Tag Archives: ai

Covering for a Gap in Workers Comp Data

What happens when a key data source becomes less available, reducing carriers’ ability to evaluate risk? This has happened during the pandemic in workers’ compensation.

In workers’ compensation, OSHA is one of the top data sources that underwriters use. In particular, underwriters will look at OSHA inspections and violations to measure some aspects of the risk. 

Here at Convr, our focus has been to help carriers with the right insights at the right time for better decision-making, and we found, using a vast data pool, that planned inspections dropped 48% in 2020.

One reason is fewer claims; as operational capacity was reduced or suspended for many industries in 2020, workers’ compensation claims dropped by over 20%. As accidents declined, inspections that normally would have followed weren’t needed. In addition, OSHA reduced the number of planned inspections for the safety of their inspectors.

The reduction in inspections has led to a lack of reliable information for workers’ compensation carriers to evaluate businesses — but this is where technology comes in. With the help of artificial intelligence and advanced analytics, carriers can still determine the risk of a business by looking at past patterns.

These past patterns include types of structured and unstructured information that data scientists refer to as “features” in machine learning models. Often, significant features are high-dimensional nonlinear combinations of company and property characteristics, such as the size of the business, the year it was established and prior violations. Other features include social media information and product and services data.

See also: 9 Months on: COVID and Workers’ Comp

Applying AI to our data lake, which is informed by over 2,000 data sources, Convr has determined that, in place of the normal volume of OSHA inspections, carriers can use a workplace safety model to accurately quantify risk. A workplace safety model consists of a machine learning model that predicts how safe a workplace will be based on OSHA data and the different data sources mentioned above.

Companies labeled as the riskiest 10% by Convr’s proprietary workplace safety risk scoring model observed three times as many future violations as those labeled as the median risk.

COVID-19 has proven that circumstances can change unexpectedly, and carriers have to become adaptable and flexible enough to implement alternative solutions to minimize the impact. Advanced AI models, like the one Convr has created to quantify workplace safety, hold tremendous value for carriers, enabling them to better understand risks even when traditional sources of information are limited or unavailable.

When armed with technological advancements such as these, carriers are equipped with the right tools for better decision-making and optimal underwriting results.

What’s Wrong With Commercial Auto?

Commercial auto is one of the biggest, most problematic cost centers in the insurance industry. Its loss ratios are incredibly high — and growing — particularly in comparison with other P&C lines. Carriers struggle to make money on the line, and they frequently incur fairly substantial losses.

With fewer claims than most other lines, why is commercial auto such a problem? Because claims are much larger; carriers may be looking at paying out $200,000 vs. $20,000 on different types.

Let’s examine the issues and see how we can turn commercial auto into a revenue generator instead of a black hole.

Commercial Auto by the Numbers

According to the National Association of Insurance Commissioners, commercial auto represents $39 billion in earned premiums. Yet, it pays out approximately $28 billion in losses, which excludes all of the internal administrative costs. Some folks in the industry thought that losses might go down this year due to fewer cars on the road, given COVID-19, but this has not been the case. One large carrier recently shared with me that they’ve seen a similar level of claims (both in terms of frequency and severity) because many commercial trucking organizations are working nonstop to keep the supply chain going.

If even a pandemic won’t lower the loss ratio, are we doomed? Let’s break this down a bit further.

Of the $28 billion in losses, roughly 60% went to medical costs associated with bodily injury. There is not much that can be done here from a carrier standpoint, short of offering slightly lower premiums for companies that incorporate the strictest safety standards and tools like electronic logging devices. Even this technology can be a double-edged sword given that the more advanced vehicles tend to be more expensive to fix in the event of an accident. While measures can sometimes minimize the severity of injury, accidents still happen.

Carriers have seen an uptick in accident rates involving bodily injury or death over the past 10 years, in part because more commercial vehicles are on the road traveling more miles each year. Then there is the role of drivers themselves. The trucking industry has a remarkably high turnover rate. For example, in the third quarter of 2019, large truckload carriers’ turnover rate increased to an annualized rate of 96%, according to Trucker.com. As companies basically overhaul their entire workforce each year, many don’t invest the resources into in-depth training programs, leading to less experienced, more accident-prone commercial drivers on the road.

