RHS 162 - Insurance Robots-as-a-Service with Quandri.io

Host: In a crude laboratory in the basement of his home. Hello everyone and welcome back to the show. Today we have a tremendous episode for you, an absolutely tremendous episode. We are bringing back a topic that was originally touched on in the episode where I talked to Jeff Roy, which we did about two months ago. So if you want to go back in the archives, you can find that episode. Jeff Roy is a tremendous guest. I'm an enormous fan and he's an awesome friend. And we got into what's being called digital workers or basically bots that use machine learning and algorithms to get into your agency management system and do highly repetitive tasks.

Host: Jeff talked a little bit about the technology. We talked a lot about other stuff that was going on, and he brought up the company that he's using for this particular work, which is Quandary. You can find them at Quandary, Q-U-A-N-D-R-I.io, Quandary.io. And this idea fascinates me. It fascinates me for a whole bunch of reasons. And I wanted to have the founders of Quandary on, Jackson and Jameson. I'm going to butcher the pronunciation of their last name, we'll say Fregeau, F-R-E-G-E-A-U. They are from Canada. They are the founders of Quandary.

Host: I wanted to have them on to talk about exactly what are digital workers? What is this tool? They basically break down what are bots, what are kind of robots as a service. And then ultimately what our topic is, what are insurance robots as a service? And what does that mean for us as insurance professionals, as agency owners, as producers? What are some of the tasks that these bots are able to do today? Where are they going in the future? And what efficiencies are they bringing to the table?

Host: What I think is particularly interesting about digital workers in particular is that I think they are going to have a major impact on the virtual assistant industry. In so much as what we are currently paying virtual assistants to do today in terms of repetitive tasks, many of those, not those that require decision-making because these bots are not making decisions, but for those repetitive tasks that don't require decision-making, why would you have a human doing those when a bot, after being properly trained and they describe what that means, can be even more effective, more accurate? It's really, really interesting stuff. I think you're going to love this episode. It's certainly something to get your brain going on building this idea of a human-optimized agency.

Host: Before we get there, I want to give two quick shout-outs. One is to the brand new community resource that I created called Finding Peak. You can go to findingpeak.com. It's a substack. You can subscribe for free. You get articles every Friday. And what we did was we added an additional feature where every Tuesday for paid subscribers, and paid subscribers are $7 a month, after the fees that Substack takes that basically make nothing, it's not about making money. It's about buy-in from you guys. But for paid subscribers, every Tuesday we're doing a deep dive video post, which allows our paid subscribers to dig into topics, which allows me to dig into topics and really break them down through videos, video walkthroughs of different tasks we're doing, of different tools that we use, breaking down different concepts, answering subscriber questions, following up on comments that paid subscribers make, and basically using this as a resource to deliver to those interested, to those looking for the next level in their business, what it means to be a digital agent.

Host: That's really what Finding Peak is all about, is building peak performance into your business, both physically, emotionally, from a process standpoint, from a tool standpoint, from a leadership standpoint. What does it mean to get to peak in your business? So we're searching for peak. The subtext of Finding Peak is in pursuit of peak performance in business life and insurance. And my friends, this is something that I've had on my brain for a long time. I kind of launched Finding Peak a few years ago as a brand. It didn't last very long. I wasn't ready. I'm ready today because I want to help those who are interested in building a digital agency, who are interested in peak performance, who want to figure out how to optimize their agency to be exactly what they want it to be. I want to help them get there. So I say go check out findingpeak.com. That's my big ask. If you love this podcast, if you like following me on LinkedIn and the different things that I do, I'm going to be very tactical, very deliberate, very consistent at Finding Peak. And I think you're going to get a lot of value out of it.

Host: I also want to give a big shout out to the parent company of Rogue Risk and that's SIA. SIA has been a tremendous partner for us. They are not asking me to do this read for them, so I don't want you to think that part of my deal is I had to do reads for them. This podcast is my own still. SIA doesn't own this podcast. I can do whatever I want with it, but I have to tell you that you never know what you're going to get when you're acquired. I had no idea. I've obviously trusted and think very highly of Matt Maciello personally, and then once I got to know his team and the rest of the leadership team, I thought very highly of them and was very impressed, but you never know.

Host: I can tell you that joining SIA, being part of this very large 40-plus-year-old network and watching the wheels turn as they adapt to the changing environment and adopt new tools and processes to help their network agencies become successful. Guys, if you are an SIA member, I think it's time to open up your ears a little bit and start taking notice of what's going on because I think there's some incredible stuff happening. And if you're looking to optimize the revenue in your agency, guys, I know that for a long time there's been mixed emotions around all aggregators, not just SIA. But I can tell you that a lot of the things that are happening inside the SIA ecosystem are very, very exciting.

