When companies study what they need to perform advanced research and analytics, Hersh explains companies traditionally have to go through the procurement process and wait four to six weeks to get a server; they would then have to set up the server and test out what is needed.

"In the AWS world, that's not even an issue," he says. "You do that in minutes with a credit card. You pay for it when you use it and you don't pay a dime when you're not using it."

Bill Hartnett, who runs a consulting firm Hartnett Advisors after years heading up Microsoft's insurance practice,  compares the reaction of insurers to the cloud to how the Internet was perceived a decade or ago.

"When [the Internet] first came around in the late 1990s, the insurance industry had the same attitude: 'We're never going to put our data on the Internet; it's too risky,'" says Hartnett. "Look at where we're at now."

The insurance industry is extremely risk averse, reminds Hartnett, and even more so when it comes to the adoption of new technology.

"This is no exception," he says. "It's accurate to say that eventually you will see a policy administration system in the cloud, but that doesn't mean State Farm is going to wind up with their policy admin system in the cloud. There are a lot of capabilities making it easier for start-ups or new entrants to take a niche business where smart people with good intellectual horsepower no longer have to be experts in data processing. You can rent a data processing department (in the cloud)."

Over the coming years, insurers may have their policy administration system operating on a private cloud, points out Hersh.

"I think you'll find a good number of vendors are looking to put together SaaS solutions that run in a private cloud," he says. "The other interesting thing is we are seeing more willingness to think about cloud. People are becoming more comfortable with multi-tenant virtualization, but it remains a tough subject for most carriers. They would just rather not be in multi-tenant environment."

Hartnett believes economics is a key factor for insurers looking to move systems or data into a cloud environment.

"You get to a utility computing model where you pay for what you use just like electricity," he says. "It's a variable cost model and not a lot of capital infrastructure to put in place. You don't need building space to house a data center or the technical help to run it. It's like your Internet jack or your three-prong plug. The economics of computation and data storage make the cloud attractive to people."

Grid Computing

Hartnett points out the capacity benefit that allows carriers to get massive computational power without building their own clusters or grids can offer a huge savings in hardware.

"[Grid clusters] would not be available to a small company in the past without a massive investment," he says. "There also is the economics of data storage. When you deal with these volumes, companies like AWS have an immense capacity to store data with huge data centers. Only companies like Amazon, Microsoft or Google could afford to make an investment like that, which can benefit everybody."

The concepts of the cloud and grid computing are similar in terms of distributing the workloads, but Hersh maintains the big difference with grid computing is carriers have to optimize their software specifically for grid because grid computing is substantially different than running off a single server.

"You are farming out individual tasks to individual machines," he says.

With AWS in the cloud, carriers need to optimize things if they are running client-facing systems or systems that are run all the time, but because of virtualization, companies don't have to spread out the work on a bunch of machines; they can run it on one powerful machine and they don't have to rewrite their software.

"There may be a lot of development work that goes into making your computer grid-compatible, or you could rent a really big Linux box for a few hours from AWS," says Hersh. "With grid, there is a pretty substantial investment up front because you have to buy a grid management system. The ongoing costs are small, but the upfront costs are not negligible by any stretch."

Public vs. Private

Using the private cloud means it is unlikely insurers are going to be able to get access to additional computing power on short notice, explains Hersh.

"The private cloud doesn't have the same elasticity as a company like AWS offers," says Hersh. "With AWS, there is virtually no limit to the amount of computing power you can get in literally minutes. With a private cloud you might be able to ramp up with no immediate notice, but if you are planning to do something very much out of the ordinary from your profile—such as a large Monte Carlo simulation for risk modeling—you need to let [the cloud operators] know ahead of time to be sure they procure the right hardware."

If you look at any kind of rating models, financial or risk modeling or even modeling for pricing, all those are tasks can be completed by carriers today on their mainframes, according to Hersh. But as carriers look to modernize their systems, it becomes appealing to them to not replace the computing power of the mainframe because of the price.

Hartnett points out that the public vs. private debate is more of a response to concerns about security and privacy. But what businesses are seeking is the ability to take the same architecture that makes the public cloud so attractive and run it inside their four walls.

"If there's a burst of activity and you don't have the power you need, you can seamlessly pull in resources from the public cloud," he says. "You don't have to write code and when you are done with it you don't pay for it anymore. I think that seamless transition in the architecture makes sense, but the true benefit comes from the public cloud. That's where you get the advantage from the massive investment companies like AWS and Google are making and not creating your own mirror infrastructure internally. Not that the private cloud discussion is a bad one. There are some things where there will be a fair amount of private cloud purchased, but over time companies will find they don't need their own data center anymore. That may be 10 years or more away, but that's not an unreasonable expectation, either."

Security Issues

Hartnett believes the security aspect is driving the move to the cloud for a lot of business, though not necessarily for the insurance industry.

"Big providers are able to offer a far more secure environment for data and computation than you can do on your own," he says. "These are centers with very few people there. No one is going to mess with your data. The ability to secure that environment not only from a physical standpoint but from an electronic standpoint is probably better—although I could be proven wrong by some data breach—but it is far less likely you are going to get a data breach at AWS or Azure than you will at your average mid-tier insurance company trying to run their own security. The notion is that [cloud] is like the Wild West—it's not safe and you can't trust it. My response is: How did you feel about the Internet when it first started."

Third-party Data

There is a long history of the insurance industry using third-party data sources and integrating them in policy and claims systems, points out Hartnett. He believes cloud computing probably will make that easier because of the standards starting to evolve around that data.

"If you get a SOA or a Web services approach to building apps, deploying them in the cloud makes them much easier to consume," he says. "If you use industry-based standards like ACORD to do the integration it also makes it easier to pull different components together and build a sophisticated app. From a standpoint of third-party services, I think you will start to see attention given to all the other data out there that no one's figured out how to use yet. There are companies that have consumer data and track your retail habits or your cellphone companies that track location."

Sharing Data

Hartnett contends insurers will rely on third-party services for storing and sharing much of the telematics information it will be collecting.  

"The analogy right now would be usage-based insurance is the tip of the iceberg," he says.

Drivers are willing to share data on their driving habits in exchange for a better rate. Soon, they might be willing to trade other private information, such as their buying habits.

"Carriers don't know what you bought online or at Target, but credit card companies do," he says. Aggregators have all this purchasing data that allows them to correlate when people are suddenly purchasing cribs or strollers or when you frequent certain places."

That data is not available to insurance companies now except through third-party sources. An insurer has your driving record and claim history, so it's hard to see whether there is any proprietary data an insurance company would have, according to Hartnett. What is proprietary are the algorithms they use to break out relevant data for underwriting or claims settlement.

"There is capacity to figure out correlations between certain behavioral things recorded by third parties to determine what kind of risk you are—good and bad," says Hartnett. "People who might be considered good risks based on traditional measures may turn out not to be and vice versa. The best example of that is how Progressive got started doing personal auto. There specialty at that time was taking non-standard risks and giving them better prices because they had a better mousetrap. The more data you get the more opportunity you have to do things like that."

Aggregated data, such as telematics, that is not tied to personal identifiable information eventually will end up in the public cloud, explains Hersh.

"The massive amounts of data that are being crunched for telematics resembles the work done on rating and pricing," he says. "It's a great candidate for the cloud. If you are doing a pilot project for telematics you are going to collect a huge amount of data that you may or may not want to store forever. You may end up buying a bunch of storage equipment that you may not need. The cloud fits the profile in terms of reliability, scalability, and cost that might fit in with something like a telematics project."