When AI Gets Pricing Right: Building Trust Through Data Integrity with Carly Gunby

In this episode of On the Move, Jen sits down with Carly Gunby, Vice President of Revenue at Transfix and a finalist in TMSA’s AI & Technology Showdown at ELEVATE. 

With more than a decade of experience in transportation and logistics, Carly offers a preview of how privacy-preserving, custom AI cost models are helping freight brokerages price more accurately and operate more efficiently. Rather than relying on aggregated market data, her approach focuses on validating and training models using each broker’s own shipment history to address common distortions like seasonal bias, carrier skew, and volume imbalances. 

Carly shares insights into the business challenges this innovation solves and the impact it can have on margin performance and RFP efficiency, while saving the full workflow and demonstration for her live keynote at ELEVATE. offers a practical look at what’s really being said, and why it matters more than you think.

Check out the Transportation Sales and Marketing Association (TMSA) website or engage with us on LinkedIn.

Transcript

Welcome And Guest Introduction

Jennifer Karpus-Romain 0:22

Hello everyone, and welcome to On the Move, a show where we share transportation sales and marketing success stories. I'm Jennifer Karpis Romain, Executive Director of TMSA, a trade nonprofit educating and connecting marketing and sales professionals inside transportation and logistics. And today on the show, I'm excited to have Carly Gunby, Vice President of Revenue at Transfix. How are you doing today, Carly? I'm doing well. Thank you for having me. So excited to have

Carly’s Path From Brokerage To Tech

Jennifer Karpus-Romain 0:52

you on the show. Can you start by sharing a little bit about your role at Transfix and your journey in the industry? Sure, happy to.

SPEAKER_00 1:00

I like most, I feel like in my generation of freight, didn't have a supply chain class or the opportunity to attend it. Probably because I liked to talk and liked people. I found a natural knack for sales. And so dived headfirst into logistics at Crowley without any formal knowledge of it here in Jacksonville, Florida. Spent years in varying roles at Crowley. It was a company that did full supply chains, so Ocean and Pull Point Distribution and LCL and LTL, all this stuff, mostly in sales, a little bit of customer success, but spent the last few years there on the brokerage side, learning operations and just kind of better understanding where the value comes from a shipper's perspective and using a broker. And so I was there for probably seven or eight years total before taking a leap of faith over to Transfix. Then it was a bit of a startup. It was still really kind of interesting in terms of digital freight and very new. So I spent, you know, years doing that at Transfix when we were a broker. Almost two years ago, we divested our operations to NFI. And so today I lead commercial strategy across two products, our cost models, which is our pricing intelligence, and our TMS built for freight brokers and three PLs. And so my whole job today is helping brokers protect their margin and win more freight, which turns out to be exactly what I was trying to do when I was a broker. I love that. I mean, yeah, that's the goal, right?

Jennifer Karpus-Romain 2:41

We're all in it. Yeah.

Why AI Beats Stale Market Data

Jennifer Karpus-Romain 2:43

And you are really leveraging AI to improve freight pricing and revenue strategy. So I'm curious, like as you were doing those shifts from brokering to transfix, as you're getting into more development, as you guys are focusing, what really initially sparked your interest in really leveraging AI in that capacity?

SPEAKER_00 3:05

Yeah. So tribal knowledge is something most brokers pinge themselves on. But in practice today, it's, in my opinion, it's a liability disguised as experience. We need everyone to be on the same page. And look, every broker is guilty of doing this, including when we were a broker, but we needed to systemize it in order to improve our buying power. And so when I was on the brokerage side, you'd price a lane based on market data that by the time the broker, by the time the load had hit your screen first when you priced it, it may have already been stale. And so moving into free tech and realized that the industry was largely doing this just maybe with a fancy dashboard or some kind of like aggregated data on top of it, AI became interesting to transfix and specifically to me, when I saw that it could ingest more signals faster, um, and to help us understand where we kind of stood in the span of that. But then it took it a little bit further when we realized um garbage in is garbage out, essentially. I mean, AI is only gonna do with what it has available, it can only have so much potential. So we were using, you know, when Transfix was a broker, we were using AI when it was referred to as machine learning. You know, like the industry is just catching up. So it's been really important to understand the true difference use case you can leverage with AI. Another part that I'll say that I'm really proud of is that Transfix stress tested our tech by having every employee, regardless of title, I mean every employee actually cover and operate freight. So the AI layer in freight pricing comes from real life experience in covering freight. And it serves a live strategy that can flex with today's market conditions. So it's it's kind of been the whole journey through Transfix, like really understanding AI, how to leverage it, battle testing it on our own PL when it was not as great as it is today. But it's it's it's been a an experience.