Host: So I wanted to do that read for our parent company, the company that's allowed me to really realize the vision of Rogue Risk. I couldn't have done it on my own, and it was only through my partnership with SIA that we're going to be able to execute our vision of the no-ceiling insurance career, of being a national small commercial insurance brand, competing against some of the other biggest brands in the country. They make that possible. So I wanted to give a big shout-out to them. So with that, I will be quiet. In terms of this intro, I will stop the intro. We can get on to this absolutely tremendous conversation that you are going to love with Jackson and Jameson from Quandary.io. Here we go.

[Transition to interview]

Host: Hey, what's going on, guys?

Jackson Fregeau: Hey, Ron. Sorry, I'm just switching my headphones here.

Jameson Fregeau: Hey, Ron. Hey, what's up? How are you doing?

Host: Not too much. How are you doing?

Jackson Fregeau: I'm doing very well. I appreciate you guys jumping on here with me.

Jameson Fregeau: Yeah, it's great to meet you. Heard a lot about you from Jeff and some of the other guys we know in the industry. So nice to finally see you.

Host: Oh, nice. Wow, that's good. And Jeff's got great things to say about you guys and the work you're doing. I don't know if your ears were ringing or not, but he was on the podcast about two months ago and was talking a little bit about what you guys have going on. At the end of the show, I said, "Hey, can you connect me with you guys? I'd love to take a deep dive and just get to know you a little better and share what's going on and ask some deeper questions." Obviously he understands the technology a little bit, but he's still just a user and gave us some good use cases.

Host: I think a lot of people have questions about staffing, which is obviously a huge issue. That leads you to things like how do we find staff and methods of servicing that can reduce costs. So people look to VAs. VAs are great, but also a lot of work. And then you look to something like what you guys are doing and some of the basic functionality. Your mind starts to spin around ways that you can be more efficient and effective while keeping cost manageable and consistent. That's very exciting.

Host: To kick things off, I'd love for maybe Jackson, we start with you, and then Jameson with you, just so you guys don't talk over each other. Maybe just a little history. You don't have to give us since birth, but just the genesis of you guys and your origin story of Quandary.

Jackson Fregeau: Sure. I'll start with my background and kind of go into Quandary, and then Jameson can give his as well. I originally did a finance degree. I was always really interested in entrepreneurship and business itself, but didn't really see that as something that you went to school for. That didn't make a lot of sense. And I really like math, so I ended up doing a finance degree. During my finance degree, I was getting pulled in the direction of entrepreneurship. So after school, instead of going the finance route, going into corporate finance or banking or something like that, I ended up working for a startup right after school.

Jackson Fregeau: I had a great boss there, a really good mentor. I was the third employee there and basically was the assistant to the CEO. I started doing basically everything he didn't want to do. I was writing policies like SOPs, standard operating procedures. I was booking travel. I was paying bills. I was paying taxes. I was helping to hire. I was basically doing everything under the sun and learning how to build a business. He was a really great mentor to teach me how to actually do this. I really looked at that as an opportunity to learn how to build a business myself.

Jackson Fregeau: It was actually at the tail end of that business that we had the idea for what ended up becoming Quandary. I had become the COO at that business after a couple of years, and I was managing a team of about 30 to 35 people. A segment of that team was doing high volume, repetitive work. As everybody probably can attest to, that work is boring. It's hard to staff. It's hard to manage. There's a lot of turnover, so it's hard to train for. And there's a lot of mistakes that happen in that work because people are just doing the same thing over and over.

Jackson Fregeau: Naturally, we were looking for technology to put in place to automate that work instead of having people do it. The system that we were working in didn't have very good APIs. There wasn't really much we could do in the back end. So we were looking for solutions to work over the user interface or over that software platform itself. I came across bots as a potential solution for that. I learned the technology. This was right at the beginning of COVID, so I had some extra time. I learned the technology, built a few bots to solve that problem in that business.

Jackson Fregeau: Jameson and I had been throwing around different business ideas for the last couple of years. This seemed like it had legs. There was nobody offering robots as a service to any businesses that we could see in the US or Canada, especially to SMBs. We saw an opportunity for businesses to really benefit from this kind of technology, technology that can save time, repurpose staff away from these high volume repetitive tasks. So we started Quandary with the mission to free the workforce from this high volume repetitive work and allow people to focus on what people do really well and robots to focus on what robots do really well, which is that high volume data processing.