Driver issues withstanding, the non-medical costs still leave $11.6 billion in losses. Where does it go? Very simple: Legal costs account for 40% of losses; the system is broken.

Losing Battle

Litigation tends to happen when claims are not resolved quickly or when something is perceived as unfair. While such concern is perhaps understandable, in practice it is not so innocent. A growing body of attorneys are ready to sign on to “help” plaintiffs, initiating cases that should never be filed, and they have become quite proficient at securing huge settlements.

The lawyers who specialize in auto claims know exactly what to look for and can be quite convincing in wooing potential clients and later in threatening the carriers with which they are attempting to negotiate. For example, the lawyers might spot something not related to the specific incident but that could be tied to the company. A savvy attorney would tack on additional charges, such as negligence, on top of bodily injury, pain and suffering. Tactics like these drive settlements higher. Complicating matters, states have very different statutes on bad faith suits filed by attorneys. Some states are more prone to settling cases early, and, as a result, lawyers in these areas increase the number of suits they file — which increases the cost for carriers.

One thing that many insurance firms and self-insured enterprises are just realizing is that plaintiff attorneys are rapidly becoming data-aware and using that awareness in a highly sophisticated and strategic way. Once upon a time, plaintiff attorneys were good at qualifying clients that they had a high degree of confidence could return them a large settlement. While that’s still the case, in the last half decade or so, even moderately sophisticated plaintiff firms have compiled significant datasets on enterprises and insurance companies, in many cases down to the general actions taken at an adjuster level. They use this data to plan their litigation strategies, select the most effective partner and manage each step of the process in an intelligent way. The result is that carriers and self-insured enterprises that do not have similar data-savvy practices are essentially being bled dry because they have nothing to counter this advantage.

See also: The Digital Journey in Commercial Lines

Attorneys are more than willing to try their hand in court, and juries can be quite sympathetic to plaintiffs they feel were wronged by a big company. As a result, there has been a certain degree of social inflation, as jury awards can rise astronomically if for no other reason than a desire to help the little guy fight back against the “evil” corporation — and winning verdicts keep going up. According to Shaub, Ahmuty, Citrin & Spratt, the median of the top 50 single-plaintiff bodily injury verdicts in the U.S. nearly doubled from 2014 to 2018 (moving from $27.7 million to $54.3 million).

As it stands today, there is tremendous variability in jury awards, just as there is with out-of-court settlements. Looking across claims, there might be very little difference in the facts of the case, yet one plaintiff walks away with millions while another receives a much smaller verdict. Carriers are often unwilling to risk the chance of coming out on the wrong side, hence agreeing to a settlement that may be uncalled for.

It is clear that litigation is the most significant hurdle to better loss ratios across the commercial auto line. If we can reduce, standardize or eliminate costs associated with litigation, the industry would be in a much better position.

A Process Evolved

New technologies, artificial intelligence (AI) and machine learning, in particular, can help. For example, you may want to know the likelihood of attorney involvement based on several claim factors, or you want to know which attorney is involved and what kind of settlements he or she negotiates for similar claims. But most importantly, you want to know what actions to take to prevent attorney involvement. AI and machine learning applications are emerging that can identify claims early in their life cycle that need the most attention.

Imagine how powerful it would be if you had an application that inherently understood the intricacies of commercial claims — one that would warn you of claims that were in danger of slipping to an attorney. Solutions are hitting the market that leverage capabilities like natural language processing and deep learning techniques to analyze hundreds of data points hidden within claims. They now can tap into structured data as well as the really interesting unstructured data, like notes or police reports, as well as decoding the sentiment of claimants. This collection of data provides pretty telling clues as to how a claim might progress.

An adjuster could get an alert about aspects of the claim that are troublesome, access to detailed attorney scores and ratings in case the claim escalates and, if need be, the optimal time to settle for a favorable outcome. With this information, the adjuster could take immediate action to head off the problem.