Breaking Silos To Train Better Models

Jennifer Karpus-Romain 5:24

I love that journey because I have worked at a variety of different places that don't take the time to do that. And I love that you said too, like, it doesn't matter your job title, like you went in, you did the system, you because every perspective is valuable. That was something that took me like I used to work for a tech company and I wasn't born as a techie brain. I was like a creative marketing brain. And so I would always feel like stupid going into a room with all these techie people. But I was like, wait, we sell CRM technology, which is going to the marketing people and going to the sales people. So if I don't think something makes sense with how they're describing it, our audience isn't. So my perspective is valuable. And so sometimes it's like just that mindset change. But I've also worked in um like having software components where I'm like, a button does not make sense. It's not where the button should go. If you're the person doing this job, you would know that that's not where the button goes. But no one's asking the person whose job it is to click the button. And I instead of clicking like go ahead to the next screen, it would, you would, it you would click cancel repeatedly. Like I did this all the time. Everyone I worked with did it all the time, but no one bothered to ask us. So, like, I love that you did that, especially now when we talk about AI and how it's learning from what we do and like building on that knowledge of how people work and what they do. That's even more important to get all of those perspectives in the room.

SPEAKER_00 6:48

Yeah, and it was especially impactful because what we had is we kind of had these silos set up, right? You had like account management, you had brokers that were covering it, you had sales who, you know, wanted to do anything. And then you had this kind of other silo for us then, which was like all of these really brilliant data scientists and economists who were trying to build a model. And they were taking these like different pieces of macroeconomics and all these indicators to build a cost model. But until they actually sat on the floor and had to talk to drivers and talk to dispatchers and cover freight and understand all the different parts that go into negotiation, um, they had no idea what they were doing. They went from looking at all of these outside factors to looking at the behaviors that drive your cost and how you buy against the model. And that's when we realized like we really have something special here. Like our accuracy is getting a lot stronger when we're when we're doubling down on behaviors and when we're looking at it. And none of that would have come had we not had everybody in the company sit down and book a load at the day, which is really the backbone of a brokerage.

Jennifer Karpus-Romain 7:50

And you've described that like original AI pricing sometimes is like giving confident but wrong answers, which I think is so true. Like I even think using my chat GPT, I put her in sparring mode all the time because if I don't, she's just like, yeah, Dan, that's a great idea. I'm like, but is it like is it like they're designed to like you know be your hype people, like not so they come out very confident, but that does not mean it's right. And so do you think that the like keeping that siloed, not having everyone involved, is that kind of what leads to that disconnect? And obviously at Transfix, you guys broke that down in that way.

