While AI systems can now learn a game and beat champions within hours, they are hard to apply to business applications.

MIT Sloan Management Review and Boston Consulting Group surveyed 3,000 business executives and found that while 85% of them thought that AI would provide their companies with a competitive advantage, only 1 in 20 had “extensively” incorporated it into their offerings or processes. The challenge is that implementing AI isn’t as easy as installing software. It requires expertise, vision and information that isn’t easily accessible.

When you look at well-known applications of AI, such as Google’s AlphaGo Zero, you get the impression it is like magic: AI learned the world’s most difficult board game in just three days and beat champions; Nvidia’s AI can generate photorealistic images of people who look like celebrities just by looking at pictures of real ones.

But most business problems can’t be turned into a game; you have more than two players, and there aren’t clear rules. The outcomes of business decisions are rarely clear wins or losses, and there are far too many variables. It is a lot more difficult for businesses to implement AI than it seems.

Today’s AI systems do their best to emulate the functioning of the human brain’s neural networks but do this in a very limited way.  They use a technique called deep learning, which adjusts the relationships of computer instructions designed to behave like neurons. To put it simply, you tell an AI exactly what you want it to learn and provide it with clearly labeled examples, and it analyzes the patterns in that data and stores them for future application. The accuracy of its patterns depends on data, so the more examples you give it, the more useful it becomes.

Herein lies a problem. An AI is only as good as the data it receives and is able to interpret it only within the narrow confines of the supplied context. It doesn’t “understand” what it has analyzed, so it is unable to apply its analysis to scenarios in other contexts. And it can’t distinguish causation from correlation. AI is more like an Excel spreadsheet on steroids than like a thinker.

The bigger difficulty in working with this form of AI is that what it has learned remains a mystery: a set of indefinable responses to data. Once a neural network is trained, not even its designer knows exactly how it is doing what it does. As New York University Professor Gary Marcus explains, deep learning systems have millions or even billions of parameters, identifiable to their developers only in terms of their geography within a complex neural network. They are a “black box,” researchers say.

Speaking about the new developments in AlphaGo, Google/DeepMind Chief Executive Demis Hassabis reportedly said, “It doesn’t play like a human, and it doesn’t play like a program.  It plays in a third, almost alien, way.”

Businesses can’t afford to have their systems making alien decisions. They face regulatory requirements and reputational concerns and must be able to understand, explain and demonstrate the logic behind each decision they make.

For AI to be more valuable, it needs to be able to look at the big picture and include many more sources of information than the computer systems it is replacing.

See also: Strategist’s Guide to Artificial Intelligence  

In inventory management, for example, purchasing decisions are usually made by experienced individuals called buyers, department by department. Their systems show them inventory levels by store, and they use their experience and instincts to place orders. An AI could consolidate data from all departments to see the larger trends — and relate them to socioeconomic data, customer-service inquiries, satellite images of competitors’ parking lots, predictions from the Weather Company and other factors. Many retailers, including Amazon.com, are doing some of these things.

AI is advancing rapidly and will surely make it easier to clean up and integrate data. But business leaders will still need to understand what it really does and create a vision for its use — that is when they will see the big benefits.

The 4th District of the Court of Appeals has reversed a puzzling W.C.A.B. decision that had awarded TTD benefits beyond five years from the date of injury, ignoring the plain language of Labor Code 4656(c)(2). In County of San Diego v W.C.A.B. (Pike),<http://www.courts.ca.gov/opinions/documents/D072648.DOCX> the appellate court had little difficulty in reading the rather straightforward statutory language to firmly reverse the WCJ and W.C.A.B. decisions awarding TTD beyond the five-year jurisdictional limit set by statute. The applicant, Kyle Pike, sustained injury to his right shoulder in July 2010 while employed as a deputy sheriff for the County of San Diego. He was awarded a 12% PD benefit in May 2011. On May 26, 2015, within the five-year jurisdictional time to reopen his case, he filed a Petition for New and Further Disability seeking TTD and Labor Code 4850 benefits. He received his 4850/TTD benefits through July 31, 2015, at which time benefits were terminated. At trial, the WCJ awarded benefits on a continuing basis, determining that while Labor Code 4656 was clear regarding benefits payable within the five-year jurisdictional time frame in the statute, it was silent as to what benefits could be provided after five years from the date of injury. On reconsideration, the W.C.A.B., in a split decision, affirmed the WCJ’s award. The Appellate Court had little difficulty in seeing through the WCJ’s and W.C.A.B.’s construct: “       This interpretation of section 4656, subdivision (c)(2) is not tenable.  As discussed above, section 4656, subdivision (c)(2) clearly and unambiguously provides that temporary disability benefits "shall not extend for more than 104 compensable weeks within a period of five years from the date of injury." (§ 4656, subd. (c)(2).) Thus, contrary to the board's decision, the relevant statutory language does provide that all periods of temporary disability for which payments are made must occur within five years of date of the injury.” The court also pointed out that if the WCJ/W.C.A.B. analysis was correct, even the 104-week limitation would not exist after the five-year limitation, in effect, eliminating any limitation on TTD beyond five years while providing limitations within five years, hardly a logical result. “…Such inconsistent reasoning further demonstrates the fallacy of the WCJ's interpretation.” See also: Why WC Needs an Outcomes Strategy   The appellate court also pointed out that all of the authorities cited by the applicant attorney, amicus for applicant and the W.C.A.B. were interpretations of Labor Code 4656 prior to the amendments limiting TTD to the period within five years from the date of injury.  The court further noted that in the one decision it found where similar language was included in the statute, the appellate court had limited the receipt of TTD to within the five-year statutory time frame. The court reversed the W.C.A.B. decision remanding the case back to the W.C.A.B. to grant the Petition for Reconsideration of the Petitioner, County of San Diego. Comments and Conclusions: The W.C.A.B.’s decision in this case is at best puzzling, at worst a flagrant attempt to avoid the legislature’s clear intent. It is difficult to conceive of how this statute could be tortured into an interpretation that allowed TTD to be paid beyond the statutory limitation. The analysis by the WCJ, adopted by the majority of the W.C.A.B., was patently inconsistent and required a tortured reading of the statute to reach the final result. If the WCJ/W.C.A.B. analysis had been upheld, all an injured worker would have to do to obtain additional TTD is file a petition within five years from the date of injury and then wait till after the five-year date to claim additional TTD. Hardly a result the legislature intended and one that even the W.C.A.B. would have a hard time justifying with a straight face. A copy of the decision can be found here.

