In a recent interview with Peter Ho, Managing Partner at the business advisory firm Brio360, Voltaire Founder and CEO Yo Sub Kwon discussed the technology behind accurate AI-generated claim letters. While artificial intelligence (AI) promises new efficiencies, claims leaders need to know if the technology can be trusted.
A full recording of podcast can be accessed here.
When Ho asked about the state of claims correspondence today, Kwon’s assessment was direct: the process is “riddled with problems.”
In their conversation, Kwon detailed the specific technical safeguards Voltaire uses to solve these problems. He explained how the right technology can eliminate risk for carriers, ensuring every letter is accurate and compliant.
Why Many Claims Letters Suffer from Inaccuracies
The risk in claims correspondence starts with the manual process itself. Adjusters must interpret unique policy documents that can exceed 100 pages, a task Kwon called “time-consuming and tedious.”
To manage this workload, adjusters often take shortcuts. Kwon explained that they “frequently use shortcuts like recycling old letters and editing them, which introduces errors.” A simple mistake, such as using the wrong name or citing incorrect policy language, can open a carrier to significant litigation risk.
How the Technology Ensures Accuracy
Voltaire solves this problem with AI that operates within strict, compliance-focused guardrails. It is not a generic tool. When drafting a letter, Voltaire’s AI “can’t cite language that isn’t in the insurance policy,” Kwon noted. This is possible because “we only send the insurance policy to the [AI] when we are looking for references to site.”
These guardrails act as the first layer of defense. The second is an internal quality check. Kwon described it as “a quality assurance system within that grades letter quality across many different dimensions to ensure compliance, accuracy and clarity.” Every draft is automatically graded before an adjuster ever sees it.
Preventing Costly, Unforced Errors
These technical guardrails have a direct impact on financial integrity. A single citation error on an otherwise valid denial can force a carrier into a costly, unnecessary settlement.
Kwon described how easily this happens. If a claimant’s attorney finds you cited the wrong policy language, “the carrier will just settle to…make that problem go away…even if they rightfully were denying the claim.” These unforced errors are a significant source of claims leakage. By ensuring technical accuracy, Voltaire’s technology protects carriers from this preventable financial drain.
——
For a demo to see Voltaire generate an accurate claim letter in 30 seconds live or even on one of your carrier’s templates, schedule time with a click.
Full Transcript
00:00:00 Yo Kwon
Within five years, no carrier is going to be writing letters manually. Currently, less than 1% of the market currently uses software like ours, so there’s this massive opportunity. And so we pushed to to go raise some funding to let us move faster essentially and to, you know, hire that the sales and marketing staff. We’ve been attending a lot of industry events. So that’s where a lot of capital is going.
00:00:22 Yo Kwon
We’ve already attended probably about 15 events this year and are just expanding our reach very aggressively to capture that market share.
00:00:31 Peter Ho
For BREO 360 I am Peter Ho, welcome to value drivers. Best use of Quang, founder of motor.
00:00:42 ANNOUNCER
Thanks for listening and appreciate it if you could take a moment to leave awaiting your review. Your feedback helps others discover value drivers and allows us at build.
00:00:51 ANNOUNCER
360 to keep delivering meaningful content, your support truly matters. Thanks for being part of the real drivers community.
00:01:00 Peter Ho
We have a special guest today. We have a fellow Virginia Tech alumni on the program. We have your subcon, the founder of motel. So water is a program that help insurance carriers by making claims correspondence faster, more accurate and cheaper with the help of AI. And this is obviously an interesting time because we’re getting into the hurricane season.
00:01:20 Peter Ho
So you’re so welcome to video drivers.
00:01:22 Yo Kwon
Thank you for having me.
00:01:24 Peter Ho
So would you mind giving me and the audience a bit about your background kind of the two-minute version of how you get to what you’re doing today? And I know you have done very much entrepreneurial, you know venture prior to vote here, if you don’t mind, give me and the audience like a a brief intro.
00:01:41 Yo Kwon
Sure. Yeah, too many version. I’ve been a technologist from a young Age, start programming at 12, launched my first company at 16. It was a web design firm. Since then I’ve.
00:01:53 Yo Kwon
Founded a number of different companies 2012. I co-founded launch key cybersecurity company that was focused on multi factor authentication. It was acquired and that experience kind of deepened my expertise in cyber security. I also got involved with blockchain early first with Bitcoin in 2010 and co-founded one of the earliest US based crypto exchanges that was also acquired.
00:02:14 Yo Kwon
Like Kraken and then the journey continued with Hosho company focused on smart contract security analysis. My entry into PNC Insurance came through AI when Chad LGBT rose in popularity, I began exploring real-world applications and stumbled upon someone.
00:02:32 Yo Kwon
Using our AI software that was completely unrelated to insurance, they were to using it to to write claims letters and I was curious. We interviewed them, their boss, and then that person’s boss eventually partnered with them, built a prototype, brought it to an insurance conference, got overwhelming positive feedback.