SPEAKER_00 8:27

Yeah, I mean, I think it does. I I I think AI is kind of like a I don't know, it's a bit of like a a middle schooler who's like trying to figure out what they're trying to do and and they're lacking that confidence. And so they're a bit of a people pleaser when it comes to when you're working with it, especially that kind of uh platform on which you shared. But look, AI doesn't know what it doesn't know. It'll give you a number with total conviction. My daughter had a math problem that I could not solve. And I tried to look into it on AI and I was like, even I know that's not right. Um, but it may or may not give you a footnote that says, hey, I've never seen this before, I don't know. In that same same thing, most models are trained on broad market data. So when you ask about a lane or a niche commodity, it interpolates from things that don't really apply. And the model just kind of outputs, most models will give you an output of a rate. And a, you know, a dispatcher or an account manager takes that as gospel. Or when they're looking at, you know, one of these market index rates, you know, they'll be like, they see it as like I always say they see it as Bible. Like the fix isn't necessarily better AI. Um, it's better inputs plus a system that tells you when confidence should be high or low. And so because of that, and because of these learnings, like we've really tried to give you a confidence level in the score. Like we'll say, here's your balanced buy, like here's where you're likely to buy, but we're gonna give you some indications that you're either buying, like your buying is moving the other way, where you're buying more at the high end of the market because something has changed in your behavior, or you're buying better at the low end. And so it's imagine having the kind of confidence when your model doesn't just learn your lanes, it learns your peak season, your biggest carrier, your highest volume freight. It it calls into it calls into place the entire picture. So the the connection is on the same page thinking that keeps you pricing in the right way. And that can hedge you from losing margin, or it can help you kind of capitalize on an area which you're buying really well. So having that data and putting that in there and having that right information is is really the best way to leverage AI when it can be confident but wrong, when you want it to be confident with the right set of data to back it up.

Jennifer Karpus-Romain 10:48

Yeah, I think that's incredibly important. And we talked about at the beginning, like in the past, it was like knowledge stayed in your head, but you need it to be across everybody. And even I think that if you think about even as people grow up and they retire and there's this information that they know, and then they leave and there's no knowledge transfer, you're creating a system that because they've used it, because they participated in it, that knowledge is then kept and then manipulated and moved forward to help other people and help your customers. I think that's really smart because there's so much of that that just gets lost. And so, how do we find a way to enable people more? I love it.

Custom Cost Models Versus Benchmarks

Jennifer Karpus-Romain 11:33

And we are so excited. So you will be one of our keynote finalists at Elevate. So, in case people listening don't know, we decided to have a tech mic drop award for our keynotes this year. We will have four finalists that are taking the stage to tell us the best technology that's going on in the industry right now. Real use cases, no promises of what it can do, has to be proven. And the audience gets to vote live in person on who wins the mic drop award. And Carly is one of our four finalists. If you have not registered yet, you can register for elevate at events.tmsa today.org. We will be in Denver June 7th through 9th, and the keynote is on the morning of the 8th, like the Monday, the day in between when we start. Um, but I'm so excited to have you come and talk more about um what you guys did, how you did it, without giving too much away, because of course we want people to come to elevate to see your full pitch, but how does your approach differ from kind of that traditional market rate or aggregated data model?

SPEAKER_00 12:43

Yes. So, first of all, we're super excited to come. Um I think it's like an aid as somebody who's in sales and in this business and running revenue for a company. I'm insanely competitive. And so I always welcome the opportunity to be evangelical about what we're doing at Transfix, especially on a platform where you're going against others. So I personally am so excited to be in Denver this summer. Um, but what I will tell you about what Transfix is doing and kind of how it differs is that traditional models are going to tell you what the market is doing. Our approach tells you what your data says you should do. And those are very different numbers. They're very different outputs. Aggregated data is a benchmark. It's useful, but it shouldn't be your pricing strategy. What we do at Transfix is we build custom cost models. So it's unique to every single broker and it's grounded in the broker's own operational data. So that can be carrier behavior, lane performance, load history, um, how long it takes for you guys to post a load. There's so many behaviors. We have over a thousand behaviors, and every model weighs those behaviors differently based on the unique broker. So the output here isn't this is what your market source says. It's here's what your margin looks like on this lane based on the behaviors on how you've moved it previously. So it's a big differentiator. It's kind of giving you a scope of where you're buying against the market to make more informed pricing decisions.

Jennifer Karpus-Romain 14:19

I love that. And then during your presentation, will you kind of walk people through visually what that looks like and what they can expect to see?