For many people, their first thought about natural disasters is the devastating property damage that is extremely visible and highlighted by the media. However, the impact of natural disasters goes far beyond property damage and includes the impact to your workforce, your supply chain and the operations of your business. During a recent Out Front Ideas webinar, we were fortunate to get the perspectives of leaders from three different segments of our industry on the impact of natural disasters on risk management. Our guests included:

• Tom Best, deputy general counsel for Home Depot
• Ryan Brannan, commissioner of workers’ compensation for the Texas Department of Insurance
• John Hinz, vice president of Vericlaim

Types of Disasters There are two basic types of natural disasters – those with warning and those without. With hurricanes and flooding, you typically have some degree of warning that allows you to initiate disaster response protocols and to prepare for the disaster. However, with events such as tornados, earthquakes and other sudden events, there is no warning and no opportunity for advance preparation to minimize the impact and maximize the response. Both types of disasters benefit from developing a disaster response plan in advance. Workplace Injuries One of the first employer concerns has to be preventing and responding to employee injuries when a disaster occurs. At Home Depot, they work with vendors on a daily basis to identify any potential weather that could affect their stores. When there is a potential event, they pull together their response team, led by a disaster captain. Their response team has functional members of all their critical business areas, including human resources, legal, supply chain and business operations. These teams meet every year before the start of hurricane season to make sure everyone understands their role and the disaster response protocols. They also connect with state, local and federal authorities to coordinate response efforts. Because Home Depot has a very important community role in disaster preparedness and response, they keep stores open as long as possible and reopen them as soon as possible. From a workers’ compensation claim standpoint, there are many concerns. Employees can be injured during the disaster itself. There is also significant potential for injuries sustained by first responders and the National Guard during the response and recovery. Texas deployed 14,000 National Guard troops in response to Hurricane Harvey, and those troops are all considered employees of the state of Texas when deployed. Traumatic physical injuries are not the only concerns. There are also occupational disease concerns because of the toxic chemicals that were in the floodwaters of Houston. Furthermore, there are concerns about post-traumatic stress. Because of the occupational disease exposure, there could be a very long claims tail from this natural disaster. Workforce Disruption Home Depot is a major employer, but they are also an essential element in any disaster response because people depend on them for building materials and other supplies. Their command center is focused on taking care of both their employees and the community as a whole. The Texas Workers’ Compensation Commission closed five field offices at various times in response to Hurricane Harvey. Their primary focus was the safety of their staff, but they were also concerned about being able to conduct the business of the commission. See also: 6 Reasons We Aren’t Prepared for Disasters   It is important to give your employees time off during a natural disaster to take care of their families and personal needs. Employers often bring in workers from other locations to assist in the affected areas so that the employees living in the area can tend to their personal needs first, then come back to work when able to do so. This allows the business to continue serving the community while also making sure that employees are settled. Workers’ Compensation System Impact Keeping the workers’ compensation system running during a natural disaster is important and challenging. In Texas, the governor suspended certain regulations and extended or tolled deadlines in affected areas to ensure that workers were receiving timely care and benefits and that carriers were focused on benefit delivery instead of bureaucratic issues. Social media was very useful in keeping people updated on when field offices were open and providing other important information to all stakeholders. Healthcare Impact One thing people do not often think about in natural disasters is the impact on the healthcare delivery system. The healthcare delivery system is disrupted in many ways:

• Technology: With electronic medical records and a wide variety of equipment powered by electricity, a prolonged period without power can make delivery of care very challenging.
• Continuity of care: Patients are often forced to treat at facilities outside their areas during natural disasters. Facilities need to not only be able to handle the influx of patients, but to deal with the potential HIPAA considerations.
• Supply chain: One impact of the hurricane that hit Puerto Rico was a significant disruption to the pharmaceutical industry, which accounts for more than 70% of Puerto Rico’s exports. There was a nationwide shortage of saline IV bags after the hurricanes, for instance, because most of these were manufactured in Puerto Rico, and those factories were shut down for a time.
• Life and death issues for patients: During Hurricane Katrina, healthcare workers and patients in New Orleans were trapped for many days without power. Providers had to make decisions around which patients to evacuate first and which patients were in such bad shape that they could not be saved.
• Litigation costs: There is always a big spike in litigation against healthcare facilities following a natural disaster because of care disruption and other challenges.

Supply Chain Supply chain is important to most businesses, and a natural disaster can significantly disrupt the normal supply chains. This was especially challenging on an island like Puerto Rico. Getting the supplies to the island was only the first step. Supplies sat for days in the ports because there were no dock workers to unload them and no trucks to deliver them. There are many lessons to be learned about disaster responses to islands after the events of 2017. On the mainland, supplies can be staged out of harm’s way in advance of a hurricane so that the trucks can start rolling in once the area is safe. Additional products are purchased in advance so there are ample supplies available. Home Depot works with local, state and federal authorities to coordinate the distribution of disaster relief supplies. Disaster Preparation Mitigating the risks and challenges from disasters takes extensive planning and practice. Every location and each facility is different and has varying needs. But as John Hinz explained, planning for emergencies can be the difference between staying in business and losing everything. There are several essential elements that should be included in any emergency preparedness plan.