00:02:52 Yo Kwon
And that’s how Voltaire started in the insurance claim space.
00:02:56 Peter Ho
OK.
00:02:56 Peter Ho
So. So you you do not have previous experience.
00:03:00 Peter Ho
In the insurance industry.
00:03:01 Yo Kwon
Correct. Yeah. That what that was part of the reason we partnered with the the the person that Brian Brian Lane to to bring all that insurance experience because it would have been really difficult to to bring this product to market without understanding exactly like the the nuances of the problem that we’re solving.
00:03:21 Yo Kwon
And also the connections in the industry to like insurance carriers and that kind.
00:03:25 Peter Ho
Of thing got it and you have obviously as you mentioned worked through a number of entrepreneurial venture. How how does that help you that experience of you know starting and being acquired as an entrepreneur?
00:03:39 Peter Ho
Help you decide to focus on your current venture here at Volterra in terms of using AI to help insurance carrier to be more efficient. How does it help you in terms of designing? That is, you know a viable idea and and build a team.
00:03:53 Yo Kwon
Yeah. I mean, there’s.
00:03:56 Yo Kwon
So like each company I’ve I’ve built has taught me lessons. Some of them I’m still learning from and realizing mistakes are kind of sub optimal decision making I’ve had.
00:04:05 Yo Kwon
One of the things like find fascinating about the problem solving of building businesses is that there are innumerable things that you can do wrong and also quite a few paths of getting things right, although substantially fewer of those. And so you can make optimal decisions and get sub optimal results. So it’s like a very dynamic.
00:04:27 Yo Kwon
Kind of environment. Interestingly enough, what works great in.
00:04:29 Yo Kwon
The company might not work in the next, so there’s no universal playbook, but experience gives you know you better. Pattern recognition speeds up decision making. Let’s see. One major lesson is to find kind of strong partnership or resources in areas where you’re weak. I’m not particularly good at like marketing, for example. And so that’s like a area that I always like need to make sure that like is being covered.
00:04:51 Yo Kwon
Well, with someone I trust, another is to clearly define the problem you’re solving like very early on and ensure customers are willing to pay for the.
00:05:00 Yo Kwon
Without that, it’s like difficult to get a business too far off the starting line. One thing that’s helps given from my previous ventures is that I’ve met a lot of like extremely smart, you know, skilled people along the way. And I’ve been able to bring those into subsequent businesses. So for example, the engineering.
00:05:19 Yo Kwon
On this business is is, you know, very cutting edge. It’s, you know, like all AI. And I wanted to have extremely talented engineer.
00:05:28 Yo Kwon
There’s, and I knew some from previous companies that I built and was able to pull them in. So I think that the experience I do have in specific industries or topics can come in handy. For example, I have a lot of cyber security experience and building like highly scalable enterprise software, things like building secure infrastructure or dealing with like compliances.
00:05:49 Yo Kwon
Are pretty trivial for me now, whereas like I’ll run into challenges in areas that have no experience that you know could probably be.
00:05:57 Yo Kwon
Easily dealt with by an expert. So I think that, you know, I think you should try to become an expert very rapidly and whatever you’re doing business in. But the my pre, I’ve been able to leverage a lot of the experience I’ve I’ve had over my previous businesses.
00:06:11 Peter Ho
Great, great. So I mean, although there’s no kind of one-size-fits-all playbook, right? I mean I I wish there is such.
00:06:17 Peter Ho
People available, but obviously I totally agree with you that there’s no such, you know, nice menu you can read and then just apply it to all businesses. It doesn’t work that way, but it sounds like building strong team and defining the problem in in where we, you know clearly manner is important.
00:06:34 Peter Ho
Now for the insurance industry. What kind of problem do they face? Kind of come back, I mean, more focus on what you’re trying to solve right now. What, what are core problems? I mean you are, you know, you have identified and and you’re trying.
00:06:46 Peter Ho
To solve right now.
00:06:47 Yo Kwon
Yeah, claims correspondence is is riddled with problems. It actually funny enough, like when we were, you know, showing kind of like our prototype initially to insurance carriers, they were really excited by it. And you know that we we took that reaction is very positive but we it.
00:07:05 Yo Kwon
Because actually a little bit longer to dig into exactly why they were so excited. And the reason why it was kind of different for different people depending on which part of the department or different occupation they had within the kind of the claims cycle. So like adjusters, they manually write letters which is time-consuming and tedious.
00:07:26 Yo Kwon
So if an adjuster saw the software, just the idea that like this part of their job that they kind of hate doing could be, you know, significantly speed up or, you know, mostly automated.
00:07:37 Yo Kwon
The reason why it’s difficult is because doing it correctly means interpreting like a unique, often 100 plus page policy document or substantially larger than that, and they’re unique because a common based policy could be modified by a special provision and any number of endorsements that make modifications to that base policy. So they’re the combination of potential.
00:07:58 Yo Kwon
Like policies is is you know.
