AI Applications: an Overview of 8 Emerging Artificial Intelligence Use Cases

Introduction

Artificial intelligence (AI) is the all-encompassing term for technologies built and developed to automate tasks previously performed by humans. Back in 1950, Alan Turing published his famous paper about the possibility of thinking machines¹. However, up until the early 2000s, humanity hadn’t made much progress on AI for various reasons. Some of the biggest reasons for that were lack of computer processing power, funding, and actual data necessary for machines to learn about the real world².

However, that’s been changing for the past couple of decades, with the rapid adoption of the internet, a trillion-fold increase in processing power since the 1950s³, and the explosion of technologies that collect and catalog data⁴.

With this rapid pace of innovation, AI became a viable business tool. The AI market is on track to outpace many industries. The business value of AI is poised to reach over 3 billion dollars within the next couple of years⁵, and many companies are trying to cash in on this opportunity.

This report was created to highlight industries that have the potential to experience or are already experiencing technological transformation brought on by AI. We selected the banking, healthcare, and insurance industries for this research to expose a good variety of AI-driven innovations and focus on industries that are ripe for disruption and are already making steps towards AI ubiquity. For example, industries like manufacturing have been experimenting with AI for quite some time now, and so such an exposé for manufacturing might not be as revealing.

Many market players haven’t yet caught up with the pace of change. However, it’s evident that market leaders in the selected industries are already reaping the benefits of artificial intelligence, machine learning, natural language processing, image recognition, and other AI technologies. In this report, we’ll review some of the specific use cases of artificial intelligence in daily life of various industries.

The business value of AI is poised to reach over 3 billion dollars within the next couple of years, and many companies are trying to cash in on this opportunity.

Disrupting the Banking Sector with AI

It’s difficult to expand the already existing growth of opportunities in banking. The market is saturated with the growing pool of fintech companies planning to provide banking products on different, sometimes better terms than banks do.

Banks carry a heavier regulatory burden, which changes constantly and in some cases stifles innovation⁶. Legacy systems that banks employ are also dragging down their growth, while fintechs have the opportunity to start with an already competitive tech stack. Nevertheless, many of the banking leaders already apply AI applications to achieve incredible productivity and efficiency gains.

1. Anti-Money Laundering and AI

The global footprint of money laundering is staggering, as up to 5% of the global GDP is laundered annually⁷, which amounts to almost 4 trillion dollars⁸. That’s why governments around the globe are tightening regulations to combat this trend. EU’s anti-money laundering directives or the USA Patriot Act¹⁰ are prime examples of such regulations.

These regulations pressure banks by increasing the compliance burden. But money laundering has other risks, which banks have to deal with.

Money Laundering Burden and Risks

Reputational

Involvement in money laundering operations can seriously tarnish a bank's reputation and negatively impact future business opportunities or client preferences.

Regulatory

Failure to comply with regulations can halt operations and result in fines, asset forfeiture, and other negative effects on the business.

Operational

Anti-money laundering (AML) operations require significant financial and talent investments.

Technological

Increased diversity and complexity of banking interactions, carried out through multiple channels, further complicate the necessary practices to combat money laundering.

Artificial intelligence has the power to influence the efficiency of anti-money laundering programs directly. Only 2%¹¹ of all suspicious transactions result in a crime. From 2012 to 2016, the number of suspicious activity reports rose by 2000%¹². Banks have to use AI to keep up with the growing number of threats and elevate the relatively low success rate.

One of the most comprehensive approaches to solving this problem is building an intelligent system, which combines AI-powered transactions monitoring and the bank’s Know Your Customer (KYC) framework. This creates a comprehensive view of banking activities and provides more context for money laundering detection. Unlike rule-based systems, AI can detect anomalous behavior a lot more effectively because it can identify patterns that lie outside of the usual purview of an AML system.

Banks that use such systems already can see incredible ROI. For example, Florida’s QuantaVerse essentially automates 70% of investigation routines¹³ and provides complex, AI-generated reports for investigators, allowing the bank to minimize the false positives rate. There is still a human in the loop, but the process is much more streamlined and accurate.

A system combining AI-powered transactions monitoring and the bank’s KYC framework can identify user behavior patterns that lie outside of the usual purview of an anti-money laundering system.

2. Risk Management with AI

Artificial intelligence is becoming a quite lucrative risk management tool for banks, especially given the abundance of robust financial datasets. The opportunities are endless as practically any banking line of business has use cases, which can be dramatically improved through advanced predictions, generated by AI.

