X

Log in

Forgot password?

Don't have an account with us?

Register Now
instech-london-podcast-129-cole-winans-flyreel-podbean-featured

Cole Winans

Founder & CEO, Flyreel

Know your risk; the self-service property inspection tool helping insurers & their clients

The most useful insurance products are often a combination of a simple concept powered by complex technology. 

Flyreel’s AI Assistant is a great example, harnessing the potential of AI in a practical application that allows policyholders to conduct self-inspections using their phone camera.

Founder and CEO Cole Winans joins Matthew to discuss how insurers are using Flyreel’s technology to create a comprehensive baseline record for underwriting, risk management and claims. Cole also shares some great tips on starting a company, finding customers and how to scale a business.

Talking points include:

  • Balancing AI with human intervention
  • How to sell to insurers
  • Developing portfolio risk characteristics through data
  • Why a mutual customer benefit is vital
  • Building a company culture and collaboration

If you like what you’re hearing, please leave us a review on whichever platform you use, or contact Matthew Grant on LinkedIn

Sign up to our newsletter for a fresh view on the world every Wednesday morning.

Continuing Professional Development – Learning Objectives

InsTech London is accredited by The Chartered Insurance Institute (CII). By listening to an InsTech London podcast, or reading the accompanying transcript, you can claim up to 0.5 CPD hours towards the CII member CPD scheme.

  • Claim 0.5 hours for listening to Episode 129 of the InsTech London Podcast

Know your risk; the self-service property inspection tool helping insurers & their clients – Episode 129 highlights

Matthew: Cole, your background is in software engineering. Where did Flyreel come from and what took you into insurance?

Cole: Before Flyreel started, we explored a couple of ideas. The early prototypes were fun to build but they didn’t amount to a great business. As part of that process, we created technology around artificial intelligence (AI), machine learning and computer vision. Our mentors and early investors tipped us off about an opportunity in insurance. That’s when the doors opened for us to invent and create meaningful technology that generated value, but this time, in a market where it could drive considerable impact.

Matthew: Did you already have the idea for what you’re now offering with image recognition and AI?  

Cole: No, we had developed an early proof of concept for smartphones where users could pan across a room and it would create a list of items in that room automatically using computer vision.

At that time, there was a lot of growth in the aerial and geospatial analysis of roofs at scale, with some early adoption and deployment of computer vision. We thought if it were valuable to process the exteriors of homes at scale, what could insurance carriers do with the interiors? That overlapped with the development of the proof of concept and it was really our “Aha!” moment.

Matthew: How are insurers using your technology in the underwriting process? 

Cole: Insurance carriers use a variety of geospatial and aerial inspection tools to get a snapshot of the exterior and roof of a property. They also do drive-by inspections and beyond that, there are in-person inspections. In-person inspections are the most expensive, but they’re also the most detailed. They require collaboration with the policyholder if someone has to go inside.

Because of the coordination and economics of these methods, insurance carriers have had to sacrifice data and visibility into properties because they can’t view them at the scale they would like to. That affects risk selection, segmentation and pricing. They’re relying on assumptions and educated guesses. There’s now the ability to know and acquire ground truth at scale through remote capture technology and self-inspections.

We work with underwriters to allow them to acquire ground-level truth on properties at scale, to make informed decisions. The outcome of that is better risk selection, better pricing and happier customers.

Matthew: Where does Flyreel enter into the underwriting workflow? 

Cole: It’s usually right after the quote, in the underwriting window. Usually, a quote is issued to the customer and a Flyreel inspection is then ordered. The policyholder receives an email or a text to walk around their house to scan it. As they scan the home, our system automatically identifies critical data like features, hazards, risks, and more. 

That’s the primary use case. The close second is renewals for homes that have been on the book for a while, but haven’t been seen or looked at ever, or at least not recently. A self-inspection link is sent to the existing customer for a quick review to ensure that everything is adequately covered and priced, and that the customer has the coverage that they think they do. 

What’s been exciting for us is seeing the behaviour that comes out of these interactions. We’re often able to inform policyholders of risks they were unaware of and provide them the opportunity to reduce that risk, or acquire additional coverage to better protect what’s most important to them. 

Matthew: How does your system work? Presumably, there has to be some human interaction with the AI? 

Cole: We do have human intervention for quality assurance. We don’t shy away from that and I use a common example to explain why. Shortly after launching, our computer vision platform picked up an image in someone’s backyard. It was round and circular with a surface that was reflecting the blue sky. The platform labelled it as a hot tub, but it was actually a patio table with a glass top. These are silly but sensical errors that, with the help of human supervision, can be reduced through human-computer interaction (HCI). 