00:08:00 Yo Kwon
Tremendous and adjusters frequently use shortcuts like recycling old letters and editing them, which introduces errors, and this opens carriers to litigation risks, especially if the wrong language or names are used or, you know, just or if they’re citing the wrong policy language from a the incorrect policy, mistakes often go unnoticed. But with AI, those errors become.
00:08:20 Yo Kwon
Easier to spot, like on the like recipient side, increasing the liability and so the attorneys and the carriers could get excited by that.
00:08:30 Yo Kwon
In most denial letters that go out to customers have some kind of issue with them, and so they might not always create legal issues, but they’re certainly not what insurance carriers would consider correspondents that they’re proud of or, you know, like, clearly inaccurately conveys their message. So Voltaire, you know, we we improve the accuracy we inconsistency.
00:08:49 Yo Kwon
Of the letter writing like dramatically, and it can’t cite language that isn’t in the insurance policy. So that alone helps a lot. And then carriers have these complex, multi layered ways of.
00:09:00 Yo Kwon
Performing kind of quality assurance QA to try to reduce their rate of error. At least you know like reduce the egregious errors and this can involve multiple layers of review like first by the you know the adjuster themselves and the claims manager, then maybe a supervisor or QA or an attorney even above that.
00:09:20 Yo Kwon
And so just a lot of man hours can go into the review and editing of these, these these letters. If there’s like a problem at one stage it gets kicked back, you know, and it they have to redo it and then it goes back up the chain again.
00:09:33 Yo Kwon
And so, like, claims, managers get excited by like the consistency increase the supervisors get excited because they’re like, OK, there’s fewer bad letters. We’re gonna come our way and then, you know, the process people and the like, VP’s, they get excited by, like, the cost savings, you know, just like the number of hours spent.
00:09:54 Yo Kwon
In their organization is is reduced.
00:09:56 Peter Ho
So basically the current process is very manual and require a lot of knowledge from the adjuster, right? So when he looked at the specific case, I mean obviously he need to understand the policy and then he.
00:10:08 Peter Ho
To you know, respond accordingly in in the letter and have to drive it up and obviously take a lot of time to do that and voter can basically, you know, ingest a lot of the policy information and, you know, have the right compliance structure in place and then provide, I guess maybe an automated letter for the adjuster.
00:10:28 Peter Ho
Coming straight from the system is is is that. Is that correct?
00:10:31 Yo Kwon
Yeah, that’s correct. It’s there. There’s different types of letters that are written, you know. So there might be like a reservation of rights letter or, you know, a payment letter, that type of thing. But if it’s like a denial letter, those are the letters that take the longest. And they’re kind of the trickiest because they.
00:10:45 Yo Kwon
They have, or even a partial denial like, you know, they’re like, oh, we’re, you know, gonna pay out on your roof. But we’re gonna, you know, deny your pool cage because it’s, you know, not listed in your policy even if you are correctly denying the claimant because it’s not covered in their policy if you are, you’re you’re legally required to cite verbatim from the policy. Why?
00:11:06 Yo Kwon
It is. You’re denying that, and if you don’t, you cite it from a different policy, like a a carrier in that instance will just like if the the person the claimant brings it up or, you know, involves an attorney, the the carrier will just settle to like, make that problem go away because they, you know, they they know it’s just gonna be like a you know an.
00:11:26 Yo Kwon
A litigation effort, even if they rightfully were denying the the the claim. So it ends up being like a a pretty large cost in in insurance carriers on you know an annualized basis to to deal with all the claims that were screwed up essentially.
00:11:42 Peter Ho
So I mean, from a pure expense standpoint, I mean I I guess the way I think about it.
00:11:46 Peter Ho
The the adjuster can do a lot more cases or maybe the other way to think about it is I mean you don’t need less adjusters to to to perform or to be built the same amount of cases right? So the efficiency of this workflow will be increased quite dramatically because I mean the computer is helping you out giving you the special.
00:12:06 Peter Ho
Well, basically to to to make it a lot faster.
00:12:09 Yo Kwon
Yep, that is the way some carriers think about it is that the adjusters can spend more time being careful, diligent, they can, you know, go home at you know 5:00 PM and they’ll have to work late like they can.
00:12:22 Yo Kwon
You know it. Make it just like a better experience for, for, for them those retention is also really low for adjusters because many of them have like poor, poor working conditions.
00:12:35 Peter Ho
Supposed got it. So I guess I mean the workflow is is very tedious. I mean as as we speak today, so the.
00:12:35
So.
00:12:41 Peter Ho
The system will provide make it I guess, take out the the tedious part and the less fun part of the workflow out out of the the the the schedule where they just let them to focus on higher.
00:12:55 Peter Ho
To add activities for for the company and then how do you ensure the the accuracy and and compliance because obviously being efficient is important, but more important may maybe is making sure the the claims respond is accurate right? Like as you mentioned earlier there are layers of review.
00:12:57 Yo Kwon
Correct. Yeah.