Risk Management Opportunities for AI in Banking

Credit Risk

AI can help banks identify the best credit allocation opportunities using not just the application data, but also the transactions or business history of the prospect. These solutions also include loan servicing tools that save time and minimize human error for the lender.

Churn Risk

Banks can use AI to identify clients at-risk of churn, based on their transactions and banking activity data.

Collections Management

Collections department can use debtors’ personal data and allow AI systems to identify people more likely to repay their debt and even identify channels optimal for engaging them.

Mortgage Risks

Banks can use mortgage repayment data, combining it with clients banking activity, to identify mortgages at risk of default and prepayment.

Fraud Detection

AI systems can be used to identify fraudulent transactions, using sophisticated anomaly detection algorithms.

Investment Risks

Banks can use the loan, investment, or mortgage data to identify risky opportunities or portfolios.

Model Risk Management

Advanced AI systems are capable of monitoring model performance and identifying models with deteriorating performance.

The examples of these innovations are abundant. JPMorgan uses COIN, a machine learning system that analyzes commercial credit data. The system saves the company hundreds of hours of manual labor¹⁴ and minimizes credit processing errors.

Customer churn prediction is a popular AI use case among financial institutions, especially given the rising competition from other banks and fintech startups. The beauty of AI apps in this niche lies in the fact that they offer intimate insights about the specific churn reasons. This data can be used to drive online service improvements and personalization¹⁵.

Artificial intelligence is becoming a lucrative risk management tool for banks, especially given the abundance of robust financial datasets.

Looking to build an intelligent banking app?

Get a quote

Improving Healthcare Outcomes with AI Apps

Healthcare is a booming industry, which is used to the spotlight attention of global corporations that create artificial intelligence. There are many reasons for this. Healthcare is a public good. Creating AI for healthcare is good PR. Healthcare organizations collect tons of data and are driven by scientific discovery that relies on data. Healthcare is a challenging industry for AI, as the available data is complex by default. That’s why even the biggest names in AI have failed here before¹⁶.

However, companies that successfully deliver artificial intelligence applications for healthcare are in for incredible revenue opportunities, akin to being the first to develop a life-saving vaccine.

Many of the leading healthcare organizations are exploring and investing in AI. The applications are practically endless, as long as there’s data to work with.

AI Opportunities in Healthcare

Image Processing

Advanced AI systems, based on deep learning algorithms can improve diagnostics by analyzing MRI, CT, and other image data.

Treatment Optimization

AI can detect irregularities and even predict treatment effectiveness by analyzing electronic healthcare records (EHR) data of patients with similar conditions and treatment regimens.

Operational Optimization

Hospitals can use AI to improve the forecasting accuracy for the daily/weekly flow of inpatients as well as outpatients and optimize staffing needs accordingly.

Service Improvements

Caregivers can use AI to predict escalation of care with a high degree of accuracy, which allows them to optimize undertriage/overtriage metrics and improve the overall quality of care.

Readmission Risk Management

AI can predict patients more likely to be readmitted, which allows hospitals to adjust their treatment, improve care, and reduce avoidable readmission rates.

Disease Analysis

Artificial intelligence can be used to analyze disease progression, as well as deliver disease propensity insights based on EHR data. This information can aid in treatment, and even in marketing efforts.

3. Medical Image Processing and AI

This is one of the biggest niches in healthcare AI, poised to become a $2 billion industry in the next couple of years¹⁷. Although medical imaging is not a new application of AI, we decided to place it into the emerging technologies list because of the incredible growth that it’s been experiencing over the past couple of years. Since 2014 over $1.2 billion dollars¹⁸ have been invested in artificial intelligence in this domain.

However, that’s just a footprint of the whole market. The leaders in this domain are already building effective AI systems for image recognition.

UC San Diego developed an AI system that can diagnose eye diseases with a 95% accuracy, just like an experienced ophthalmologist¹⁹.
Samsung’s AI technology that detects breast lesions in ultrasound images is said to have increased the diagnostic accuracy by 5%²⁰.
Nvidia made the SDK for its Clara Compute Platform available to the general public. It was already used by a tandem of hospitals in the Boston area to develop an algorithm that recognizes abdominal aortic aneurysms²¹.
The leading educational institutions create separate research units to explore AI in medical imaging, like the Stanford’s Center for Artificial Intelligence in Medicine & Imaging²².

These and other developments will push AI in medical imaging to the next level. But for now, there are still many obstacles and issues that AI software developers in this domain have to overcome. For example, the interpretability of models that make predictions is an issue since most of these models are based on deep learning algorithms, which are almost impossible to explain.

Medical image processing is one of the biggest niches in healthcare AI, poised to become a $2 billion industry in the next couple of years.

