NewTech Friday: Daisee – uncovering true insight in spoken language

Richard Kimber is CEO and founder of Daisee. He spoke with Efma’s Boris Plantier about his company’s AI speech analysis solution that “hears between the lines.”

I have spent my career at the forefront of evolving technology and its application to providing business and customer solutions initially in financial services, insurance, and payments and evolving into broader internet and industry applications. Prior to founding Daisee, I could see the next big evolution in technology was rapidly evolving around Artificial Intelligence (AI) and Machine Learning. I could see that there were a number of opportunities to address business needs using AI, not just in Australia, but globally. I looked at many areas of AI research and looked for an area with global application. I founded Daisee with a view to develop business solutions using the most advanced AI and machine learning available. 
 
We explored both computer vision and voice AI and had developed several products in both areas. Even though it was extremely complicated and had significant barriers to entry, we took the strategic decision to focus solely on speech analytics because we had developed a highly effective solution that we wanted to develop further. We set the objective to be a world leader in using AI to uncover true insight in spoken language. 

 

Daisee's Artificial Intelligence (AI) speech and text analytics software helps businesses know exactly where to focus by analyzing every word their customers say to them. Our solution ‘hears between the lines’ and can evaluate the quality of every voice interaction using our proprietary algorithms. 

Daisee has the unique ability to automatically analyze and dynamically score 100% of customer interactions identifying issues that require human intervention. This is achieved using a world-first digital quality scorecard measuring the most important elements of each interaction. The benefit is that you can dramatically improve your control framework, deliver significant operational savings, and most importantly, increase your revenue generation and customer experience. 
 

The problem that our solution solves begins with the fact that historically companies used 1-2 percent manual sampling to attempt to evaluate how well they are servicing customers. Our product allows them to sample 100% of every word in every conversation with their customers and thus be completely compliant with regulations and rate and rank every interaction for communication, conduct, and compliance. 
 
We identified this problem following the Hayne Royal Commission into financial services misconduct that showed a systemic issue across many regulated financial services companies. The core issue is that complex products are often not understood, or features are not explained. The original focus was life insurance - but the issue is widespread - and although many companies say they are monitoring their calls for coaching and quality, it is humanly impossible to actually listen to all the calls. Daisee solves this issue. 
 
Our solution is described simply in that we translate voice recordings to text and then analyze the transcripts with advanced natural language processing algorithms and machine learning to derive a set of scores that evaluate the quality of the call across a range of criteria including sentiment, professionalism, active listening, and many more dimensions. The system then assigns every call an overall score. This score identifies high risk interactions and allows a company to focus on the calls that matter. They can then proactively intervene and fix the issues. 
 
Daisee helps businesses discover not just what happened and why but work to predict what WILL happen and how they can make those things happen to stay ahead of their customers and competitors. 

 

One of the exciting new features that we have developed and are just starting to roll out is Daisee Essence - a fully autonomous capability to summarize a phone call using deep learning and then derive the topic of the call by mapping every statement to each other and detecting the most relevant topic in the call. This can then be used to evaluate many factors, including process wastage, customer effort, and many other elements of improving customer experience. This has been turned into a production feature and is live with our first customers. 
 
We are also working on revolutionary technology to evaluate and improve live person chat and bot chat outcomes. This product will evaluate the text interactions between customers and agents, score and rate and rank them for effectiveness, timeliness, and how well they resolved the issue. We have developed a world first chat scorecard that leverages our voice algorithms and machine learning. We believe this is a world first and will have the potential to open up many new markets.  
 
Our roadmap is extensive and we plan to leverage channel partners to offer a range of additional functionalities that will allow them to configure and deploy low code scorecards for their insurance clients. 
 
Another big focus for us is the RG271 legislation and helping insurance companies prepare and improve their complaint handling identification and processes. We have developed a specific module for insurance complaints that relates to this legislation that we will be rolling out to customers imminently. 
 
We have created a series of predictors that enable us to use tagged data to create a model to identify future events. This is particularly relevant for insurance claims. Our unstructured data is being used to identify behavioral links to customer behavior and outcomes. We believe that this is a very new area and very few companies have the capacity to evaluate this data.