Matthew: In that field of computer vision, are you able to buy in the libraries, or do you have to build them from scratch?

Cole: That’s one of the things that some of the larger insurance carriers were wrestling with when we were entering the market. Should they build it with their data science teams or partner with someone like Flyreel? Many of these insurers went through a cycle of experimenting and building things themselves and then, around a year later, concluded that they should partner. 

To truly excel, machine learning requires an incredible amount of engineering focus and data discipline. Data science teams at insurance carriers shouldn’t be grappling with patio table vs. hot tub problems. They have more important things to do. 

We’ve also found that the system has to work extremely well with imperfect data. Images from open datasets and real estate websites are fine for demos, but those aren’t the images that a policyholder sends from their smartphone. Often, the video bounces, sometimes images are pixelated, the lighting isn’t always great. Well-lit real estate photos aren’t the same as a dark utility room with an electrical panel, where we’re expected to detect, for example, whether or not there’s a Federal Pacific Stab-Lok breaker there. 

Matthew: Are you also looking outside the property? 

Cole: We do the full interior and exterior. We’ve developed our system in a way that it isn’t simply uploading photos into a browser or filling out a form. The policyholder is interacting with an automated AI assistant in a conversation. Based on what it sees or doesn’t see, the automated assistant will change that conversation to follow up and get more information. 

One of the examples I give is a pool. If the system detects a pool, that should trigger some follow up questions. If it hasn’t seen a fence, it should ask to see the fence because insurers don’t want someone accidentally walking into that pool. The AI assistant will ask questions specific to the risk to get a complete understanding.

Matthew: Flyreel started in residential, but what about commercial? That’s a big area where companies are struggling to get good information. 

Cole: The overarching value of Flyreel is most prevalent in high volume use cases. In the end, computer vision and machine learning models simply rely on statistics and probabilistic modelling to arrive at reasonable conclusions to make decisions. AI deployed in a nuanced commercial use case, like looking at a stadium, isn’t able to learn from a large quantity of those to transfer knowledge to the next one. 

Our commercial focus is on high volume use cases like restaurants. We’re able to pick up on sprinkler heads and ventilation, presence or absence of exit signs, slip trip and fall risks where there may or may not be a rubber mat in front of a soda machine. Other examples would be multi-family housing, retail, barbershops. Those are the areas where we’re either entering or operating today.

Matthew: Are you able to understand things like construction type? That could be another benefit in terms of working out what a potential loss could be from an earthquake, hurricane, or flooding.

Cole: We get a good glimpse into that, but we can’t compete directly with someone that’s able to get behind the walls. However, compared to not having any visibility into the property, or just a couple of photos from real estate sites, we provide a considerable value advantage in comparison. 

We automatically detect exterior building materials and we can provide visibility into the interior structure and more. That’s one of the main pain points we have solved for several residential and commercial customers. 

Matthew: A lot of the cost for a building, if it gets damaged or destroyed, is the interior fitting. Two buildings might look similar, but one might have a very high-end kitchen or expensive fireplaces inside. 

Cole: Detecting the items in a room is important, but there’s a larger-scale problem in that insurance carriers need to understand the structure. If something happens, they need to be able to accurately account for that and get the policyholder on a path to resolution. 

A 20-minute minute walkthrough done by the policyholder creates a comprehensive baseline record of the interior and exterior, creating considerable value upstream for any future claims. If something happens, we can pull up that original scan and look in the living room, see what the finishes were and use that to quickly validate claims. 

Matthew: You started with one clear use case and are now spinning off into more areas. Often, people try to do too many things with the product and fail to land on one. 

Cole: There’s no shortage of opportunity, but we take the relationships with our customers very seriously and that brings a challenge of focus. We need to be responsible and pick and choose where we’re going to go. 

We’ve maintained a tight focus on underwriting to date, but we’re excited about where we’re going and the technology we’ve innovated on for claims and rolling that out. That’s a great opportunity and a challenge at the same time.

Matthew: You’re also making it easier for customers because they can now get help in what they need to cover. 

Cole: When insurtech was coming to the market, almost everyone was focused on speed. Do customers want speed? The answer is not that simple. They want an efficient experience that’s not going to take a lot of time, and they want to know that their stuff is protected and that they’re paying the right price. If making a better decision means 15 to 20 minutes walking around and scanning all the things that they want to be covered, they’re willing to make that trade-off. 

The customer experience is really important and the market as a whole will see a little bit of a correction there. The future of insurance isn’t one-click insurance. It’s an efficient, consumer-grade experience and making sure the insurance product is the right one.

Matthew: When we last spoke, you had 25 clients in the US and one in Canada. Can you share some examples of who you’re working with?