4. Disease Analytics with AI

Being able to predict disease progression and propensity allows healthcare organizations to intervene promptly and make a difference for any type of disease. Machine learning can effectively predict certain diseases with a >90% accuracy rate²³.

Some of the Benefits of AI in Disease Analytics

Improved Treatment

The ability to start the treatment earlier improves clinical results and treatment effectiveness.

Optimized Cost

Early disease predictions can dramatically costs associated with the treatment.

Improved Payer Experience

Payers get better visibility into the future costs of treatment and can save on them thanks to a potential improvement of treatment effectiveness.

Improved Marketing

Patients who are identified as having propensities towards certain diseases can be identified early and marketed to.

iQuity uses a proprietary machine learning technology to predict autoimmune diseases, identifying patients almost a year in advance of traditional methods.

Scientists at the Francis Crick Institute created an AI system that better predicts the risk of death from heart disease, compared to humans. The AI system used hundreds of health and healthcare variables to make the final predictions, while doctors used about two dozen²⁴.

Mount Sinai hospital uses artificial intelligence to predict kidney disease²⁵, which can affect a host of decisions and variables in the treatment process, from decreasing the need for dialysis to drug adjustments in the treatment plans.

Unlearn.AI is using unsupervised machine learning to predict Alzheimer’s progression. Based on their findings, a demo version of a clinical trial simulator²⁶. Early diagnosis of this disease could save over $7 trillion in medical costs in the long run²⁷.

These and many other developments in healthcare AI are the reason why ABI Research estimates the saving brought on by its adoption in the healthcare sector can reach over $50 billion in 2021 alone²⁸.

Machine learning can effectively predict certain diseases with a >90% accuracy rate.

Managing Insurance Risks with AI Applications

The insurance market for AI is growing at a very rapid pace, as legacy approaches to insurance predictions are being disrupted by various technological advances. The industry has a history of working with data²⁹, which is the foundation of any modern AI system. Insurance is also famous for hoarding data about its customers to improve various operations, like underwriting, claims management, and reserving. That’s why AI is on track to revolutionize the insurance market with use cases popping up in all parts of the insurance lifecycle.

According to a survey by SCOR, these are the operational areas that will experience the biggest transformation due to artificial intelligence³⁰.

Apart from these generalized areas of interest, there are specific insurance use cases that can be greatly augmented by artificial intelligence and machine learning systems.

Major AI Opportunities in Insurance

Insurance Lead Scoring

Use AI to identify the most profitable insurance applications that are more likely to convert. This allows you to optimize application review and focus on the most lucrative customers.

Underwriting Automation

Automate the process of application review by using AI to identify policies eligible for automatic acceptance.

Claims Acceptance Automation

Offload your claims management team by using AI to identify claims that could be automatically marked for payout.

Price Optimization

Use AI to find the right price equilibrium, which is competitive, profitable, and which improves the likelihood of a policy conversion.

We’ll focus on the areas that could potentially experience the highest impact from artificial intelligence: claims management and underwriting (based on the Insurance Nexus study, cited above).

5. Claims Management and AI

Claims management is a resource-heavy process in the insurance lifecycle. It directly influences the profitability, retention, and reputation of the insurer. In the UK alone, the average daily payout on auto insurance claims reaches £33.3 million (≈$43.4 million)³¹. Insurers are presented with an opportunity to optimize costs associated with these payouts or maybe even optimize the payouts themselves through better claim insights. AI apps can deliver fascinating results in claims management.

Automated claim classification (with AI) can be 30% more accurate than manual processing and has the potential to increase claims management productivity by 80%³². This can be used to identify urgent claims, claims with the highest payout potential, claims with fraud risks, etc. An AI system for such classification can potentially reach an almost 100% accuracy. The biggest bottleneck in the process is human error, which leads to inaccurate data being used for classification³².

In healthcare, insurance claims classification with AI can lead to almost perfect case prioritization, improved quality of healthcare outcomes, and lower expenses for the payer³³.

Fraud detection is another important aspect of claims management, especially given the limited resources that insurance companies are working with. SIUs are a precious resource. You only want to engage them on cases most likely to be fraudulent. Insurance fraud in the US alone costs $40 billion annually³⁴, and getting ahead of such fraudulent claims can be a game changer for insurers. The same AI-powered classification systems described above can be calibrated to classify claims likely to be fraudulent and allow insurers to dispatch the most appropriate resources to investigate in a timely and effective manner.

AI-powered claim classification can be 30% more accurate than manual processing and has the potential to increase claims management productivity by 80%.