• Focus on prevention: If there is any way to prevent a disaster from happening, that is your best defense. The first step in the process is to assess your risk and the potential impact to see how you can be more effective in disaster planning. Once you know the type of disasters for which you are most at risk, take steps to minimize potential damage to your facility and harm to your employees. Think of the actions you might need to take and what you would need in the event of a fire, flood, severe storm or other disaster.
• Evacuation plan: Every facility should have primary and secondary routes and exits that are well-lit, marked and easily accessible. There should be an outside area designated as a meeting place for employees to gather once they are out of the building. Staff members that may require assistance during an evacuation due to physical limitations should be noted in the plan.
• Communication: In addition to emergency contact information for local police, fire and ambulance numbers, you should have a contact list that also includes information for your customers, suppliers and distributors. This list should be updated continually, and copies kept both in your files and in offsite locations so you will be able to access them regardless of the situation. You may want to preset conference call numbers in case that is needed. Be sure you have a way to contact key players in and outside the organization.
• Protect vital company information and critical data and programs that are imperative to keep your operation running. Make sure these things are backed up and that the backup is kept in a location separate from the primary facility.
• Understand your insurance coverage: Review your insurance policies with your agent or broker so you know your deductibles and how they are applied to your coverages. You should know the limits and nature of your insurance, including coverage specifics. You may want to make changes to some policies, as all coverages are subject to limits and exclusions.
• Keep insurance information handy: The names and numbers of your insurance representatives should be kept in a safe, accessible place, as this will expedite the claims process when the time comes.
• Plan for contingencies: Despite your best efforts, your preparation may not be enough. Have an offsite location or allow personnel to work from home, if necessary, to keep the business running.

See also: Cognitive Biases and Risk Management   Final Thoughts There is no foolproof plan that will protect your organization from every disastrous situation, but you can be well prepared for most emergencies. If your company does not yet have such a plan, you can work with carriers or agents and brokers to begin the process. There are also a number of consultants that specialize in this area. After developing a disaster preparedness plan, you need to continually review and update it to make sure that it is current and that everyone understands his or her role if there is a disaster. You can listen to the archived Out Front Ideas with Kimberly and Mark webinar on this topic here.

The accident in which an Uber driverless car hit and killed a pedestrian Sunday night in Tempe, AZ, got us talking about how innovation should progress in a risky world. We worry that Uber hasn’t always been as careful as, say, Google/Waymo and GM/Cruise in testing its autonomous technology. An investigation will tell the tale soon enough in this particular instance (one of the advantages of all those sensors on driverless cars), but we’ll still be left wrestling with a world in which innovating can place people at risk, but in which not innovating might even be more dangerous. We can’t just sit still in a world where 1.25 million people die in traffic fatalities each year.

Our CEO, Wayne Allen, took up the pen to address the role we intend to play on innovating amid risk:

"I want to be super clear. We at ITL and Innovator’s Edge are going to increasingly take positions about what 'should be.' The rate at which certain technologies are being introduced into the marketplace outpaces the ability of lawmakers and even regulators to keep up, including those in the insurance world. Governmental authorities have jurisdiction over bits and pieces of the technology ecosystem, but it is rare for a single body to have complete authority over all the concerns about privacy, discrimination, consumer safety, etc.

"It is, therefore, incumbent on our industry, the folks who are supposed to manage risks of all kinds, to step forward and lead. Let’s provide innovation at the speed of reason.

"We will never advocate for the tempering innovation for the sake of antiquated processes. No. But we advocate for everyone to cinch up their chin-strap and do what’s right. 

"Consider chatbots. If chatbots are as much a commodity as many tell me they are, then anybody could roll them out to a carrier, right? Well, what if it turns out that the data transmitted to a chatbot is personal financial information or personal information about a health condition or relates to an accident and includes health information? Shouldn’t we be asking questions about how secure these chatbots are? How safe is the data being transmitted? Shouldn’t any company integrating chatbot technology in its claims process—or any customer communication, for that matter—verify that the bot and the path through which the information passes meet a standard for privacy? 

"What about autonomous vehicles? Should we be surprised if not all autonomous technology is created equal? Could one company’s technology be more secure, more heavily tested, just plain better than another? Are there standards for testing and product rollout that should be implemented? Should we as an industry simply take the position that we will not insure any vehicle (irrespective of the creative ways insurance can and will be built into the experience) that does not meet certain standards? If we do not, I am certain some enterprising politician will roll out a solution that will be the worst possible result, will likely be in line with what some lobbyist representing some corner of some industry wants, will stymie innovation and will be nearly impossible to overturn. 

"If you agree with me and want to help provide innovation at the speed of reason, for the benefit of us all, please email me at wallen@innovatorsedge.io. The table stakes need to be known, and they need to be raised."

Have a great week.

Paul Carroll
Editor-in-Chief

For several years, AIR has written about, presented about and created infographics around the problem of the “protection gap”—the difference between economic losses from extreme events and what is actually covered by insurance. For the most part, our discussions have focused on the macro-level issues of global (re)insurers and governments, catastrophe models and data. However, to learn more about the micro-level issues that may be contributing to the protection gap, I recently attempted to buy flood insurance on my townhouse condo in a suburb of Boston. As with any condo, there are two insurance policies at work here — a condo master policy and my condo HO-6 policy. In my case, there are two separate local insurance agencies. I called the agent on the master policy and asked whether I was covered for flood, knowing full well that I wasn’t. He correctly answered no, but quickly countered, are you in a flood zone? I responded no, and he appeared to be unsure as to why anyone would want flood coverage unless their bank required it for their mortgage. The same scenario played out with the insurance agent on my HO-6; she was confused about why anyone would want coverage that they wasn’t required by their mortgage lender. This time, I pushed a bit and explained that I had a friend who lived outside of a flood zone and had suffered a flood loss. That motivated her to initiate the process to sell me flood insurance. See also: The Myth of the Protection Gap   She emailed me a form that looked photocopied and asked me to fill it out, which I had to do by hand. Filling out forms by hand is one of the activities that I find most irritating in life, as my handwriting is absolutely illegible, even to me. Thankfully, the necessity of actually handwriting anything has declined dramatically in the past few years, but the technological changes that have made that possible have apparently not reached the independent insurance agency system in the U.S. But I had bigger problems than my handwriting. Many of the questions that were being asked on this form were ones I didn’t know the answer to, even after working at AIR for 10 years. The questions I easily knew the answers to were occupancy, year built, whether I had a basement and whether my house was elevated. Construction type was also an easy one for me, but I’m not sure that would always be the case for the typical customer. Here are some of the questions that I and the average consumer would find challenging:

• Foundation: slab or pilings
• Type of pilings: wood, concrete, driven or poured
• Base flood elevation
• Lowest floor elevation
• Post-firm or pre-firm enclosure (luckily, I don’t have an enclosure)

At AIR, we’ve been talking about the protection gap and how to close it at a very macro level, but this exercise gave me a more practical feel for why the protection gap exists at all—including in developed countries. The ability to properly quantify flood risk does exist, from AIR and other modeling firms. And as we know all too well, there is currently enough capital to insure the risk. But those capabilities break down somewhere within the insurance value chain. As my own experience illustrates, there is a lack of willingness of the front-line insurance distribution system—in this case, insurance agents—to push the coverage. To the extent customers are aware of the risks and ask their agents to buy flood or earthquake coverage (a rare situation, to be sure), the process of getting this coverage is cumbersome and antiquated, to say the least. On this point, there are a slew of venture-backed startups dedicated to making the insurance purchase “mobile first,” but to date these startups appear to be more focused on auto, renters or lower-value contents coverage, and have not yet made inroads into streamlining the process of purchasing flood or earthquake coverage. See also: Future of Flood Insurance   Closing the protection gap will require a concerted effort on the part of every player in the insurance value chain—from agents, to carriers, to reinsurers and to those of us in the modeling industry. I also believe it will require technology that identifies those who need coverage and places that coverage in a seamless way, as the status quo doesn’t appear to be doing the job. Getting the entire insurance value chain on the same page to make the necessary investments to close the protection gap is easier said than done, of course, but that’s the real problem that needs to be solved if we are serious about closing it.

Throughout this series, I have stressed the importance of connecting workers’ comp patients with great doctors at their time of need and how this results in an average of a 10% reduction in claims costs. I’ve also talked about the role of PPOs and EPOs. Next up: outcomes strategies for each jurisdiction. Workers’ compensation is determined state-by-state, requiring any national workers’ comp company to adhere to variations in the laws. Some states provide excellent control over care, while others allow so much choice that it can seem like there is no strategic use for an outcomes-based network. States use three main categories of jurisdictional control: panel (employee has to select from your list of doctors), care direction (you tell the employee who to use) and employee choice (employees choose their doctors on their own). The graphic below provides a visual breakdown of each state’s model: Despite what you might think, outcomes can play a meaningful role in any of these jurisdictions. Let’s discuss each model in more detail. Panel States In a panel state, employees are required to choose their doctors from a list created by payers or employers. These can range from exclusive provider networks (EPOs) to simple panels. The employee could be mandated to use doctors on the list for the life of an injury (as in the California MPN), or the designated doctor could maintain control for at least a meaningful portion of the claim (e.g., Pennsylvania workplace panels). The important consideration is that the employee can typically select anyone on your list. There is no way to manage around average doctors in your network, so it is in your best interest not to file your entire PPO as your exclusive provider listing. If you do, your overall average outcomes drop to the average across all doctors included, and you will have little recourse if injured workers choose the worst doctors on your list. For this reason, it is critical to stay on top of who’s included. Remove poor performers to ensure that any provider on that list can be trusted at any time. See also: 2018 Workers’ Comp Issues to Watch   My best practice recommendation is to “right-size” your EPO by first predicting how many doctors are necessary in each coverage area (typically about a 15-mile radius). Once you know your needs for appropriate coverage, then pull only that number starting with the best and working down the ranking. The tighter you keep participation, the more impact you will see on claims outcomes. A well-constructed model can leverage provider outcome and drive overall claims costs down by 10% or more. The downside of this model is the level of effort required to maintain your list. You should keep inclusion/exclusion criteria documented up front, so you have a structured plan for whom to keep and whom to remove. You also need to check in at least quarterly to remove anyone whose performance has dropped or who has moved or closed his or her practice. Typically, you are handing over a list and letting the employee select a provider. If there is a problem (such as a provider closing the practice and not notifying you), you probably won’t realize you have an issue until you hear from the injured worker’s lawyer. Key strategic driver: Focus on the providers in your listing at all times and keep the list tight. Care Direction States In a care direction state, the insurer or payer maintains the exclusive right to tell the injured worker which doctor to use. The biggest difference over a panel model is that providers can be selected in real time. This makes maintenance a lot easier because you don’t need to “right-size” the network and anticipate needs. What you do need is an integrated view of your networks that includes your scores. In care direction states, it is critical that the search engine your team uses to direct care shows the scores of the doctors. Any time a patient needs a doctor or specialists, you want your team selecting the best available provider proximate to the injured worker — not just the closest one in network. When outcomes as a measure first matured to the point where it was widely adopted, most people focused on using outcomes to select their networks for panel states. However, the care direction model actually has a higher potential ROI because you are always selecting the best provider at the time of need. In a panel state model, when I hand over a list, the average quality of all the doctors on the list will drive the impact on your claims population. In care direction, I am always selecting the best, which drives impact to the quality of your best doctors. In addition, because the scoring updates in real time, you are always making your care decision on the latest information at the time of need, not on the information available when you built your list. Also, your best strategy is to not limit yourself to your existing network. If you are looking for a top neurosurgeon in a rural area, and there are not any good options in your PPO, you should look for a high-scoring doctor outside your network before you send the patient a longer distance to an average doctor in your PPO. As I explained earlier in the series, the outcomes savings will almost always be more than the PPO discount. Key strategic driver: Always go for the best doctors available, keeping your scores integrated with your list. Employee Choice States Many people falsely believe that outcomes-based control strategies cannot work in a state that allows the employee to select whomever they want. This is not true. Soft channeling is a strategy that involves making targeted recommendations throughout the cycle of the claim. For example, selection of the doctors that show up on the workplace poster should be based on the best doctors nearby — not just the closest. Using outcomes scores when creating workplace panels helps encourage good choices without stepping on the rights of the employee. Another strategy for the employee choice model is to integrate scores into your billing workflow. If you see an initial consult from a poor-performing ortho, you have about two weeks before your claimant goes under the knife. If you can operationalize this strategy, you have the opportunity to reach out to the injured worker with a crafted message such as: “We typically recommend a second opinion in cases like yours; there is a great specialist not far from you. Would you like me to set up an appointment?” See also: The State of Workers’ Compensation   The point is to not interfere with the injured worker’s choice while taking every opportunity to provide great options for the person to consider. If your scoring model is focused on quality-of-life outcomes such as less time away from work or lower disability rates at the end of care, you are providing a valuable service to that injured worker when you steer the person away from a doctor who has high infection rates and a high incidence of failed surgeries. Certainly, the potential ROI is not as significant as in other states, but you only need a small number of injured workers to choose a better doctor to recoup typical costs of implementing these strategies. Wrapping Up There are methods and strategies to leverage outcomes in all jurisdictions. Some jurisdictions are easier than others, and your ROI and level of effort will vary, but don’t lose sight of why this is important: Better care improves everyone’s outcomes. Better outcomes improve not only the bottom line but also the quality of life for people who get injured on the job. Better care results in happier and more productive employees, fewer costly complications, less legal friction and, ultimately, a better workers’ comp system. Throughout this series of articles, I’ve covered multiple aspects of outcomes-based networks, including important considerations and strategies. These are recommendations and best practices pulled from my experiences. You don’t have to hit a home run every time. If you simply get 5% to 10% of the claims away from bad doctors and into the care of great doctors, you will have significantly better results overall. Please reach out with questions and feel free to comment with ideas for additional articles on the future of workers’ comp. As first published in Claims Journal.