Jackson Fregeau: We started Quandary to solve that problem initially in industries that had high volume repetitive operations. Healthcare, legal, finance and accounting, retail, and insurance were the first couple of industries that we worked in. After working with our first couple of insurance agencies, a little less than two years ago, we realized pretty quickly how massive of a pain point there is in the insurance industry for this high volume kind of work, and that it's not going away anytime soon. There are a lot of structural reasons in the industry why that's not going away. There was almost a need to solve it. We were seeing a demand for this technology. In other industries, it was helpful, but in insurance, people really needed it.

Jackson Fregeau: So we pivoted the company at that point to focus on insurance specifically, on insurance brokers and agents. We have been going deeper and deeper into the distribution channel of insurance ever since. That's now all we do is build robots for insurance agents and insurance brokers.

Host: Tremendous. I got a bunch of questions there, but Jameson, I'm interested in you.

Jameson Fregeau: Yeah, of course. First, thanks for having us on. Jackson told his story of going into and learning about startups from the business side of startups. I took the other route. With a tech startup, there are kind of two different sides to it, the other side being the technology side. So I did a computer engineering degree. One of my first actual roles as an engineer, as a computer engineer, was at a startup that was building robots in a sense. It had similarities, but these were actually boats to go out into the ocean, autonomous boats that were solar powered to do scientific research and gather all this data.

Jameson Fregeau: I was on a small team. I started when there was just the two founders and myself and one other employee. I immediately fell in love with the startup environment. I very quickly realized how much you can get done and how much of an impact you as the person working for a startup can have on the company itself, but also how much of an impact that startup can have on you as an employee and just give you so much opportunity to learn and grow. I was there for a while, just building all sorts of systems, all sorts of prototypes for this company.

Jameson Fregeau: After that was right around the time that Jackson and I were bouncing around business ideas. We've been doing this for quite a few years. There were a few that we thought we wanted to pursue and then stopped after a little while because generally we thought for most of those that they weren't really as big of an impact as we wanted to have. When we first started looking into robotic process automation, bots and automation in general, I think after working with the company that Jackson worked for and a few other companies after that, we quickly realized just how massive of an impact this can have on so many different industries.

Jameson Fregeau: Like Jackson was saying, that led us to starting Quandary and kicking it off and working across all sorts of different industries. One of the key points about when we first started working with insurance brokers and agents that Jackson touched on is that all of these other industries that we had worked with, they had a pain point around high volume repetitive processes that were taking up labor. But the pain in the insurance industry was probably about 10 times what we were seeing in all of these other industries. It was a very immediate need. That led us to pretty quickly pivot and focus solely on the insurance industry because that's where the biggest need is and was. We haven't looked back since.

Host: Thank you for that, guys. That's awesome. It's a cool journey that you have, someone coming at the business side, someone coming at the tech side. I know having founded Rogue Risk a little over two years ago, if I had had a technical or more technical founder alongside of me, I think we would be a lot further along than we are today. I love technology and I'm a highly functioning user of technology, but I cannot create technology. I'm not super good at taking what I would like something to do and turning it into something that someone could actually build. That skill of translating a user's feedback or what a user wants and translating that into a document or a series of tasks that you would need a system to do, that's a whole other skill. I think that's really cool that you guys have that between you.

Host: My first question in general, and some of these are going to start fairly remedial, but I think it's important in setting the stage for the audience. Just so everyone's clear, they're not talking about actual robots like you see in Star Wars or something like that. When they say robots, I'm assuming that's not what you're talking about?

Jackson Fregeau: No, not yet. I can see some of the audience going, "There's this mechanized humanoid going to come into my office and start banging on my keyboard now." No.

Host: So when you say bots, what does that mean? We hear on social media, on the news, bots. If you're into cyber insurance, then bots are a big part of that. What is a bot? What does it mean? What is it actually that a neophyte could grab onto?

Jackson Fregeau: Good question. Before I jump to that, one thing I'd like to call out is that these mustaches are not a fashion statement. Quandary is doing a Movember fundraiser. Jameson and I are having a competition, just so you know, Ryan.

Host: I liked them though. I'll tell you, I was a little... mine's a little further down my face, but I do like it.