SPEAKER_00 14:26

Yeah, absolutely. So we'll walk through kind of the life cycle of a load, maybe from a spot or an RFP standpoint. The RFP standpoint's really exciting because you can see it from like a full life cycle. You can get the RFP, you can price it, you can apply your rules to it, you can make those changes. Um, and we're gonna take it further than that. We're gonna show you um were you awarded, what you actually priced, same volume, same price. There can be oftentimes discrepancies between there. And then we're also gonna help you track what you've been tendered, what you've been warded. And based on any bit of seasonality, any bit of you know, change in the market, how do you buy or respond to those changes? So you're gonna have an understanding of where your margin lies throughout that whole life cycle. Um, it's it's really exciting, it's super informative. I honestly can't wait to have a big screen behind me and point out all the beautiful features of it.

Data Integrity And Trust In Rates

Jennifer Karpus-Romain 15:20

I am excited to watch it as well. Um, data integrity really seems to be a central theme in your work, which I think is great. That's that's the goal, right? We we don't want to just have data, we want to have the right data to help the right people at the right time. But why is that so critical to building trust into the AI-driven pricing?

SPEAKER_00 15:41

Trust in AI pricing is downstream of trust in the data. And most people don't audit their data nearly as often as they audit their models or they audit their RFPs or what that looks like. Um I think the edge is in the combination, not in the source, but it has to be right. There's no single source of truth. That's why people have multiple subscriptions and they're looking at, you know, the market, they're looking at their historical data, they're trying to find it. If the training data has gaps, outliers, or sale lanes, the model learns the wrong patterns. And you don't find out about it most of the time until you've already bid wrong. This is critical. And it's why at Transfix, we give you a full reporting suite to help you understand what that looks like. One of the reports that we give, which is like one of my wow factors, I actually just showed this to I think a top five broker, and they were like, this is incredible. But we're giving you this reporting to show you your cost model training report and it provides an overview of model training validation errors to help you address issues quickly and efficiently. So you'll be able to see whether, you know, is the issue due to a carrier invoice didn't match or lanes got miscoded, or your out time is like three days before your in time. You know, there's just things that happen, whether accessorials get buried or fat finger slips, like data integrity issues are common. I actually love it when people tell me they have bad data because I'm like, great, we have a whole team that helps you understand what that looks like and how to improve it. And then your accuracy improves. So the brokers who get the most out of AI pricing are the ones who do the work to clean up their data hygiene consistently. Because that's where you're going to get the advantages in your MPE and your MAP and understanding your accuracy to make better decisions, whether in spot markets or in long-term pricing for RFPs.

Jennifer Karpus-Romain 17:40

I think that makes a lot of sense. And then what type of business outcomes have organizations experienced when they then

Margin Gains And Faster RFPs

Jennifer Karpus-Romain 17:47

do improve pricing accuracy? I'm assuming they're able to like streamline the RFP workflows and their processes move faster, better, more efficient. But what does that actually look like for a company?

SPEAKER_00 17:57

Yeah, so I'd say the outcomes cluster around three things. Um, first is margin that they were leaving on the table. Um, the second is time they were wasting on the manual bid process, spending hours normalizing a file, pulling multiple rate sources, not tracking changes or doing any sort of drift analysis. And then possibly a third, which is um very impactful and a bit harder to quantify, but we've heard it um from our brokers is the potential to know where you don't buy well and taking a more conservative approach on those rate on those lanes. So you're actually not acquiring freight that you typically buy bad on because maybe a rep wanted to get more aggressive to win that freight. And so when you're looking at those three, it can actually have a massive impact on your overall margin and your cost per load. On the pricing side, brokers who move from like a gut plus market to more of like a model-informed approach typically see a meaningful improvement in bid accuracy, which means fewer losses and more margin. Um, on the RFP side, the workflow compression alone is significant. Um, and you can also track the changes of that. So there's no guessing as to like, did why did we quote this this way? Who made the change in the file before we submitted it? Um, you can lock it, you know, you can put a lane or a set of lanes in for review. So what used to take a team days can run in hours when the cost model is doing the lane level math, especially when you have your rates control panel where you can apply rules to the RFP long before you receive it. So that way, again, that that account manager's tribal knowledge doesn't live in their head and they're like, oh, wait, there was a lane they told me to be aggressive on. You can set a rule for it prior to receiving the RFP, and then it's already applied. And then the subtler outcome, which is is pretty impactful from a brokerage standpoint in their relationships with shippers, is to have defensibility with clients. Like when you can show a customer how you priced their freight and why, what support do you have? You know, what kind of lane density, what overlap, what backhauls? That's a real relationship asset.