00:13:15 Peter Ho
In place, right? How do you?
00:13:16 Yo Kwon
You.
00:13:17 Peter Ho
You know with.
00:13:17 Peter Ho
Hotel make the accuracy I mean within at all certain tolerance, right?
00:13:23 Yo Kwon
Yeah. So you know, like we use AI, we use large language models, same kind of technology that powers Chatty BT, but with our own technology stack prompting and it’s outfitted specifically to deal with, you know, these kind of challenges, right, the challenges that that adjusters deal with and also with accuracy. So one of the things that.
00:13:42 Yo Kwon
We built was.
00:13:44 Yo Kwon
A quality assurance system within that grades letter quality across many different dimensions to ensure compliance, accuracy and clarity. Another thing is that the.
00:13:54 Yo Kwon
We only send the insurance policy to the to the LM when we are looking for references to site and then we check that any language that is cited is, you know 100% verbatim quoted from from their from their policy. So yeah. Critically, when the same policy.
00:14:14 Yo Kwon
Language system only uses the relevant insurance policy and then it pulls that cited language adjusters. Can we do tell adjusters that like you know they are the ones that have to make the decision?
00:14:25 Yo Kwon
Of what they’re going to deny, but our software can help them find the language that’s that’s relevant in their in the within the policy.
00:14:32 Peter Ho
Policy garlic, garlic and are there other companies trying to do similar things as what hotel is doing now? Who who are your competitors?
00:14:41 Yo Kwon
So I was just at a insurtech insights yesterday and the day before that, which is a a big insure tech conference in in, in New York and I you know I was, I was walking around the floor, you know you see AI everywhere, probably 80% of the companies that are doing a.
00:15:00 Yo Kwon
I I there really isn’t anyone doing letter writing. From what I’ve seen so far, there’s like a couple of companies that it seems like they’re it’s not their main focus, but they, you know, potentially could be moving into that direction with like some of the examples that they’re showing of like, you know, the possibilities. So I kind of I feel like it’s.
00:15:20 Yo Kwon
Probably like right on the verge of.
00:15:23 Yo Kwon
A bunch of companies, either startups or, you know, some AI companies kind of like entering into that this arena because it’s a very obvious use case, like when we go to, when we talk to, you know, anybody in like a a claims department all the way up to the CEO, the the Chief claims Officer they you know they’re they’re like oh wow, I was waiting for this.
00:15:44 Yo Kwon
Product to you know, to come to life or I ohh, I had this idea myself, you know and.
00:15:48 Yo Kwon
So it’s it’s a pretty obvious idea to if you know anything about like the the latest advances in AI or if you’ve ever used chat, TVT, and you’ve ever written a claims letter, it’s a pretty, you know, small leap to make to like, ohh. If I combine these things like you know it, it should be able to make this process more efficient. But surprisingly we’ve seen.
00:16:08 Yo Kwon
Very little on the competitive front.
00:16:10 Peter Ho
Got it. But some of these are insurance carriers. They have pretty good scale and size. I mean, and also technology infra.
00:16:19 Peter Ho
Structure, wouldn’t they consider doing some of that in house as well? I mean, if they have a a good technology team, some of the carriers maybe considering doing some of that in, in housings of outsourcing it to a third party like motel.
00:16:34 Yo Kwon
I I mean, there certainly are insurance carriers that will try. I think that generally the way that these entities, you know the insurance carriers are generally.
00:16:44 Yo Kwon
Pretty large and the their technical teams are not particularly strong and in most cases they’re actually like, you know, they kind of like bundle the technology teams with their IT teams and they’re like heavily backlogged like you know way behind schedule on integrations of other things that they want done. So I don’t really see.
00:17:04 Yo Kwon
That being the case for like the vast, vast majority of insurance carriers, maybe like some of the the very the very top top level insurance carriers, they could consider it. I don’t think that they’ll do it very quickly though if they do.
00:17:21 Yo Kwon
It’s just not like a space that seems very quick to innovate it like a lot of these insurance carriers are using, you know, software that that you know is 15 plus years old.
00:17:32 Peter Ho
Yeah, no, make makes sense. I mean the sometimes, I mean the the scale but also we come with the bureaucracy and the inertial to do things the old way, right? Yeah, they they they probably do have a sizable technology team, but they already have a lot of backlog to just to deal with the existing.
00:17:49 Peter Ho
We’re close, so I agree with you now. You mentioned something pretty interesting. I mean some you quantify some pretty interesting stats in the press release. I mean, can you kind of walk us through in, in the press release you mentioned there are some pretty interesting metrics that show that Votel can really help the efficiency of the insurance carrier.
00:18:10 Peter Ho
And create value. Can you kind of talk us through?
00:18:13 Peter Ho
What I mean how? How do you, how should we think about those metrics? Sure.
00:18:16 Yo Kwon
So I mean, just from a time saving standpoint, that right alone is quite valuable and you know it kind of depends like you know what the adjusters are are paid and like if they’re you know using that time for for other things. But generally if like it might take them let’s say 30 minutes or to two hours to write like you know one of these letters.