Jackson Fregeau: Thank you. In terms of a bot, basically why we call it a bot and what exactly that means is that this robot or this program, this piece of software, is going to interact with your software in a similar way that you do. They're going to get a username and password. They're going to access your system that way. They're going to click buttons on the screen. They're going to read text off the screen. They're going to interact with that system in the same way that you do. They're just going to do it a lot faster. They're going to make fewer mistakes, and they can run 24/7. That's why we call it a robot, because a lot of these systems don't have that backend access. You can't automate things through an API or through a backdoor. So you need to go through the front door. You have a robot that emulates what that person is doing.

Host: Now, if you are working inside a system that has an API with capabilities, can you set it up that way too? Does it have to go in through the front door, or can your systems plug in through an API connection as well?

Jackson Fregeau: Yeah, we can plug in. If there is an API, we'll use the API. That's actually easier for us to do. So we'll interact with it that way. But if it doesn't, then we'll use the username and password method.

Host: Sweet. What are the processes? The biggest question I got after Jeff's interview was, he gave some kind of high level examples of what these robots could do. I think it would really help people if, in as broad a stroke as you could, help them better understand exactly what kinds of stuff they are capable of doing. I think some people basically assume it's almost nothing. Other people have the assumption that it's everything that a human could possibly do. The truth is somewhere in the middle. Kind of break down what are the general functionality features, and then maybe what are one or two actual procedures that you've seen people using this for, just to allow people to visualize what you're talking about.

Jackson Fregeau: Sure. At a high level, looking at any process that is rules-based, high volume, and repetitive. Rules-based meaning you can break it down to "if this happens, do this, if this happens, do this." It's high volume, it's repetitive, and it's all digital inputs. Meaning that what that bot is actually reading is some kind of electronic digital input, like a number in a spreadsheet, a number in a software platform, a PDF document, not a written signature or something that somebody has put in by hand. At a high level, those are the types of processes that you can automate.

Jackson Fregeau: In terms of actual specifics that we're doing right now for different agents, the way that we approach this is by having productized robots. Basically a robot that is pre-built, that's already solving a problem for an agent that they can then deploy into their system and do a little bit of configuration to make it fit for their operations, but not something they need to build from scratch or work with us to build from scratch. I can quickly run through what those three products are. They all automate different aspects of an agent's operations.

Jackson Fregeau: The first one helps with the daily download process. The EDI process that happens on a daily basis. All of those clicks that need to happen and then all of the matching that needs to happen afterwards. All of the policies that don't end up routed to the correct account, all the e-docs that don't end up routed to the correct account. The bot can identify those, figure out where they need to go, send them to the correct account so that a person doesn't need to do that every single day in the morning before the brokerage starts for the day. That's the first one.

Jackson Fregeau: The second one, we call the e-doc executive. That will focus on categorizing, storing, and naming documents. When those e-docs download, the actual policy documents, the bot opens them up, figures out what they are, names them something contextual for internal standardization. They're named properly. When those policies get sent out to customers or accessed through a customer portal, they're named something like "Personal Auto 2015 Ford $1,483," not whatever they come down through a carrier safe standard naming conventions across your business. That's the second application of it.

Jackson Fregeau: The third, and I think our most interesting one, is the renewal reviewer. This robot will look at last year's policy. It will look at this year's policy. It will extract all of the data out of both of those policies. It will prepare a report and give that report back to you. That report is a comparison from last year to this year, showing you any major changes. Premium changes, things that have been pulled off, things that have been added on, discounts, coverages, surcharges, that sort of thing.

Jackson Fregeau: It will also compare this year's policy to a rule set. Let's say you want every policy to have sewer backup coverage, or you want to know every premium increase over 20%, things like that. For example, every premium increase over 20% over two years or over three years. The robot can look at things like that, give that back to your team in a report so that they can scan that in 20 or 30 seconds instead of somebody spending 20 or 30 minutes preparing that report. What it allows their team to do is only focus on the ones that are most important, where there's a risk of churn, a risk of that person not retaining, or where there's upsell or cross-sell opportunities. You can focus on the 20% or 30% that are really important, and the robot can handle the rest of them.

[Host interjects with a mid-roll request]

Host: What's up, guys? Sorry to take you away from the episode, but as you know, we do not run ads on this show. In exchange for that, I need your help. If you're loving this episode, if you enjoy this podcast, whether you're watching on YouTube or you're listening on your favorite podcast platform, I would love for you to subscribe, share, comment if you're on YouTube, leave a rating review if you're on Spotify or Apple iTunes, etc. This helps the show grow. It helps me bring more guests in. We have a tremendous lineup of people coming in, men and women who've done incredible things sharing their stories around peak performance, leadership, growth, sales, the things that are going to help you grow as a person and grow your business.