In Part 1 of this blog series, we shared how a microservices architecture can bring value to a constantly changing business environment. In this segment, we will share our views on the benefits of a microservices architecture for the insurance industry.The traditional insurance value chain and subsequently the customer experience over the last two-plus years has operated across a number of functional silos. Each of these silos has unique characteristics and is organized around its own KPIs. For example, underwriting/risk management is focused on risk assessment, underwriting quality and booking of premium. Claims is focused on claims management, fraud detection, leakage reduction and processing turnaround time. Typically, each of these functions operates with limited organizational integration. This organizational design is several decades old and inspired by Henry Ford's assembly line innovation that optimizes processing costs through specialization and automation. This design has served insurance well during the industrial and information ages, leading to specialized functional IT systems targeting the business needs of the silo functions from underwriting to policy management, billing, claims and more. The core system components are integrated as a suite. The components and suite serve well-defined functions that have had less dynamic integration needs. This organizational model and IT system landscape has been effective for decades. But as we enter the digital age, the painful and expensive business and IT modernization projects over the last decade, coupled with portals and complex integrations to these core systems to improve agent and customer experience, do not align with new market needs. Today’s insurance landscape demands agility to adapt with ease, innovation to reimagine the possibilities and speed to capture the opportunities. The digital age demands so much more to stay relevant and competitive. Customer Experience is the Differentiator Customer experience is front and center in differentiating insurers in the digital age. It is a key factor in driving higher customer acquisition and retention, which, in turn, drives growth. Customer experience is much more than offering a better user interface. It is about the customer journey that creates a unique, compelling and engaging experience that makes it “easy to do business” with insurers. Customer journeys must cut through functional silos, which are currently optimized for internal operational efficiency. These silos, however, as customer journeys are changing, are now contributing to the degradation of the customer experience. To design and refine customer journeys in today’s digital age, insurers will need to collect siloed capabilities into a new virtual capability designed to optimize the customer journey. This new virtual capability will require hyper-integration and micro-granularities of system capabilities to achieve the desired result. See also: Insurance Hasn’t Changed, but… (Part 5)   Insurance Value Chain Disrupted As we highlight in our Future Trends 2018: Catalyzing the Shift to Digital Insurance 2.0 report, the insurance value chain is rapidly shifting to adapt to new business models, innovative products, expanded distribution channels, new competition with entrants from outside the industry, elevated customer expectations and emerging technologies. Digital transformation is redefining the value chains and each component. New products such as on-demand products, connected products and micro-insurance are reshaping business assumptions and fundamentals. We are seeing innovative product design that uses new sources of data, new risk assumptions, micro-segmentation, expanded services, new customer engagement approaches and new channels to reach customers. These designs leverage new technologies such as artificial intelligence (AI), cognitive, analytics and microservices. The result is the disruption of the insurance value chain. With the value chain disrupted, the underlying systems must be disrupted, too. Rise of Ecosystems and the Platform Economy As we enter the digital age, the blurring of traditional industry boundaries is seeing the rise of ecosystems and the platform economy. Companies like Apple, Alibaba, Google, Amazon and Facebook are at the forefront of this shift. They are using an ecosystem with connected services from different parties to create a seamless customer experience. An ecosystem is the DNA of the platform economy, enabling a business model to exchange and share value among its partners and customers. To meaningfully participate in the platform economy, insurers must embrace ecosystems and be prepared to partner with competitors, other industries and innovative technology-based service providers. ZhongAn, an online Chinese insurer, generated 70% of its 2017 car insurance premium in one month (January 2018) by using AI and big data with the ecosystem, including carmakers, dealers, after-sales service providers and lenders that created market reach and a loyal customer base. The ecosystem approach eliminates traditional industry or organizational boundaries in designing products and creating a new customer journey. However, it necessitates the need for a flexible and granular system composed of different services running on different technology platforms that can easily integrate with any ecosystem. A New System Paradigm for the Digital Age A common theme is emerging that highlights the need for a new set of capabilities to support the paradigm shift. To succeed, let alone survive, insurers will need to respond to the value chain disruption, elevated customer expectations and the rise of the ecosystem and platform economy by using granular (single responsibility principle) API/microservices to build an on-demand business solution with loosely coupled microservices and find-n-bind capabilities that can leverage any ecosystem. A microservices architecture enables the building of new capabilities to meet these needs. The graphic below contrasts the anatomy of a traditional “pre-digital age” monolith insurance app and a “digital age” innovative microservices-based insured app. Today’s monolith insurance systems, although partially accessible through APIs, are built as a large deployable monolith unit. This architecture does not easily adapt to the rapid pace of change because the change is to a large-system single codebase and specific localized API. A separate API layer exposed over the single-monolith code base makes it difficult to integrate with ecosystem partners as well as making it extremely complex to orchestrate services across various systems or apps. In contrast, a microservices architecture decomposes a large unit into fine-grained single purpose, self-contained and independently deployable business services that enable the ability for rapid change and open the possibility of multiple change deployments daily instead of waiting for the periodic release cycle. Using microservices across various apps, insurers can orchestrate a composite user interface that is a tailor-made customer journey. It can be enhanced quickly based on customer feedback. The graphic below shows how a microservices architecture can assist in the design of a unique customer experience using a product offering and ecosystem.  Multiple customer journeys can be assembled by orchestrating functional microservices and ecosystem services available outside the insurer enterprise. The times are changing. And it is exciting! The ability to leverage powerful microservices architecture to build a new foundation for the digital age of insurance is game-changing. It will enable new business models, new products, refined customer experiences and timely responses to new business needs (in hours and days instead of months and years), and it will help insurers remain relevant and competitive. While microservices is exciting and will accelerate the industry’s ability to innovate, it is not the Holy Grail. The smaller, focused services have many advantages but also create complexity in the orchestration of those services. Employing best practices in designing microservices size and data model sizing is critical. Most importantly, determining the gradual transition to microservices rather than a big-bang approach will help insurers build a platform that can withstand the test of time and constant change to help insurers participate in the digital age and platform economy with agility, innovation and speed. See also: It’s Time to Accelerate Digital Change   In Part 3, we look forward to covering our views on best practices in introducing and scaling microservices within the world of the monolith IT system environment. We encourage you to read our thought leadership, Cloud Business Platform: The Path to Digital Insurance 2.0, to gain a deeper insight on these topics. Please share your views on this exciting topic in the comments section. We would enjoy hearing your perspective. This article was written by Manish Shah and Sachin Dhamane.