As applications get smarter, they will be able to determine what a claim settlement should look like and why with a much higher degree of certainty based on similar claims. This can be instrumental in the adjuster’s or defense counsel’s ability to negotiate with the claimant’s counsel. Armed with this kind of hard data, the organization could walk into settlement negotiations in a much stronger position and begin to counter the formattable data advantages that plaintiff attorneys have been amassing.

AI and machine learning systems also help organizations close claims faster, and, in doing so, relieve some of claims management teams’ administrative burdens. Additionally, by closing claims quickly and fairly, claimants receive settlements faster, return to their everyday lives sooner and thus generally wind up in a better financial position without ever involving an attorney.

Considering the high loss ratios of commercial auto insurance today and the propensity for them to increase further, emergent AI-based applications are our best hope for improving profit margins and repairing the commercial auto line.

As first published in Claims Journal.

5 Things to Know When Integrating AI

In my 30 years in the industry, I have seen several great advances in claims processing, but nothing as exciting as AI. 

The benefits of applying AI can be transformational, not only when it comes to cost control and claims acceleration, but also in providing an exceptional customer experience. 

Several players are promoting AI technologies now — whether it’s for auto or property, underwriting or claims – which means it’s important for claims professionals to be able to see past the AI buzzwords and evaluate solutions against solid criteria. 

According to MIT Technology Review, 40% of AI companies don’t actually have operational AI to apply. As a result, it can be challenging to see which provider actually has a functioning solution, versus one that aspires to do the work in time. 

Here’s my recommendations of what to look for when evaluating an AI platform for claims processing — is the AI making predictions that drive business value or are they just nice-to-haves?

1. When the AI processes photos it has never seen before, does it make accurate predictions? 

An effective AI system can provide a recommendation in real time on a photo it’s never seen before. A quick way to weed out potential providers is to ask for a live demo and submit your own pictures. Grab random pictures from Google or from your own historical claims to really test the waters. Did the AI accept the image? How long did it take to process? What was the confidence of the recommendation? Most importantly, is the vendor comfortable with the approach or coming up with an excuse to dodge this particular exercise? These results will separate a seasoned AI platform from ones that are closer to beta. 

2. How accurate are the results against my existing data? 

Now that you have seen if the AI works – you have to grade it against your historical figures. Ask for a calibration period where you send past photos from your own operations and ask for the AI’s recommendations. Use this as a “report card” to grade the accuracy of the technology. 

3. Do you have an open platform in which to deploy your AI? 

An open platform is more advantageous now and in the future. It allows for flexibility and will make scaling easy by adding, upgrading or swapping in the best vendors and their software for your organization. An open ecosystem also helps in reducing cost by avoiding vendor lock-in and increasing competition to get the best technology out there. The ability to integrate technology with your existing stack and processes will enable each system to “speak” to each other and give a greater picture of your claims operations. Examples of open claims platforms include MitchellGuidewire and Duck Creek. If your claims management system is home-grown, it is also an open platform. 

See also: How AI Can Tackle Claims Staffing Gap

4. Will the AI be adopted by the end-user? 

While the concept of AI can sound daunting to those on the ground, they should see the value that will enable them to focus on other duties and be much more efficient. A good AI provider should be easy to use, with a seamless implementation process across regions. 

5. How will my customers benefit? 

Submitting a claim is already stressful for your policyholders – not including the possibility of being without a car for weeks, or even months. A core key performance indicator (KPI) when evaluating an AI solution is, “How soon can my customer submit a claim?” Be sure to ask how an AI can benefit the customer experience. The concept of generating an instant claim shouldn’t be far-fetched – insurers are already doing it.

How AI Will Define Insurance Workforce

Prior to COVID-19, the U.S. boasted historically low unemployment and a roaring economy. Nearly every industry was expected to face a severe talent shortage within the next 10 to 20 years. But then March hit, and the world turned upside-down.

Since then, the pendulum has swung in the other direction. The current unemployment figures are reporting as many as 10.7 million people are out of work, and, despite this sudden abundance of available workers, staffing issues remain — they’ve just become more complex.