Host: They all check out comments, ratings, reviews. They check out all this information before they come on. So as I reach out to more and more people and want to bring them in and share their stories with you, I need your help. Share the show, subscribe if you're not subscribed, and I'd love for you to leave a comment about the show because I read all the comments. Or if you're on Apple or Spotify, leave a rating review of this show. I love you for listening to this show and I hope you enjoy listening as much as I do creating the show for you. All right, I'm out of here. Peace.

[Return to interview]

Host: So let's get back to the episode. That's really interesting. What about automated rating? If a renewal comes in, you probably get this question all the time, but that's all right. We're on a show, so I have your attention. The idea of in that renewal review, you get that this renewal came in 20% over. Can you ultimately get, or is the goal to get to where that robot could then actually re-rate that business through a platform? Depending on who you use, some sort of rating platform could take the information, plug it in, and re-rate it, come back and say, "Hey, your Hartford's renewal is a 20% increase this year, but Travelers is actually only an 11% increase. This might be a good place to move it." Is that level of sophistication possible over time? Is it teaching the system, or is it just as sophisticated as you want to get if everything is in place?

Jameson Fregeau: That's absolutely part of the goal. It's very possible to get to that level of sophistication. What it takes is time and training. We need to train these robots to be smarter, to be more intelligent, to be able to process more information and more complex and more nuanced information. That is something that we're looking at pretty seriously, being able to add that onto the renewal process. Ultimately, where we want to go with this is really being able to use a robot for all your backend processing, high volume, repetitive activities, and have people focused on what people are really important for: making decisions, talking to customers, upselling, evaluating risk. Allowing all that data processing to happen in the background through a robot.

Jackson Fregeau: One of the first things I wrote down when I was doing the business plan for Rogue was the term "human optimized." Everything that we do at Rogue is pushing towards that. We're only two and a half years old, so we're still very much in startup mode too. What you're describing fits very much in with our core philosophy: humans should be used in the activities and in the touch points where they add the most value, and they should be removed or have their role greatly reduced in the places that don't have value.

Host: It's why we're always looking and trying to find more automated ways of producing COIs. I know a lot of agencies, and hopefully over time this will change, but a lot of agencies still see one of their primary value mechanisms being the production of certificates of insurance. That was at one time a very valuable thing that we did. Today, it's kind of like the barrier to entry. It's the bar. If you can't rapidly pump out with limited friction a certificate of insurance, they're almost immediately looking for someone else, especially contractors or anyone who has to do more than a few certificates a year.

Host: The question we always ask ourselves with processes such as these is, where are the places that, like you producing a COI today as an independent insurance agency, adds zero value to your customer? It's just an expectation. If you can remove the human or reduce the human's touch as much as possible from a process like that, then you're giving the customer what they want, but it's not something that adds value. So don't spend your people's most valuable resource, their time, on that activity, because it's just the bar. It's an expectation.

Jameson Fregeau: Right. We spend quite a bit of time just exploring, as you can imagine and as you know, there are countless of these processes that are just necessary for doing the business that you do. We spend quite a bit of time just exploring all of these processes and what has the most value for building a bot around. Those three that Jackson mentioned are our starting point of the processes in these BMSs that we think add a lot of value to the brokerages and the agencies. We are constantly looking and evaluating what the next productized bot will be because there's an almost infinite number of processes that can be automated through these bots.

Jackson Fregeau: To your point, what you were mentioning there, Ryan, what this is all about is optimizing the person, right?

Jackson Fregeau: Getting people away from the work that is really important — like issuing a COI or doing a renewal review — it's important, it has to happen, it adds a lot of value to the brokerage, but there's no additional value of having a person do that. It's not like talking to a customer. It's not like trying to sell somebody. It's not like you need all of that insurance knowledge to evaluate a risk profile. It just needs to happen. That's really where you can put a bot in place to optimize people to focus on what people are really good at.

