Settle.
Improved Voice Search Functionality: Sharpen’s customers use built-in reporting to look for keywords and phrases in collected audio data rather than a faulty outputted transcript, enabling them to seamlessly pinpoint specific timestamps and gather helpful insight. “Now that our customers have transcriptions they can rely on, they can layer custom use cases on top of other contact center metrics, which is really helpful from a quality assurance standpoint.” said Settle. “Instead of just randomly grabbing calls, coaches can use specific keywords, in addition to other metrics, to identify the right call at the right time.”
Looking Ahead
A Partner to Help Foster Innovation
Likely as a result of the COVID-19 pandemic, there has been a recent resurgence of coaching and quality assurance automation needs in the contact center. The ability to provide fast, accurate and reliable transcripts allows Sharpen to provide more granular insights to its customers. Looking ahead to the future, taking advantage of Deepgram’s real-time capabilities will enable Sharpen to launch new features and products that will continue to differentiate the company from other contact center as-a-service (CCaaS) players.
With the speech recognition element of its platform squared away, Sharpen can focus its data science and development resources on the tools and analytics that its customers need to be successful. Settle says, “Having a technology partner such as Deepgram that innovates rapidly and delivers quality audio transcriptions, made possible by deep learning, will continue to open up new opportunities for our business down the road.”
