There is no question that the enterprises that successfully leverage AI’s potential will be the ones to get ahead. While businesses are more and more certain about the goals they want to use AI for, how to execute and deploy it successfully remains a difficult question.

Enterprise AI – evolution not revolution

In today’s world, the dynamics and efficiency of operations are crucial factors, and data is the key to flourishing in the rapidly changing business reality. Data provides insights from various areas of business activity. Solutions based on artificial intelligence or machine learning not only support the analysis of huge amounts of data, but also provide a new approach to optimizing and automating core processes. Enterprises across the full spectrum of industries can build competitive advantage based on AI. Such a strategy need not come down to one-off AI implementations. Rather the most successful will take a holistic approach to treating data and high-end technologies as core organizational assets. From this perspective, enterprise AI can be considered as the next, natural stage in the digital transformation of enterprises.

New approach to AI team augmentation 

Many companies want to implement the enterprise AI approach, but don’t know how to go about it successfully. Companies complain that they don’t have the necessary in-house knowledge or skills, and building a team from scratch is too time-consuming and risky. In fact, the effective implementation of AI projects requires building interdisciplinary teams – ones that command extensive expertise in available technologies and approaches to data analysis, while at the same time possessing a thorough understanding of the business processes taking place in the enterprise. They must be able to recognize the factors that influence the core processes. As business challenges go, this one certainly qualifies as daunting.

The answer to such a challenge may be to combine in-house and outsourcing models. In one respect, an in-house team provides crucial know-how and knowledge about their company’s operations. The outsourcing partner, meanwhile, provides cutting-edge knowledge of AI technology and a flexible, innovative approach to realizing the use cases. Ultimately, the vendor’s team becomes an equal member of the client’s team. Such an approach seems to have a number of disadvantages, as the knowledge of core business processes somehow goes beyond the enterprise. However, the benefits of this approach far outweigh such disadvantages. It might go so far as to say that truly remarkable results are not possible without combining the in-house and outsourcing models. At least not without a great deal of time and much larger financial outlays. Cooperation with a reliable outsourcing partner ensures flexibility, cost control and it also brings a fresh perspective to the data being analyzed.

deepsense.ai’s holistic approach to enterprise AI transformation 

The unique approach and work methodology developed by deepsense.ai maximizes synergy. Our cooperation with the in-house client team always begins with workshops that help map and understand the business needs. Joint ideation sessions define the main processes, data sources and use cases that can become part of the implementation. The deepsense.ai team then begins to review and analyze the data in close cooperation with the customer’s team on a daily basis. This model was implemented in cooperation with our client, the retail research leader Nielsen. Working alongside Nielsen’s experts across the globe maximized the synergy between our two companies. Together we developed an advanced AI solution to automate extraction of content images of varying lighting conditions, viewing angle and quality. After implementing the solution we conducted technical workshops for Nielsen’s team to transfer the know-how and code. We also provided assistance in cloud readiness and scaling.

Summary

The dynamic development of technology makes it difficult for companies whose core operations are not related with hi-end technology to keep up with the latest solutions. Close cooperation with a technology partner allows them to maximize the potential of state-of-the-art technologies and focus on the industry-related aspects of building competitive advantage.

Innovative leaders from the insurance industry are starting to recognize the benefits of machine learning and deep learning applications. As business approaches to using AI-based solutions evolve, concerns about the transparency and explainability of AI models are also diminishing. While the trust in this type of solutions is growing, AI offers a new look at the product and service portfolio, as well as improve the sales process itself to best suit customer needs. As predicted by McKinsey experts, the technological revolution in the insurance industry has already started, and its progressive development will lead to a massive technological shift over the next decade.¹

AI-based customer service automation

One area where insurance leaders look to implement AI solutions is customer service automation. Chatbots and biometric identification are becoming permanent features of the customer service landscape.
Chatbots automate customer service, increasing customer experience and satisfaction. Behind the scenes, automation significantly reduces customer service workloads. As insurance industry leaders have indicated, over 90% of customer service engagements follow basic patterns, leaving excellent potential for automation while retaining high customer satisfaction.
Biometric identification supports explicit or unnoticed identity verification within remote channels. This can include voice identity verification in call centers or typing manner verification in online channels. Similar solutions are poised to gain traction, particularly in the post-covid world.

Back-office optimization

Another area where AI-based applications are highly effective is back-office optimization.  The documents and forms processed in the insurance industry are voluminous, so improvements in this area quickly bring noticeable results. Given that, deepsense.ai developed an insurance claims assessment system for one of the largest financial institutions in Eastern Europe.  Employing advanced computer vision algorithms, the system recognizes car parts and classifies which are broken, assesses the severity of the damage and estimates the cost of repairs. It allowed nearly real-time assessment of claim value based on image documentation sent by the client. With 70% of claims handled automatically, the manual work required in the process is significantly reduced, while the time needed to assess damage falls from days to seconds.

deepsense.ai claims assessment system

Automating text and image documentation analysis can be widely applied in insurance, including for cash operations, trade finance, insurance application processing as well as classifying incoming emails to go to the appropriate department or customer segmentation.

Customer insights

Applying deep learning to customer analytics makes it easier to combine insights from various data sources (e.g. transactions, online banking logs, call center interactions) as well as a 360^o customer view and personalized sales activities. This helps to better understand customers and to build personalized recommendations, making the business more responsive and efficient. For one client in the banking sector, deepsense.ai designed a machine learning model that created personalized recommendations, showing what should be proposed to a given client based on the type of customer they are. The model had a 93% recall rate, which is 60% more effective than traditional product recommendation techniques.
Thanks to accurate AI algorithms, we can also improve customer retention by predicting churn probability. This is important as customers often churn without obvious warning signs. The roots of deepsense.ai’s experience with anti-churn use cases go back to the banking industry. We developed an intelligent anti-churn system based on historical internal data and external databases for a leading financial institution in Central Europe. To identify risk groups that were more likely to churn, we leveraged multiple data sources and gradient boosting trees and also evaluated customer lifetime value. The results were used to prioritize communication with highlighted clients, 50% of which confirmed they were considering leaving the bank.

Insurance risk management

Machine learning enables nearly fully automated fraud detection, adapted to individual patterns and changing behaviors. It can be applied in areas where a large volume of events needs to be analyzed in real time. Our client, a leading CEE insurance company, suspected that some end customers were abusing access to private healthcare. deepsense.ai was tasked with analyzing data in a search for anomalies and spotting the data-marks of fraudulent transactions. With the knowledge it gathered, the team developed algorithms identifying common schemes and techniques of private healthcare abuse. The schemes included excessive medical diagnostics and exploiting flaws in billing systems. The team also identified potential fraud being committed by service providers abusing their agreement with the health insurance company. The model we delivered spots suspicious activities and has enabled the company to reduce losses by over 3M EUR annually.

Summary

A wide range of ML applications is increasingly being used to solve real business problems in insurance. As AI becomes more popular, these applications will become the market standard. Nevertheless successfully leveraging the new opportunities offered by cutting-edge technologies will require insurers to undertake a shift in corporate governance. However, if applied successfully, AI-based solutions will benefit both insurers and clients alike.

¹ https://www.mckinsey.com/industries/financial-services/our-insights/insurance-2030-the-impact-of-ai-on-the-future-of-insurance