Host: Yeah, I mean, while you were doing those first three, which are an awesome place to start — automated rating, accounting, COIs, data gaps — I know Chris Paradiso, a buddy of mine, at one time he had three virtual assistants working overseas basically on data gaps, plowing through records, making sure there were 17 pieces of information he wanted on every account. Now you're paying three humans who make mistakes. Not that a bot can't make a mistake, but it probably won't make that mistake twice. If it makes it once, it learns. Humans will make the same mistake multiple times. That's a lot of cost, a lot of training, having to come behind those three humans and QA their work regularly. Something like that would be fairly easy for you guys — just ping every record in the database and make sure all 17 specific pieces of data are captured. If they're not, call them out or fire off an email. That kind of stuff makes so much sense. Those tasks in most agencies aren't even getting done, which is why you have inconsistent phone records, inconsistent email records, no FEIN for 40% of commercial clients. Most agency owners don't consider that 99.9% of the time it doesn't matter that you don't have Jim Jones Bakery's FEIN. But when you go to quote their workers comp and you don't have it, now you have to call them, they don't pick up, you email them, they don't get back to you for three days. Do that for hundreds or thousands of clients and you're wasting so much time in a year. Tools like what you're building at Quandri are the natural next evolution. Humans make decisions. Robots, automation, bots — they do all the process, the grunt work. But we still need humans for the decisions and the relationships. There's so much that can be taken out of the hands that is just busy work today and not actually valuable. You're paying your most expensive people to do things that don't add value to your clients. That should be scary to a lot of agencies.

Jameson Fregeau: It makes people feel productive doing that busy work. We've all had that — you end up answering emails for four hours and feel like you've been really productive. But have you actually moved the needle? Have you brought a lot of value to the business? If you remove a lot of that busy work, people don't need to have their nose to the grindstone for eight hours a day, but they can produce way more value because they're focused on much higher leverage activities.

Host: Right. You can create a lot more with the time you're putting in if you don't need to do that low-value busy work. So let's say I'm an independent agent listening to this and I'm thinking the daily downloads and the edoc thing would be amazing for me. What does an engagement with you guys look like? How do you get started? I don't need a hardcore sales pitch, but I want you to talk about what you do. Does my AMS connect? How does this work? How much time does it take? Talk a little bit about the buildup process. We'll give a human a thousand mistakes, but if something automated makes one mistake, it doesn't work and it's all a sham. How do you deal with the get-off-the-ground stage?

Jameson Fregeau: You're right — a bot at 97% is almost worse than a person at 85% because we're more comfortable with people making mistakes. But typically, once these bots are up and running, they're as close to 100% accurate as they can be. We always err on the side of a bot not processing something if it's not 100% sure. So the error rate is very, very low. A bot might process 90% of edocs and leave the remaining 10% for a person to do. It won't make errors on those 10% — it just won't process them.

In terms of getting started, it depends on the product. Some are easier to implement, some are more complex. On the easier side is our edoc executive — we can get that up and running in under 30 days. For our renewal reviewer, it takes about two to two and a half months to train that bot on your data set and get it deployed.

The process looks like this: we have a discussion to figure out what processes interest you and look at your internal processes. We get a really good understanding of how you're running a renewal review, see if there are differences from how the product works. Then we go through a design phase where you get to configure the bot to yourself. Using the renewal reviewer as an example, that robot will look for about 30 to 40 data points on every type of policy, pulled automatically. You as an agent get to determine which data points are important to you — which ones you want to be made aware of, which ones you want in a report, which ones you want flagged for your team. That design phase and configuration takes a couple of weeks. After that, we train your bot in an environment running on your data set while our engineering team watches everything it's doing, building in new rules and exceptions to make sure by the time it gets to deployment, it's working perfectly. Your team reviews some of those outputs during the deployment phase. Once we actually deploy it and it's running autonomously at that one to two month mark, it'll be processing everything in your system at a rate of 90 to 95% of items processed with an accuracy rate of 97, 98, 99%.

Host: One follow-up question. If I heard you right, you could actually take the physical downloaded document or attachment that comes with the download, and the bot can scan it and pull a piece of information off of it, compare that against what the downloaded information was, and fill in any gaps the download didn't pull? It can pull it off the doc and bring it over?

Jameson Fregeau: That's correct. It can run off the application data — the EDI data in the AMS system — and/or the actual policy data. We've got machine learning models internally to extract all of the data from the actual PDF documents themselves. Although EDI is getting better, it's still not 100% accurate, and people still look at the policy documents even if the EDI is 90% accurate because there are one or two things that are really important. So you can scan the data off the actual policy document, off the PDF attachment, and reconcile that with what's in the system.

Host: That's awesome. That PDF document is generally the best source of truth for all of this data.

Jameson Fregeau: Yeah, pulling the data off those PDFs is a pretty important functionality for these bots to know that the data they're interacting with is actually the right data.