To navigate this wildly fluctuating environment, companies will rely on data for decision-making about hiring, training and countless other matters that affect the bottom line. This will require tools, like artificial intelligence (AI), to make sense of data and to adjust quickly amid uncertainty.

The best way to examine AI’s value in today’s uncertain world is to look at how it can work within a specific industry. Doing so makes it possible to show practical applications from which lessons can then be applied to other industries.

Commercial Insurance: A Case Study

Like other industries, commercial insurance faced a significant hiring crisis pre-COVID-19. The average claims adjuster remained in the industry for just four years — about the time it takes to gain full expertise — and those workers who stuck with claims have inched closer to retirement. So, this multibillion-dollar market is at risk of losing much of the human brain trust that enables current systems to run, as new workers cannot be hired, trained and retained fast enough to balance the scales.

Fast forward to today. Commercial insurance looks markedly different. The types and volumes of claims are changing. For example, claims related to COVID-19 contact or work-from-home circumstances are rising quickly, as are post-termination claims, while traditional claims have dropped.

At the same time, access to traditional healthcare has been in flux. To combat the limitations on available providers, telehealth solutions have exploded, opening up a whole new set of providers that claims reps need to become somewhat familiar with to facilitate claims accordingly — claims that bear a greater potential for fraud and litigation, which cost companies millions of dollars each year.

In short, almost everything about claims operations has changed — and, like many other industries that have been traditionally slow to adapt to new challenges, commercial insurance faces real hurdles.

The Importance of Data and Intelligence

Data is the key to overcoming dramatic changes within a relatively static industry. Maintaining a pulse on what’s happening across a business, or with a specific claim, and how it relates to things experienced previously is important; spotting trends early is vital. Organizations require data to determine if their plans and practices are working — and, if they are not, data should be used to drive intervention and adaptation.

But thousands to millions of data points alone won’t save the day if an organization doesn’t have the capability to understand what the data is telling them. What is the context? How are points connected? If a trend continues, what will be the effects six months or two years from now?

AI systems unlock the meaning of data to make it useful, pinpointing where organizations need to make adjustments. In commercial insurance, AI could allow for expanding provider networks to offer better, faster access to care. To actually expand networks using quality providers, systems need to tap into more data to learn which providers have achieved the best outcomes on which types of cases.

What is particularly exciting about implementing AI in this rapidly changing environment is that interpretations of data are not fixed. Machine learning capabilities are constantly refining and updating insights so that organizations — and their people — can respond accordingly.

See also: How AI Transforms Risk Engineering

Designing the Future Workforce

So, if data analytics and AI become staples in modern business, how do they solve the human resource problem? What do they mean for the future workforce? The answer is threefold.

Data determines what your hiring needs actually are: In a world that is changing so quickly, your business might not need as many people specialized in a certain area, whereas new opportunities or divisions may emerge. Your business may be forced to alter its offerings to match customer needs. Data is the guide; it lets you home in on exactly what skills are required.

AI guides training: Because AI is able to analyze so much data so quickly, new hires are able to access the information and prompts they need to do their jobs well as soon as they need it. There is not as much feeling around or dependency on senior colleagues. This is not to discount the value of experience, but it means that workers can reach a competent level much faster; what they lack in experience and intuition is replaced by data-driven insights and standardized practices.

AI augments jobs: AI solutions take care of many of the rote tasks workers are routinely bogged down with today. As a result, employees can focus on making more efficient, informed decisions; they can actually use their brains more. AI flags potential errors or problems so that they can be addressed before they escalate. Reps can focus on delivering compassion at a time when people need it most.

While COVID-19 has fundamentally altered the future workforce, tools like AI help get it back on track. In leveraging it effectively, organizations will become nimbler and more responsive to conditions while employees are more knowledgeable and effective.

Beware the Dark Side of AI

Within the Biden administration’s first weeks, the Office of Science and Technology Policy has been elevated to a cabinet-level position. Biden has appointed Alondra Nelson as deputy director. She is a scholar of science, technology and social inequality. In her acceptance speech, Nelson shared, “When we provide inputs to the algorithm, when we program the device, when we design, test and research, we are making human choices.” We can expect artificial intelligence (AI) bias, ethics and accountability to be more significant issues under our new president. 