00:18:37 Yo Kwon
And that that can be brought down to, you know, 30 seconds. And then like maybe a, you know, a few minutes of review and then on top of that the reduction of.
00:18:46 Yo Kwon
QA that’s needed from the, you know, their managers and the supervisor supervisor. So that’s on the time savings front on the the other, the other big front is on the reduction in litigation and also something called claims leakage where you know if the sometimes adjusters make mistakes they will accept.
00:19:07 Yo Kwon
You know, coverage decisions for things that actually don’t have coverage or they will try to do the opposite and then they’ll.
00:19:14 Yo Kwon
Sued, you know, they’ll deny for things that they shouldn’t be denying for, and then, you know that can that opens up, you know, can of worms for them. And it’s a lot more difficult to do that with our software if, like, you can’t, if you you try to cite language that doesn’t exist in the policy it it, it can’t do it. So and that has happened before with our software where someone.
00:19:36 Yo Kwon
You know, and just just trying to cite certain language and they are unable to do so and they’re like file a ticket with us telling us that there’s like some sort of problem. And we’re like, show us the policy and they’re like, ohh, it’s actually not in there. Never mind.
00:19:49 Yo Kwon
So on the the claims leakage side on the reduction of litigation, that one’s a little bit more, you know, long winded or more you know you have to think a little bit more ahead on on those savings. But between the the time savings and the reduction in cost to due to the increased.
00:20:09 Yo Kwon
Accuracy. It’s it. It’s a quite a substantial.
00:20:12 Yo Kwon
Improvement to the to the carriers bond.
00:20:14 Peter Ho
Next.
00:20:15 Peter Ho
Line and how do you charge the the carrier? Do you charge them on the number of cases they use this material they charge them the number of users. What what is your business model?
00:20:26 Yo Kwon
It’s on a per claim basis, so they can write as many letters as they want for using our software for that claim, but they because there are a lot of times multiple like back and forth, it’s not just like you know, selling it might might be the last letter that goes out or the payment letter, but the there might be letters that happen in between acknowledgement letter or.
00:20:48 Yo Kwon
Reservation of rights or, you know, things like that in.
00:20:51 Yo Kwon
And the reason we charge on a per claim basis is because like a lot of the carriers that we support currently, they deal with catastrophes like in the Southeast, they have reinsurers and they can kind of pass on those costs to the reinsurers if they’re on a per claim basis, sometimes depending on, you know like kind of their agreement with the their their reinsurers.
00:21:10 Peter Ho
And I know you and your team have recently closed around funding $4.2 million seat funding round.
00:21:17 Peter Ho
And how do you plan to use the the capital I mean for for the near term, I guess building on your team?
00:21:22 Yo Kwon
Yes, we’ve yeah built out the team. We hired a marketing person and a salesperson. We are, you know, hiring a couple other people now as well. On the customer success side to manage the new accounts that we have coming in, we we had, you know even before the the seed round, we had raised some pre seed.
00:21:43 Yo Kwon
Around.
00:21:44 Yo Kwon
Money. Not too long before that, we still have plenty of runway left, but what we realized was the biggest bottleneck to adoption was awareness. You know, carriers, they love the product it, you know, unequivocally benefits them, you know, saves time, reduces costs, lowers risks, but they don’t know that we exist. So.
00:22:04 Yo Kwon
Most of them don’t. So in internal discussions with our team.
00:22:08 Yo Kwon
You know, like we we believe that like within five years like no carrier is going to be writing letters manually, but currently less than 1% of the market currently uses software like ours. So there’s this massive opportunity. And so we pushed to to go, you know, raise some funding to let us move faster essentially.
00:22:28 Yo Kwon
And you know, hire that the sales and marketing staff, we’ve been attending a lot of industry events. So that’s where a lot of capital is going. We’ve already attended probably about 15 events this year and are just expanding our reach very aggressively to capture that market share early.
00:22:43 Peter Ho
And how was the funding?
00:22:46 Peter Ho
Experience. Was it smooth? I mean, how long does it take you to to close the the seed funding round?
00:22:53 Yo Kwon
It was pretty quick, like we decided I think in like December that we were going to actually go and try to fundraise and we had most of the money locked up by the end of January.
00:23:07 Yo Kwon
I think we raised roughly like half of the round in, you know only a few days and then kind of like looked for more strategic partners and stuff that would, you know be able to do things like facilitate introductions to carriers or like, you know, kind of help us out in in other ways.
00:23:22 Peter Ho
I mean, what what’s the driver for the basically quick closing? Is it because the proof points that you were able to explain to the investors or maybe maybe your some relationship you might have with your prior ventures, what what are the key I mean in your opinion to a fairly smooth clothing, because I’ve heard like different stories.
00:23:39 Yo Kwon
Yeah, I.
00:23:42 Yo Kwon
That that was that you?