Following on from last week's post on Regulation, it is time for our final 'Key Theme', on Product Development, after which we move on to our Regional Profiles. Product development is the bread and butter of the insurance industry – the question is not whether to build products but rather which products will best serve the needs of today’s increasingly demanding insurance customers. What has become clear across this content series is that product development can no longer occur in silos, with one function creating products for another function to sell. Insurers are determined to make every moving part of their business serve the customer, and what this means in concrete terms is that every division of the business has a contribution to make toward the creation of customer-centric products. The line between product developers and product salespeople, as in so many industries, is therefore becoming blurry. In this sense, Product Development is the central node into which feed all our other 'Key Themes' (Analytics & AI, Digital Innovation, IoT, Marketing & Customer-Centricity, Distribution, Claims, Fraud, Cybersecurity, Investment Management and Regulation). And all the indicators suggest that insurers are only just beginning to tap the opportunities for technology-driven product creation.

"The biggest problem these days is, although everybody is talking about Digitisation and Disruption and are modernising their core systems with huge investments in order to support these new trends, the underlying products are not yet 'digital ready' – even new ones." — Oliver Lauer, former Head of Architecture / Head of IT Innovation at Zurich

In this installment, we start by looking at where innovation is occurring (by insurance line and insurance department) before considering some of the broad trends in product development – from bundling and upselling to diversification and Usage-Based Insurance (UBI). The following stats and perspectives are taken from our Global Trend Map; a full breakdown of our survey respondents, and details of our methodology, are included as part of the full report, which you can download for free at any time. So first up: which departments is product innovation currently coming from? ... The three highest-scoring departments were Underwriting (with 70% of respondents naming it), Pricing (55%) and Marketing (54%). Other areas that warrant mention were Actuarial (51%) and Distribution (36%). We noted in our earlier post on Marketing & Customer-Centricity that the roots of today’s consumer-led disruption are in the rise and ease-of-use of new distribution channels – so insurers that leave Distribution outside of their product discussions do so at their peril! See also: Next for Insurtech: Product Diversity   Which insurance lines are driving the greatest degree of product innovation? In addition to seeing product as department-driven, we also investigated the extent to which it is line-driven. The chart below shows Auto (voted by 56% of respondents), Home (45%) and Health (41%) to be the three lines experiencing the most product innovation (according to carriers taking our survey). This is corroborated anecdotally by the sheer number of in-market IoT products we see across these fields, from in-car telematics through to smart-home controllers and connected-health armbands. Life and Commercial are relative laggards in this regard, although we do believe there is ample opportunity in both these areas. This may follow the same pattern we identified with IoT (itself an abundant source of product innovation), where we saw platform implementation in Commercial currently trailing but quickly drawing level with other lines (see our earlier post on IoT). Regional trends for this question warrant some high-level comment:

#smarthome #mhealth #AutonomousVehicles #telematics See where insurance innovation is greatest in the #InsuranceMap: https://t.co/QH8z3aYH7y pic.twitter.com/qj6KKBrUF3

— Insurance Nexus (@InsuranceNexus) July 26, 2017

Health is a substantial driver of product development for carriers in Asia-Pacific and Europe. In North America, however, it appears relatively insignificant

1. Life appears tagged to Health in terms of how it trends regionally and is consistently the least innovative of the lines#
2. P&C/General, Commercial and Home insurance are relatively consistent across our different geographies
3. Auto is ahead among carriers in Europe and Asia-Pacific compared to North America

The relative prominence of the different lines, as well as the broad regional tendencies outlined above, remain the same when we widen our scope beyond carriers to consider the industry as a whole.

"The biggest risk the insurance industry faces when it comes to innovation is not taking enough risk. True innovation requires experimentation, which most of the time results in failure. Insurance organisations are built to eliminate failure from their culture. Without failure, you can have no innovation. CEOs demand a positive Return on Investment (ROI), they now need to seek out and understand what it means to have a positive Return on Risk (ROR)." — Steve Anderson, President at The Anderson Network

[caption id="attachment_30597" align="alignnone" width="570"] Human hand pointing at touchscreen in working environment at meeting[/caption] Key trends in the development of products It’s clear from this section so far that product development is a strategic priority for a diverse spread of departments and lines. But how are insurers actually going about product development on the ground? Let us now present our trends on a number of product approaches that we identified among our carrier respondents: product diversification, Usage-Based Insurance (UBI) and product bundling/upselling.