The financial services industry has a long and dark history of redlining and underserving minority communities. Regardless of regulation, insurers must take steps now to address the ethical concerns surrounding AI and data. 

Insurers are investing heavily and increasingly adopting AI and big data to improve business operations. Juniper Research estimates the value of global insurance premiums underwritten by AI will exceed $20 billion by 2024. Allstate considers its cognitive AI agent, Amelia, which has more than 250,000 conversations per month with customers, an essential component of its customer service strategy. Swiss Re Institute analyzed patent databases and found the number of machine-learning patents filed by insurers has increased dramatically from 12 in 2010 to 693 in 2018. 

There is no denying that AI and big data hold a lot of promise to transform insurance. Using AI, underwriters can spot patterns and connections at a scale impossible for a human to do. AI can accelerate risk assessments, improve fraud detection, help predict customer needs, drive lead generation and automate marketing campaigns. 

However, AI can reproduce and amplify historical human and societal biases. Some of us can still remember Microsoft’s disastrous unveiling of its new AI chatbot, Tay, on social media site Twitter five years ago. Described as an experiment in “conversational understanding,” Tay was supposed to mimic the speaking style of a teenage girl, and entertain 18- to 24-year-old Americans in a positive way. Instead of casual and playful conversations, Tay repeated back the politically incorrect, racist and sexist comments Twitter users hurled her way. In just one day, Twitter had taught Tay to be misogynistic and racist. 

In a study evaluating 189 facial recognition algorithms from more than 99 developers, the U.S. National Institute of Standards and Technology found algorithms developed in the U.S. had trouble recognizing Asian, African-American and Native-American faces. By comparison, algorithms developed in Asian countries could recognize Asian and Caucasian faces equally well.

Apple Card’s algorithm sparked an investigation by financial regulators soon after it launched when it appeared to offer wives lower credit lines than their husbands. Goldman Sachs has said its algorithm does not use gender as an input. However, gender-blind algorithms drawing on data that is biased against women can lead to unwanted biases. 

Even when we remove gender and race from algorithm-models, there remains a strong correlation of race and gender with data inputs. ZIP codes, disease predispositions, last names, criminal records, income and job titles have all been identified as proxies for race or gender. Biases creep in this way. 

See also: Despite COVID, Tech Investment Continues

There is another issue: the inexplicability of black-box predictive models. Black-box predictive models, created by machine-learning algorithms from the data inputs we provide, can be highly accurate. However, they are also so complicated that even the programmers themselves cannot explain how these algorithms reach their final predictions, according to an article in the Harvard Data Science Review. Initially developed for low-stakes decisions like online advertising or web searching, these black-box machine-learning techniques are increasingly making high-stakes decisions that affect people’s lives. 

Successful AI and data analytics users know not to go where data leads them or fall into the trap of relying on data that are biased against minority and disadvantaged communities. Big data is not always able to capture the granular insights that explain human behaviors, motivations and pain points. 

Consider Infinity Insurance, an auto insurance provider focused on offering non-standard auto insurance to the Hispanic community. Relying on historical data, insurers had for years charged substantially higher prices for drivers with certain risk factors, including new or young drivers, drivers with low or no credit scores or drivers with an unusual driver’s license status. 

Infinity recognized that first-generation Latinos, who are not necessarily high-risk drivers, often have these unusual circumstances. Infinity reached out to Hispanic drivers offering affordable non-standard policies, bilingual customer support and sales agents. Infinity has grown to become the second-largest writer of non-standard auto insurance in the U.S. In 2018, Kemper paid $1.6 billion to acquire Infinity. 

Underserved communities offer great opportunities for expansion that are often missed or overlooked when relying solely on data sets and data inputs. 

Insurers must also actively manage AI and data inputs to avoid racial bias and look beyond demographics and race to segment out the best risks and determine the right price. As an industry, we have made significant progress toward removing bias. We cannot allow these fantastic tools and technologies to enable this harmful and unintended discrimination. We must not repeat these mistakes.