Host: It brings up the whole clean data thing. You think about how much information you're filling in when you're prospecting based on what people are saying, and then you turn that prospect into a client. Sometimes in that process the data actually changes. You go from "I make $4 million in revenue" to "really I make $3,675,000" and that's what the policy is rated off of. If you still have the $4 million from prospecting, now you have a mismatch. Your policy is rated on the actual PDF at $3.6 million, but you have $4 million in your management system. Now you try to re-rate on a rewrite and there's this mishmash of data. The ability to say the PDF is source of truth, match that up against what's in your agency management system, see discrepancies, and fill in missing data — that whole argument around clean data feels like a major aspect of success moving forward. You can't do these automations if everything's a mess.

Jameson Fregeau: You hear it all the time — garbage in, garbage out. What we've found from working with agencies is that agencies' assets are their data. All they do all day is process information. You get information from carriers, you process that information, you give that information back to a customer. One core competency is really processing data, and to do that as effectively as possible, you need that data to be as accurate as you can.

Host: During that buildup phase where there's going to be some learning and some error, what does that look like? How does it work? I'm trying to take the scary out of this for people. The scary parts for me are one, I have no idea what this is — it seems very ethereal. We talked about that — it's a robot, a piece of code, acts like a human except it can do these repeatable tasks. You gave us the rules-based, high-volume, repetitive, digital. That makes sense. Downloads are a pain point for me. I can see how this could help. The next question people have is, is this going to be a nightmare once I get it going? Am I going to have more work for a period of time? Am I going to have to task an entire employee to make sure it's running correctly? What is the buildup time to acceptable accuracy — 99% plus? What does the work look like during that period for the agency? Are they getting a lot of call-out reports?

Jameson Fregeau: A few different things. When we first scope this and put a bot in place, we work with an agency's team. There's always one person who knows that process the best — usually the person actually doing that process. We work really closely with them to understand their specific process, how it might differ from another agency, what exceptions or differences they might have, before we actually put the software in place. We spend a lot of time upfront with your team.

Once we do that, we do a phase called supervised deployment. We put a robot into your system, first show you what that robot is doing so you can actually see it running. Your bot will be processing data, but it won't actually make any changes in your live system. One of our engineers watches everything it's doing. We ask the agency's team to review all of those outputs during that supervised deployment period. We do that for anywhere from one week to potentially three weeks, depending on how complex the data set is. Your team reviews all of those outputs. By the time it moves to autonomous deployment, you've reviewed all of it, your team is really confident with the results, and we only deploy once we have sign-off from our engineering team, our QA team, and your team that those outputs are exactly what you want.

End to end, we recommend people set aside about one entire day of work — about eight hours — to validate the outputs and work with us on the upfront process. It does require somebody on your team to work with us, but it's not extremely laborious because we take a lot of that work on our side. On an ongoing basis, we have internal spot checks and QA testing to make sure the bot is doing what it's supposed to be doing. There's a daily report that comes out from the robot, usually to the person who was responsible for this process previously. Once the robot is done with its run for the day, it sends out a report saying, "Hey Ryan, finished your download today. I was effectively able to process 95% of all these items. Here are the remaining 5% that still need to be managed by a human." The person goes in, looks at that remaining 5%, processes those quickly, and does that on a daily basis. It still requires a person to be involved for a very small percentage, but a person doesn't need to monitor it, start it, or stop it. It runs at a set time, processes all that information every day, and just notifies your team what's complete.

Host: I know you guys are just getting started, so you probably don't have hundreds of deployments out there. But I remember the guy who launched QQ Solutions, which is now owned by Vertafore — his name was Mark. About a decade ago, when I was working at TrustedChoice.com, I got to know him pretty well. When QQ first launched, it was revolutionary — the future of agency management systems. I asked him, "Mark, how come you're not in every agency under 20 people in the country?" He said when they first started selling, one of their pitches was that you could never have to hire another CSR, or this would replace one or two of your CSRs. He said they struggled at the beginning because the people whose jobs would be in jeopardy would immediately find every reason why this thing was going to blow up and wasn't going to work. They had to work around and find ways to say, "No, we're not recommending anyone get fired." I'm sure you guys run into something similar. If I'm the person who does all these tasks and that's my job, and now there are these bots that are going to quasi-replace me, do you get pushback? Do internal people look at this and start questioning it?

Jameson Fregeau: We've definitely dealt with that on a smaller scale. Typically, we'll go into these organizations and it's that person — the subject matter expert doing the downloads or the renewal reviews — who has a lot of doubt as to whether this robot can actually do the process. We usually spend time with that person upfront, really showing them what the bots can do, how they work, how we train them, and how we work with their team to make sure we get it exactly right for their organization. Ultimately, we need to build trust not just with the person buying the software, but with the rest of the organization because they're going to be the ones using it and interacting with the bot.

