How AI can uncover the value in unstructured data

Extracting value from unstructured data is one of the biggest challenges in the insurance industry, but advances in AI are opening up new possibilities, as Pamela Negosanti, head of NA sales & sector strategy - insurance and financial services for Expert.ai, told Intelligent Insurer.

​The use of AI in insurance is growing, but it’s a continually evolving field that can be fraught with frustrations and unmet expectations. On December 1 a webinar moderated by Pamela Negosanti, head of NA sales & sector strategy - insurance and financial services for Expert.ai, will explore how to get the best out of AI.

Titled Discover How AI Uncovers the Hidden Value of Unstructured Data in Underwriting, the webinar will explore the challenges encountered by insurers when dealing with unstructured data and ways in which AI can help by automating processes that were previously time consuming and fraught with uncertainty.

Expert.ai for Insurance is a suite of AI services that reduces risk, improves win rates, and increases capacity for commercial insurance carriers and brokers. It automates the reading, understanding and extraction of meaningful data from text at speed and scale.

Through a unique mix of natural language understanding and machine learning techniques, expert.ai supports the broadest range of commercial insurance use cases in claims management and underwriting, meaning Negosanti is well-placed to dive into the subject of AI and how it can improve processes for insurers. Here she outlines what she hopes people will gain from the webinar.

What topics do you aim to cover in the webinar?
The main topic is the practical usage of technology that is already bringing value to underwriting; so rather than discussing AI in theory, and where can be in a couple of years, we will be looking at concrete projects that are active today and deliver tangible results.

We will also cover what didn’t work and why; statistics have shown that 85 percent of AI projects are failing. When AI is not delivering successful results, in many cases this is not a matter of the technology but more to do with mindset, attitude, skill and education.

Why did you choose this topic – why is it important today?
There is a lot of hype around AI and a lot of misinformation and miseducation, which is still leading to unrealistic expectations, disillusionment and scepticism: people end up thinking nothing will work. But AI can help deliver real data points that are not opinion, but facts.

What challenges does the insurance industry face in relation to unstructured data?
Today there is increasing awareness about unstructured data, and the automation of processes. More than 80 percent of documents that are processed in the insurance world contain unstructured data – these documents are not predefined in terms of format and so you cannot read them and come to any firm conclusion; examples of this type of document include policies, emails, claims, and medical reports.

A lot of insurers do not have a clear data strategy; they know they have to do something but they don’t understand where the sources are and where the documents are distributed across the multiple business units in multiple countries, so there is no unified strategy when it comes to managing the data.

One of the challenges is that the unstructured documents are complex to manage because you cannot forecast what they will look like and they create an increasing variability that you have to manage.

Unstructured data contains a lot of value for companies and the challenge is to understand and quantify it. Think about how many opportunities are not being won properly because you are not analysing the data that you have. There is an opportunity for improved operational efficiency, and strategic, medium to long term gains.

Please outline expert.ai’s offering for the insurance industry
Our main focus is on unstructured data. Our AI has the capability to not only read but also understand the data. To extract value from unstructured data in insurance you have to go deeper into the understanding of it – you have to have advanced knowledge behind the scenes. You have to be able to connect the dots.

We apply our technology to extract knowledge and insight from insurance documents, making that embedded knowledge available, and automating processes that are based on knowledge and cognitive processes - processes that some years ago people thought could not be automated. We can transform unstructured data into decisions and actions.

What benefits does it bring?
A key benefit is that we are smart from the start thanks to this embedded knowledge - so we don’t have to train the system each time with hundreds of thousands of labelled documents; we don’t have to have that learning stage – we can start from day zero and return value.  Another important point is the accuracy level we guarantee in terms of comprehension of the documents – which is one of the highest in the market. Give us a document and our system can understand it more than 90 percent of cases without any training being needed.

What do you hope people will get out of your webinar?
I would like to offer a different perspective on the reality of the concrete application of AI, and to give people a toolkit for analysing and understanding where they can start and what elements they need to take into consideration. We will also share potential ideas for use cases. Attendees will hear from people who are really applying technology with success and gain real examples of how to collect insights and use the knowledge.