Cole: We’re very grateful for our relationships with companies like State Auto. We also recently announced our relationship with Kingstone, who are seeing great results from pushing the boundaries of innovation. It’s the same with QBE and Mercury. These are really interesting groups that are getting great results from their deployments of this technology. 

Every insurance carrier is able to personalise our application. The AI assistant has its own voice and way it communicates, and the workflows are entirely configured by the insurer. They have incorporated the tone, voice and values of their businesses into this customer experience and are attaining great results.

Matthew: Another thing that strikes me is once you’ve assessed a certain number of homes, you can use that data to start understanding what the risk characteristics of a portfolio are.

Cole: That’s where we’re headed and where we are being used at a healthy scale. Acquiring that data in a structured way means we can perform some valuable analytics about risk selection. We can ask questions like what parts of a book are performing better than others? What are the attributes that are potentially driving that? 

The customers that have adopted us at the greatest scales are now able to pull that data and advance their models, advance their understanding of risk, and they’re reaping the rewards. That was our hypothesis and it’s nice to see that starting to prove out with our customers.

Matthew: What advice can you give people who are new to the industry about selling into insurers? 

Cole: The way that companies sell changes over time. A start-up with hopes and dreams, but no proven business model, has to sell differently than a start-up with 25 insurance carrier customers, a good sized team and financial backing. 

We’ve been both of those. When we first entered the market, it had to be a relationship sale. We had to sell the vision and de-risk it for the prospective customer as much as possible. State Auto was one of our first customers, and they have an incredible vision and focus on deploying the best of today’s technology to deliver the greatest experience to their customers. That’s so ingrained in their culture that it led them to be open-minded in terms of experimenting, testing and deploying new technology. 

That’s the persona of an ideal early adopter customer for a start-up that has a great idea. Try and have conversations with them to say, “I believe, to my core, that there’s something here. I have no proof, but I’ll do everything I can to deliver this.” If we can’t demonstrate a proof point, they owe us nothing, and de-risking it that way is how we got some of our first chances to enter this market.

Once the idea has been proven, there’s still the rollout and longer sales cycles, but there are technical ways to bridge that gap. One of the things we did was around integration. There was potentially a delay related to integration where IT teams are very busy, and if we needed an integration into our system before the customer can go into production, that was a risk. We created a standalone dashboard where they could access data while we were waiting on those integrations to happen. 

Matthew: Fundraising is a big challenge for businesses looking to scale. Are there any lessons you’ve learned along the way?  

Cole: A big lesson learned is to raise a little bit more than is needed to last through 12-18 month sales cycles. Working with prospective clients and understanding what their drivers are, where they’re hurting, sometimes takes 12 months. Being able to ride that out without a tremendous financial risk to the business is important. Customers want to know that partners will be around for a while and raising a bit more capital in this market is a pretty wise move.

Matthew: Guidewire is one of your investors. Why did they get involved and what are you doing with them? 

Cole: Guidewire is such an incredible organisation, going back to its founding and the early vision it had for the industry. It’s established itself as a market leader and the role it plays in deploying capital into start-ups matters so much. It has great connections to venture capital firms and we were connected through one of those. 

It has a vision for where the industry is going, and it’s one of the underlying platforms powering that. Aligning with groups that are driving the future, in a way that also benefits Guidewire, creates a real win-win. That’s what their team saw in us and it’s made the foundation for a great partnership.

Matthew: What advice can you give people looking to scale a business quickly in a fast-growing area? 

Cole: Anyone inventing technology needs to have convictions about the future. They then need to validate these along the way and putting the customer first is so important. I hope and believe all of our customers feel that every single day when they work with Flyreel. 

Internally, start-ups need to have a vision for the culture and people they want. Diversity, diversity of thought, and cultural diversity has to be the core of a successful company, and it takes a lot of very deliberate work. Now that we’ve gone from six people to the high 30s, we don’t have the luxury of being with everyone in a single room, so putting extra effort into values and communicating those is super important. 

One of the most important parts of that is linking up customers with really happy employees at our company that understand their business and have a vested interest in their success because it’s not possible to fake that.

Matthew: What’s next for Flyreel and property insurance?  

Cole: We think that the inevitable future of property insurance is that when a customer wants it, they take out their phone and walk around the house or business with the camera on. That policy will be dynamically priced and constructed specific to the customer and their property in 20 minutes. 

If we achieve that vision, it changes the entire experience. It improves the economics, the insurance carriers will do better, the customers will be happier, and it’s the experience that everyone deserves. That’s what we’re after in the long run.

For more information on Flyreel, go to flyreel.co

Stay up to date with the latest from InsTech