00:23:43 Yo Kwon
You touched on them like I we only really. We didn’t do like, kind of like what you would expect, I guess like in a like the a larger fundraise where you like, you know, reach out to all the VC’s and have the, you know, the VC meetings or anything like that. We basically reached out to, you know, investors that we knew directly. My my Co founders and me and just.
00:24:02 Yo Kwon
You know, kind of told them what the idea is. A lot of them we, you know, just kind of sent, you know our.
00:24:08 Yo Kwon
Like I had a a perspective like a business case. Kind of just explain like what the business is like, what you know, like where we’re at and like, what what our current financials are customers you know, just like kind of, you know, all the information and a lot of them, the investors just you know, kind of made a decision off of that and we you know got it done pretty quickly.
00:24:26 Peter Ho
Great.
00:24:27 Peter Ho
Great. And then how how do you think about the current AI hype cycle? I mean or AI cycle rather?
00:24:35 Peter Ho
Obviously, I mean, you know your last, you know, 2-3 years, right? I mean that is kind of the the focus for many technology investors. I mean that is basically the the, the new wave that driving a lot of the investment pieces right. This is basically the new mobile or new Internet, maybe even bigger than the prior waves of these.
00:24:54 Peter Ho
You, you know, with full technology where where do you think we are. I mean today when we look at it, do you think we are like in the early stage kind of in the mid stage or kind of towards the end how?
00:25:06 Peter Ho
How should we think about it?
00:25:07 Yo Kwon
My opinion is that we are in an early stage in the sense that the the world is going to change quite drastically. I think in over the like the next 5 to 10 years the the pace of AI advancement is extraordinary. It’s on a parabolic trajectory.
00:25:27 Yo Kwon
And like so, for example a a couple of months ago we gathered the team, which is fully remote. Otherwise in Montana for an AI hackathon and we encouraged everyone, technical or not, to use AI to build something that would be helpful to them with their regular work. Some amazing applications came out of that hackathon.
00:25:47 Yo Kwon
And more importantly, it kind of like opens people’s eyes, you know, on on my staff for for for non engineers like to understand that AI.
00:25:56 Yo Kwon
As unlocks automation in ways that would have been previously only possible with the help of like a software engineer, and that’s like the enabling the enablement there is, it’s like it massively increased the impact that each individual within our organization is able to affect and is the speed at which they can do so and so empowering the team with AI has been one of the.
00:26:18 Yo Kwon
The most like transformational moves we’ve made and I think you know, we’re a start up where we can be a little bit, you know faster and you know more nimble about those kinds of things but.
00:26:27 Yo Kwon
That’s. That’s where that’s like, kind of like the next Phase I see happening just across the board is that, you know software engineering is is already changed, right? Like you you read about, you know Amazon or Meta or these companies where they’re, you know, they’re outputting so much more code at, you know, higher quality with, you know, significantly less, fewer.
00:26:47 Yo Kwon
Few.
00:26:48 Yo Kwon
Were hours spent by by engineers. A lot of it’s trend, you know, already transitioned from engineers writing code. So now just more reviewing code and kind of like guiding AI along. And I think that, you know like the this company Voltaire is you know, automating a lot of like you know more of the boring, tedious processes and that’s just.
00:27:08 Yo Kwon
On a a trend to continue for every company like mine, there’s, you know, thousands of companies like mine that are doing it for very niche specific.
00:27:15 Yo Kwon
Like use cases and automating things that are you know can be automated. And I think that most of remote work is going to be like within 5 to 10 years is probably going to be automated to a pretty heavy degree and it’s gonna drastically change like a lot of things about the world I think.
00:27:33 Peter Ho
Yeah, no, I think totally agree. I mean, you know, from what some of these CEO’s were saying, Microsoft matter, I mean, 20 percent, 30% of the code base now are generated by.
00:27:45 Peter Ho
Die and and then from the small company side. I mean, I’ve heard, you know, you you probably know. You know, there’s some of these unicorns are built by 15 employees or something like that, which is kind of scary. I mean, in, in, in the old days or maybe even go back five years. I mean, you want to build Unicorn upon it, you know, thousands.
00:28:06 Peter Ho
I mean at least hundreds of employees.
00:28:08 Peter Ho
Right now we’re talking about Tiny team that can basically build something so quickly in a matter of, you know, quarters in instead of years, which is quite, you know, remarkable. And a lot of that obviously is coming from AI, right. So so you think the the AI cycle is still early, there’s still plenty of Greenfield for folks to to explore.
00:28:27 Yo Kwon
Yeah. I mean, especially now.
00:28:29 Yo Kwon
Well, given that you can be like a very small team or even a solo entrepreneur and build like, I mean some of the apps that like we’ve built or that I’ve even built like over like a weekend just to like help out with something it are things that would have taken me six months or something, you know in the past previous to LM’s to to build. And there are things that.
00:28:51 Yo Kwon
Like 1 issue that we’ve like run into is like formatting from like you know PDF documents and putting in it’s like a Word document with the the letters and that was the.