"You don’t really have to invent new products in my eyes, you just have to make the existing ones easier and more 'digital native'. Today’s products have been and are still created for non-digitals. And this situation not only makes new customer-facing digital processes complicated, it also makes core replacements and automation more complicated and expensive than necessary." — Oliver Lauer, former Head of Architecture / Head of IT Innovation at Zurich

i) Product Diversification The pace of change in the insurance sector is picking up, and many ecosystem players are quickening their iterations both on new and existing products in a bid to stay relevant. 76% of (re)insurers are pursuing product diversification as part of their organisations’ growth strategies… This move towards diversification is not limited to any one kind of insurance market but is driving insurance growth all around the world. It is naturally very important for insurers looking to break into totally new markets, as is the case in many developing economies, where traditional insurance products may be inappropriate for lower-income demographics (microinsurance being a case in point). In mature markets, growth will come primarily from addressing existing demographics with more tailored products to fill in under-penetrated lines and segments; there is also the perennial threat that existing customers, tired of products that are suboptimal, will churn to competitors and new entrants who can offer better-suited ones. Whatever the market conditions, by trialling multiple products, insurers can both broaden their appeal and arrive at optimal products more quickly. ii) Usage-Based Insurance (UBI) The emerging generation of insurance products differ from those that have gone before in several key ways. One is the on-demand or usage-based component of new products – so that, rather than having idle and inflexible policies that subject the policyholder to the Tyranny of Averages, consumers can enjoy insurance-as-a-service. 32% of insurer respondents have a Usage-Based Insurance (UBI) strategy… This trend towards Usage-Based Insurance (UBI) is a global one, and there was no significant regional variation across our three key regions. More information about UBI models can be found in our earlier Internet of Things section and in our section on Europe in the upcoming Regional Profiles. "Most of the innovation in product development will happen where smart connected devices drive new business models based on behavioural data. I particularly expect improvements in pricing. UBI is a bigger game changer than covering events that are not insured today. In any case, there will be a huge need to understand, measure and manage rational and irrational behaviour." — Andreas Staub, Managing Partner at FehrAdvice iii) Product Bundling and Upselling A key trend we have witnessed in other B2C industries, like Amazon and Netflix, is the rise of the recommendation engine. This is not just a core component of the customer experience but also an important enabler of new business, insofar as it lets companies push new products to customers that they are actually likely to want and to find useful. We also see this in insurance, in the form of data-driven product bundling and upselling. 47% of insurers have a strategy to bundle and upsell products based on customer lifestyle analytics… Insurance has always had relatively few customer touchpoints, and even though insurers are now seeking to increase that number, insurance remains a product that is sold rather than bought. Taking full advantage of every selling interaction they have is therefore the surest way for insurers to increase their customers’ lifetime value. Bundling and upselling products is – like diversification – a strategy with strong applications regardless of what sort of insurance market you are operating in. Whether your focus is to chase new customers or to retain existing business, it is better to reap maximum reward on each customer from the outset than to re-engage them (perhaps unsuccessfully) later on. "The cost of customer acquisition is a critical metric for marketing efforts. Lifetime customer value also helps us know how much we should be spending to acquire as well as how we should expand our share of wallet within the products we offer." — Michael Shostak, SVP and Chief Marketing Officer at Economical Insurance See also: Insurance Product Development (Excerpt, Part 3)   There are obviously many aspects of product development beyond diversification, UBI and bundling/upselling which we were not able to investigate directly. For this reason, we asked our carrier respondents to indicate additional product-development talking points in open text. Two key points that emerged across their comments were:

1. The need to increase the service element of products, to keep up with evolving consumer needs and to drive customer retention
2. The problem of constrained development resources, especially in expert personnel and IT

Now that we have worked our way through all our Key Themes, it's time to move on to our Seven Regional Profiles, exploring on a geographical basis the stats and perspectives presented in the series so far. If you'd like to access all our Regional Profiles straightaway, then please feel free to skip ahead and download the full Trend Map here (it's free!).