The second part is the replacing aspect. What we've found across brokerages and agencies all across Canada, the U.S., and really every other industry is that people can't hire enough staff. You cannot hire enough qualified, trained staff and retain them on an ongoing basis. People need more labor to continue to maintain their business and grow. What these robots allow you to do is take staff who have been with you for a long time, who are well trained, who know your business and insurance, and put them on much higher value processes instead of needing to hire other people. A bot eliminates the need to hire and train new people. We've never seen somebody actually lay somebody off because they probably have four or five open positions in their agency, if not more.

Host: That's a great way of getting after it. So what's beyond building productized robots for additional processes inside the agency? What's on the future for you guys? What excites you? To me, this feels much more like the future of how these tasks are handled. Virtual assistants will always be a part, but this feels like the jump — more than cheaper humans in other parts of the world. Where do you see this ultimately going? When you're sitting there fantasizing about what this can do in the future, what does that look like? What are some of the crazy problems you think you can solve with what you're building?

Jameson Fregeau: I think it's something that's becoming more apparent every day as we do this. A lot of people think of a future with robots walking around with people every day doing tasks. But what we've seen in the last couple of years is this huge shift to everything being online — work being online. All the work I do at Quandri is done from my computer for the most part. So I think what's becoming more apparent is that there is a future where those two things merge — robots doing certain things within this society. But what's much closer than having that in the physical world is having it in the digital world.

For us, what is really exciting is just pushing the limits of what we ourselves and what everybody else thinks these robots can do. There is so much within the insurance industry alone that we are going to sit and focus on for quite a while because there's so much value to add. But we want to continually push both what processes the bots are working on and the complexity of what these bots can do. Ultimately, we want to push into many other areas. Healthcare is an example where this can have massive impacts on people's everyday lives.

Host: I think that's really good. I think this is the jump. Other humans outside the country who work for less money is a good way to add scale to your workforce at a lower price point. But to really optimize your business, systematize it, and put it into that next gear — which allows the people you have to be as efficient and effective as they can be — a tool like a bot that can learn over time through machine learning is going to be way more effective, way more easy to manage, control, and make adjustments when necessary. The hope is you don't have to QA it as much. Once it's humming, you can give it the limited brain cycles you'd hope some of these processes would have.

I think the work you're doing is awesome. I wanted to have you on to shed more light on this topic. These digital workers — how you explain them — I've talked about VAs a lot on the podcast, and I still believe in VAs. That's not a knock. But the true optimization is going to come from jumps like this — jumping past cheaper humans to technology that helps augment what you do and do it faster and more effectively. The agency management system is a bugaboo for most people listening. Their agency is in some way in that system, and the data is a bugaboo. What you guys are doing seems like something people should consider. How do people get at you guys? Is it LinkedIn? The website? Are you taking new customers?

Jameson Fregeau: We're definitely taking new customers. We do have a bit of a waitlist right now because we have a lot of customers who want to sign up, so it takes a couple of months to kick off a new customer. But if anybody's interested, they can go to our website and book a meeting directly with me, or send me a message on LinkedIn. We'll get in touch and talk to people as soon as possible.

Host: That website is quandri.io. I'll have the links in the show notes. If you're listening and you forget, just type in Q-U-A-N-D-R-I insurance and it comes up. Our good friend Jeff Roy is using you guys. I know a couple other friends of the show are in talks or moving forward. I'm excited about what you guys are doing, and hopefully we can come back in the future and talk about all the cool, crazy stuff you're working on again.

Jameson Fregeau: Awesome, Ryan. Thanks a lot for having us. It was really nice to meet you. I enjoyed the conversation today.

Jackson Fregeau: Yeah, thanks so much. It was great being on the show. Thanks a lot, Ryan.

Host: Great to meet you guys. Cheers.

[Host's closing remarks begin] Host: Twice as many deals by this time next week. Sound impossible? It's not. With the One Call Close System, you'll stop chasing leads and start closing deals in one call. This is the exact method we use to close 1,200 clients in under three years during the pandemic. No fluff, no endless follow-ups, just results fast. Based in behavioral psychology and battle-tested, the One Call Close System eliminates excuses and gets the prospect saying yes more than you ever thought possible. If you're ready to stop losing opportunities and start winning, visit masteroftheclose.com. That's masteroftheclose.com. Do it today.