6. Automating Underwriting through AI

Underwriting is a manual process that includes a variety of repetitive tasks, which, by default, means it has the potential to be automated with artificial intelligence. The modern insurance customer is not going to wait for application approval but instead will go to a different insurance company. So instead of wasting underwriting resources on rubber-stamping generic applications, AI should be used to identify policies that can be automatically accepted or, on the contrary, require an underwriter’s review. And many insurers are already betting on digital AI applications to disrupt their legacy approaches to policy management and pricing.

In healthcare, AI is expected to completely automate the underwriting processes for 60% of applications in the near future³⁵. The accuracy of these AI systems got to the point where insurers are issuing policies to traditionally risky and uninsured customer categories, like people with chronic diseases and non-citizens.

These potential productivity gains are why the underwriting AI market is already getting its "unicorns", like Planck Re, which received $12 million in its round A of funding³⁶.

However, there are companies offering much more exotic tools for underwriting. One of the examples is EagleView that uses artificial intelligence to catalog and identify property details, which can be used in daily life to refine the underwriting process and ultimately fine-tune pricing for property insurance. Such image recognition technology could also be used to assess damages after major insurance incidents, like hurricanes and floods.

In healthcare, AI is expected to completely automate the underwriting processes for 60% of applications.

Cross-industry trends

In this section, we will cover some of the trends that are emerging not only in the markets that we have highlighted in this report but in other industries as well.

7. Augmented Analytics

If we take a look back at some of the use cases in this report, it becomes obvious that you need people to generate insights and operationalize them. All of the available data science specialists and AI engineers can’t possibly cover the demand for machine learning and AI.

That’s why Gartner predicts³⁷ that augmented analytics (AA) will be one of the biggest trends in the coming years. AA means that data scientists, engineers, and business analysts will be aided by automated systems that can build, deploy, and update machine learning models automatically. This will save thousands of person-hours and take productivity to the next level.

These capabilities have the potential to transform how healthcare, banking, and insurance companies are exploring AI and iterating on their findings. Some of the major products in this niche are offered by Dataiku, H2O Driverless AI, RapidMiner, DataRobot, and AWS SageMaker.

Augmented analytics will be one of the biggest trends in the coming years, and it will enable data scientists, engineers, and business analysts to build, deploy, and update machine learning models automatically.

8. Robotic Process Automation and AI

Robotic process automation (RPA) has been a growing trend for quite some time now , even before AI became an enterprise buzzword. The difference is that RPA copies routine human actions, while AI copies human intelligence . Think of RPA as a robot that runs the production and AI as a system that runs the robot. However, the future lies in combining RPA and artificial intelligence to create diverse and highly autonomous systems to tackle a variety of business problems, from process automation to customer interactions.

That’s why companies that deliver these synergic solutions are materializing in many industries. For insurance, it’s startups like GuideWire and Shift Technology. Companies like WorkFusion offer AI-powered RPA solutions for the healthcare space. Capgemini singles out customer experience management as the most impactful domain for RPA and AI blend in banking.

The future lies in combining RPA and AI to create diverse and highly autonomous systems to tackle a variety of business problems.

On a Final Note

AI has become an integral part of modern business processes. While transforming everyday life, there are examples of artificial intelligence revolutionizing the way people address challenges in various areas.

The heavily regulated banking sector can use the power of artificial intelligence to increase the effectiveness of anti-money laundering investigations, thus being more accurate in identifying fraudulent activities and getting enough time to prevent them. AI also broadens the banks’ risk management capabilities by allowing them to predict credit risks, customer churn, collection, and investment risks with higher precision.

Owing to great volumes of data processed in healthcare, the industry is to benefit immensely from AI. From treatment and operational optimization to readmission risk management—the scope of healthcare domains that AI consultants can foster through intelligent technologies is impressive. Medical image analysis is one of the fastest developing technologies and it’s able to improve the diagnostics outcomes significantly. At the same time, healthcare providers can use disease analytics to excel in predicting diseases, thus getting the opportunity to foresee them or cut their progression.

When AI comes to the insurance area, it can foster such domains as claim management, underwriting, fraud prevention, customer services, sales, and marketing recommendations. The greatest thing about AI in insurance software is that it brings automation to various resource-heavy and time-consuming processes, which allows insurers to reduce human effort while improving the insurance outcomes substantially.

Apart from several industry-specific AI applications, there are at least two increasingly popular AI domains that will influence various spheres. While augmented analytics (AA) is being developed as a new paradigm of automated data preparation and sharing, robotic process automation (RPA) coupled with AI can become a new turning page in the history of autonomous systems that will likely accompany or even replace humans in a variety of business processes.