In programming circles, there is an aphorism known as the “90-90 rule.” It states that the first 90% of code accounts for the first 90% of the expected development time—and the remaining 10% of code takes another 90% of time. The rule is a tongue-in-cheek acknowledgement that technology projects always take longer than you expect, even when you know that they are going to take longer than you expect. Sacha Arnoud, director of engineering at Waymo, recently used a variant of the 90-90 rule to characterize Waymo’s self-driving car program. Waymo’s experience, he said, was that the first 90% of the technology took only 10% of the time. To finish the last 10%, however, is requiring 10x the initial effort. Arnoud’s remarks were given at a guest lecture at Lex Fridman's MIT class on “Deep Learning for Self-Driving Cars.” He offered technical insights on the history of the Waymo program, how it is applying artificial intelligence and deep learning and how it is moving from demo to industrial-strength product. The Waymo engineer’s lecture goes beyond most Waymo management presentations and press events. He provides vivid details on the complexity of the effort to date and insight on challenges to come—both for Waymo and for those trying to catch up to its pioneering efforts. Here are 5 takeaways, though I recommend watching the entire presentation. 1. Industrialization requires 10x the effort. Arnoud emphasized the large amount of work needed to go from a demo that works in a lab to an industrialized product that is safe to put on the road: “You need to 10x the capabilities of your technology. You need to 10x your team size, including finding effective ways for more engineers and more researchers to collaborate. You need to 10x the capabilities of your sensors. You need to 10x the overall quality of the system, including your testing practices.” 2. Deep learning enabled algorithmic breakthroughs. Arnoud noted that deep learning techniques were much less advanced in 2010 when Google started its work on self-driving cars. But, in the years since, deep learning has advanced to enable algorithmic breakthroughs in several critical areas for autonomous driving, including mapping, perception and scene understanding. Arnoud gave numerous examples, such as using deep learning to analyze street imagery to extract street names, house numbers, traffic lights and traffic signs. The ability to precompute such data and store them as maps in the car saves precious onboard computing power for real time tasks. See also: When Will the Driverless Car Arrive?   Deep learning is driving breakthroughs in real-time tasks as well, such as analyzing sensor data to identify traffic signals, other vehicles, obstacles, pedestrians, and so on. Deep learning capabilities also help in anticipating possible behavior of other drivers, cyclists and pedestrians, and driving accordingly. 3. Synergy with other Google units is key to Waymo’s progress. Arnoud acknowledged the importance of Google’s “whole machine learning ecosystem” to Waymo’s progress. This includes the seminal software advances by the Google Brain team and on-going collaboration with other Google teams working on deep learning at scale, such as in vision, speech, natural language processing and maps. The Google ecosystem also provides specialized infrastructure and tools for machine learning. This includes accelerators, data centers, labelled datasets and research that support Google’s TensorFlow programming paradigm. 4. Waymo’s testing program might be its secret sauce. Arnoud emphasized that however great Waymo’s algorithms, sensors and overall package might be, driverless cars are still complex, embedded, real-time robotic systems that must work safely with imperfect data in an unpredictable world. He highlighted Waymo’s three-prong testing program of real-world driving, simulation and structured testing as key to iterating on and productizing the technology. Much is made of the millions public-road miles that Waymo’s cars have driven autonomously. Arnoud described this as the equivalent of about 300 years of human driving experience and 160 times around the globe. Real world driving is critical, he said, but what is more important is the ability to simulate. Simulation is critical because it allows for Waymo to test each new iteration of software against all previously-driven miles. Even more important is the ability to test against “fuzzed” versions of those millions of miles, such as seeing how the software would handle cars going at slightly different speeds, an extra car, pedestrians crossing in front of the car and so on. Arnoud described Waymo’s simulation-based testing capability as the equivalent of 25,000 virtual cars driving 2.5 billion real and modified miles in 2017. The third component of Waymo’s testing program is its structured testing program. Arnoud said that there is a “long tail” of driving situations that happen very rarely. Rather than trying to encounter every possibility in real-world driving, Waymo set up a 90-acre mock city at the decommissioned Castle Air Force base where it can test its cars against such edge cases. These tests are then fed into the simulation engine and fuzzed to create variations for more testing. 5. Waymo’s next steps are big (and hard) ones. Arnoud closed with a discussion of the engineering challenges in front of Waymo. He described two big next steps. One next step is expanding the “operational design domains” (ODD) of the cars. This includes expanding into “dense urban cores,” such as San Francisco (in which Waymo recently announced it is expanding its testing program). The other ODD was additional weather conditions, such as hard rain, snow and fog. (Waymo CEO John Krafcik recently told an audience that he was “jumping up and down” recently when it snowed 12 inches near Detroit, because it would enable Waymo’s testing in snow.) See also: 7 Steps for Inventing the Future The other area of focus was what Arnoud called “semantic understanding.” As an example, he pointed to the chaotic Place de l'Étoile traffic circle around the Arc de Triomphe in Paris. The circle is a meeting point of 12 roads and notoriously difficult to navigate. Arnoud says he has driven it many times without incident, however, and that such situations require a lot more than perception and vehicle operating skills. They require deep understanding of local rules and expectations. They also require constant communication and coordination with other drivers, including signals, gestures and so on. This kind of deep reasoning is key to numerous edge cases and improving the general abilities of driverless cars. * * * While Waymo has clearly made tremendous progress towards the driverless future, Arnoud closed his presentation by emphasizing the engineering infrastructure and the complexities of scaling that have to be addressed in order to turn driverless cars into safe production systems. How far along is Waymo in the last 90% of that industrialization process? Arnoud never said. But, to put a point on the complexities, he showed a closing video of a Waymo car stopped at an intersection as a gaggle of kids bounced on frogger sticks across the street on all sides of the car. Some things are waiting for, he seemed to imply.

There are a lot of reasons why negotiating a workers' comp settlement at a mediation is better than at the board, such as control over scheduling and lack of time constraints. But how does that compare with an informal? The presence of the mediator makes all the difference. Here are three examples: Stop the Posturing When opposing counsel sit together they keep their cards close to the vest. They magnify the strengths of their own case while denigrating the opposing viewpoint. Once I separate the parties into separate rooms (caucuses), the motivation to aggrandize diminishes. Attorneys and their clients can reasonably discuss the good and bad sides of the issues with the mediator without giving up their negotiation position. Then it’s my job to convey that position so that another reasoned discussion happens in the other room. See also: Work Comp Mediation in a #MeToo World   The Neutral Sounding Board As a professional neutral, I do not have a stake in the outcome. I want to help the parties reach a settlement, which is the optimal result for all. I can provide untainted feedback and sometimes point out overlooked data. Some clients refuse to listen to their lawyer’s case assessment, and some lawyers have learned that their continued employment mitigates against contradicting their client’s overconfidence. As mediator, I can deliver an unwelcome message about the prospect of success, opening the door to a more frank discussion between attorney and client. Stakeholders Have Active Roles Unlike at the WCAB or an informal, the stakeholders are encouraged to take an active part in mediation. In caucus, clients and their attorneys can have frank discussions with the mediator—and each other. When claims adjusters and risk managers attend mediation, they maximize their understanding of the dynamics of the negotiation. Injured workers can receive settlement offers in real time. Some applicant attorneys keep their clients away from the negotiation in an informal meeting. The injured worker may be hidden in a back office or on telephone stand-by. There may be important reasons to prevent interaction between the defense attorney and the injured worker, but this approach prevents the injured worker from buying in to the negotiation. Sometimes, the result can be disastrous when the injured worker later repudiates that carefully crafted compromise and release. <See also: 25 Axioms Of Medical Care In The Workers Compensation System   In contrast, once the parties are separated into their caucus rooms, the stakeholders, their attorney and the mediator can have a confidential, free-flowing discussion without the presence of opposing counsel. It may be the only time the injured worker gets to tell the story to a neutral. The neutral intermediary is missing in an informal. My job is to steer the proceedings, frame communications to facilitate the negotiation and help parties decide their course.