00:29:00 Yo Kwon
Actually like an unsolved problem in computer science, I feel like for a long time and now it’s just like you have a I kind of figure it out and and a lot of those those challenges are are just getting easier and easier to solve.
00:29:13 Peter Ho
Now then how? How do you and your team keep up with all this latest development? Right. So it is a way. It’s basically right, everyone. I mean, you know, in industry is probably in touch with you know, the latest version of the model and the latest capability things are moving at a pace that you know quite faster than ever. How do you and your team?
00:29:33 Peter Ho
Keep up with all this like you know, latest development. Do you? You keep reading stuff every.
00:29:36 Yo Kwon
Yeah, I I think.
00:29:37 Yo Kwon
It’s.
00:29:39 Yo Kwon
I yeah, I I do read a lot of like, you know, what’s happening. And I think really in order to fully understand a lot of the advancement you have to use it like you have. And so like, we have, you know, subscriptions to like pretty much every major foundation model for, you know, our our team and, you know, encourage them to be using and trying things out.
00:29:58 Yo Kwon
That we carve out 25% of the time for our engineering staff to just be experimenting and like testing out models and you know playing around with AI, researching new things about AI. And then we also, you know the hackathon I mentioned, we wanna make that like a.
00:30:12 Yo Kwon
Regular thing where we, you know, just allow everyone to to try to like, build and automate. You know, anything really that could be beneficial to the, to the business. I think that that you really can’t keep up with everything. It’s just it’s moving too fast. But if you wanna like ride the wave, you know at least.
00:30:33 Yo Kwon
Trying out the latest model, the latest models.
00:30:36 Yo Kwon
Is is important because you like whatever is best practices now is no longer gonna be best practices in a month from now when there’s a new model and it’s like a lot of the things that you have to do to like kind of tweak or trick, you know the the model to get it to do whatever it is that you want to do. It just becomes even more intuitive and like you can and even.
00:30:56 Yo Kwon
More plain English or even, you know, fewer instructions. Get it to to to deliver the outcome that you’re looking for.
00:31:02 Peter Ho
How how do you think about these big players who are building foundational models, right? You know, matter? Open AI, Google and and Microsoft. I mean, do you think like open AI is is leading?
00:31:14 Peter Ho
Or how how?
00:31:15 Peter Ho
How do you think? I mean, you know, are they gonna be like 1 winner in in in the end in terms of the the core model, I mean how how, how should we?
00:31:23 Peter Ho
Think about that.
00:31:24 Yo Kwon
So it’s it’s kind of interesting. I do think that there is like a race that’s happening and that there’s going to eventually be like this model that you know, like they already have like some like self improvement like methodology, right? Like with the latest reinforcement learning models where they don’t really need to necessarily have a human in the loop to.
00:31:45 Yo Kwon
Said.
00:31:46 Yo Kwon
Increased its quality and the the the rate at which it’s getting better. So if that happens in like a real like meaningful way without like there’s restrictions obviously like things like hardware but that would give whoever gets there first like the significant edge where their model just.
00:32:05 Yo Kwon
And you know just spirals and continues to get better and better and better, but.
00:32:10 Yo Kwon
The models that exist currently are already so good, and they’re so capable that there is there’s not that much of A Moat anymore in the sense that, like you can go to hugging face and download, you know, like a a very powerful model, run it on, you know, your own hardware.
00:32:29 Yo Kwon
And that becomes, you know, significantly cheaper for better performance all the time.
00:32:35 Yo Kwon
And because of that, I think that, you know, people can probably like there is going to be a point where the intelligence is good enough to to do a lot of the things that you might want to do and you don’t need to use the foundation model necessarily to to do it. But for newer use cases and for more advanced things.
00:32:55 Yo Kwon
And or the things that people you know expect to be coming, we still.
00:32:59 Yo Kwon
Need to rely on.
00:33:00 Yo Kwon
Kind of like these continuing advanced advancements in in the foundation models, but a lot of those advancements will trickle down to like these open source models as well. So I think that it’s, I don’t know, it’s it’s kind of a really difficult landscape to to fully comprehend and predict what what’s what’s gonna happen.
00:33:19 Peter Ho
Yeah. No, I mean, when, when, when we look at the the size of the investment that these companies are putting into, we thought about, you know, multiple 10s of billions of dollars in in each quarter. And I think to to your point, I mean where where the curve get to a certain point, it gets steeper and steeper.
00:33:39 Peter Ho
And and I think some some of these players are are just nervous that whoever get to that point of the curve first, then they will win and capture the market, right. But in in a way that’s quite good for the application developers like you know there because I guess you’re indifferent in terms of the foundation model, right? You’re gonna use whatever is the best.
00:33:58 Peter Ho
The moment to.
00:34:00 Peter Ho
Buy the solution to your customers, right?
00:34:02 Yo Kwon
Yeah, yeah, that’s correct.
00:34:03 Peter Ho
So you’re so thanks again for joining us today. So we have time for one final question before I get there. I would like you to share with me and the audience books or podcast that you work recently or you enjoy. I mean just for for leisure.
00:34:20 Yo Kwon
I don’t really listen to any podcasts, honestly a the a book I read recently or reread is the kind of like Christo, A phenomenal book.
00:34:31 Yo Kwon
I’m also reading Great Expectations, but with Chachi BT I’ve been able to dive deeper into like the historical context like while I was reading kind of my christo like, what was the situation in France when the original readers were reading that book? What you know, like all the details like in French history, that they would have been acutely aware of like.
00:34:52 Yo Kwon
At the time to kind of get like a a a deeper, like richer understanding and like appreciation of like, you know, the book, the book and and the time that.
00:35:00 Yo Kwon
Reflex.
00:35:00 Peter Ho
It’s interesting. So when you read the book, you basically use ChatGPT to kind of like help you dig deeper into what you’re reading.
00:35:10 Yo Kwon
Yeah, I would. I would keep it in voice mode. So I would just be reading and then I would see like this thing about like Napoleon or something or like, you know the year or like you know 1815 and I’d be like, you know, tell me what was happening in France in that year and like what, you know what? Where was Napoleon and like, had he already left the island or, you know, that that kind of thing and would give.
00:35:29 Yo Kwon
You know the description and.
00:35:32 Peter Ho
Well, that’s great. I mean, never thought about that again. I mean learning something new everyday. Thank you. So finally, you have been in entrepreneur mode for some time, OK, even before Voltaire. What kind of special sauce or key lessons do you use to motivate yourself and your team?
00:35:51 Peter Ho
You know, building companies is not.
00:35:54 Peter Ho
You know, it’s not very easy, right? What are the things that you use to keep yourself excited and keep your team, you know, aligned and marching towards the ultimate goal of success? Hmm.
00:36:07 Yo Kwon
Hmm, I’ve I’ve never really had too much problem on the motivation front. I think that the the act of building a business itself kind of like strokes.
00:36:19 Yo Kwon
That that, like need of or that interest like that’s like that is my interest is like problem solving and it used to be I used to do it with programming. I guess I still sometimes do stuff around that but building a business is a significantly more challenging like problem solving exercise and.
00:36:39 Yo Kwon
I think that we’re in this time period right now.
00:36:42 Yo Kwon
You know, with with AI and you know, we discussed the the I feel like it’s very early and so being involved in something that is going to have this like tremendous impact on humanity while it’s from the inside and seeing it all happen I think is is exciting. And I think that you know the.
00:37:02 Yo Kwon
Plays a large part in. You know what brings me, you know, brings me to work.
00:37:07 Yo Kwon
Every day.
00:37:08 Peter Ho
Got it. So.
00:37:09 Peter Ho
Basically what you’re saying is, I mean, obviously this is a special moment, right? Being in the arena, you know, playing with this technology, helping using this technology to make everything quicker, faster and easier for for the customer is.
00:37:22 Peter Ho
It’s a good challenge and it’s a, you know, useful motivation for yourself. What about your team? I mean, how, how do you get them excited?
00:37:30 Yo Kwon
Yeah. I mean, a lot of the team is, you know, people I’ve worked with previously. And then also I I I specifically look for people and try to build into the company’s culture people that are excited about AI and what’s happening with it and you know want to to join a company that is AI native from the core.
00:37:50 Yo Kwon
And and so there’s always people talking about, you know, things they’ve seen and news they’ve read and discussing that, you know, and how maybe like a new model or some sort of new technique might help us with our with our business. And so I I think the the team is very centered.
00:38:10 Yo Kwon
Grounds like that excitement as well of like, what’s happening in the world with.
00:38:15 Yo Kwon
AI and all the advancements that are being made and they also enjoy being involved in in a company that’s that’s doing something with that.
00:38:24 Peter Ho
Right. So, so thanks again for spending time with us today. And if the audience want to connect with you or if they want to work for Voltaire, what’s the best way to to find you?
00:38:35 Peter Ho
And learn more about Voltaire.
00:38:37 Yo Kwon
Let’s see, you can connect with me on LinkedIn. Yo sub kwan. Voltaire dot claims is our website. So you can go there. I think we have, you know, job postings there as well and.
00:38:51 Yo Kwon
Yeah.
00:38:52 Peter Ho
Great. Please keep us in the loop on how you and your team are doing. Would love to have you back in the future and I’m I’m sure the the state of AI will be quite different if we speak again in a few months.
00:39:03 Yo Kwon
Sounds good. Thank you for having.
00:39:04 Yo Kwon
Me. Thank you.
00:39:05 ANNOUNCER
Thanks for listening. We’d appreciate it if you could take a moment to leave awaiting your review. Your feedback helps others discover value drivers and allows us in visual 360 to keep delivering meaningful content. Your support truly matters. Thanks for being part of the value drivers community.