Change Management And AI Adoption

Jennifer Karpus-Romain 20:28

Yeah. Do you find that people get immediately on board when you present these types of concepts and idea, or they're still skeptics on AI and the power of it? And if there are, how do you get them to understand the power of it and like the safety of their data, what they're sharing, things like that?

SPEAKER_00 20:49

Yeah, I think I think as we move more towards um AI adoption in today's environment, it's less about the efficiencies that can be gained and it's more about like change management. I think it's harder for people to kind of adjust to the new ways of doing an RFP, knowing that they're actually providing more value in the way that they're pricing and the way that they're gathering the data and it's being presented so that they can have those conversations as opposed to seeing it more as a threat. Um, in terms of the AI as a whole and accuracy, we we run uh before we turn the model on, we'll we'll give our broker maybe 20 lanes for them to kind of spot check, run it internally, make sure you feel comfortable with those rates. You can you can have a week or two to do it to kind of see before we turn the model on. So we really work to help people feel more confident in the rates that we're giving them. Um and it's also like we're not saying this is it. We're giving you a distribution and we're showing you here's where you're likely to buy, and here's the behaviors that support.

unknown 22:00

It.

SPEAKER_00 22:00

I'm not running your business. I'm just giving you the data to help support your strategy and what that looks like moving forward. Great.

Jennifer Karpus-Romain 22:09

I was just curious because I feel like we're getting further into it where people are like, okay, AI does have power. It's not necessarily replacing, but changing. I loved the change management piece of it because it's also like you might have to like invest in more time of updating your processes because that's going to change and let people know what you're doing and how. But um I still see some hesitancy, but it's getting better. So I was curious, like, because you live and breathe it every day, if that's still what you're seeing, or if we're moving a little bit faster and more understanding now.

SPEAKER_00 22:43

I think there's like overall hesitancy, but like when you have a champion internally, it makes a massive difference in adoption and what they're doing. I had just made a LinkedIn post yesterday. We had an enterprise client that did over 65 RFPs in last week alone. Uh, and there was like not a peep from them, right? Like we have we have weekly check-ins, they have their own customer success agent that can help them throughout the process. And we were like, hey, you guys ran a lot, wanted to check in. And they're like, Yeah, like once you move through the process, it's it's really easy, it's really intuitive, and it's there. So I think once they realize the ease in which it provides you to move through the product, right? Like on our in my circumstance, it's it's RFPs, it helps with that adoption and kind of like helps people take their walls down a little bit for how they're using it. Great.

AI As Collaborator In Revenue Strategy

Jennifer Karpus-Romain 23:35

And then looking ahead, how do you see AI continuing to kind of shape the pricing and revenue strategies in the industry?

SPEAKER_00 23:42

Yeah, I think people were using AI originally as a pricing calculator. Now I think we're shifting for using AI as a collaborator, and that's a really big difference. Um, right now, most brokers use AI to get a number, and it's a good start adoption-wise, right? Like you want them to try to find ways to move it into their workflows, but you have to graduate from that. And the next wave is that AI explains the number, it flags when it's uncertain, it helps you understand the why, going back to that like defensibility into a with a customer, you know. And then it's gonna it's gonna constantly learn as you override it or as your behaviors shift internally. On the revenue strategy side, I think we'll see more real-time margin management, not just bid pricing, but in cycle visibility into how a book of business is performing against the model. I think the winners, the winners will be the brokers who treat AI as an input to a human decision, not a replacement for it. Yep.

Jennifer Karpus-Romain 24:51

Yes. I still think humans are important. It should be amplifying what we're critical about replacing it.

unknown 24:58

Yes. Absolutely.

SPEAKER_00 24:59

Like we we want to give it to, we want to give you all of this data, all of this as an input, right? There is still every every brokerage is running their their operations differently, they're incentivizing behaviors differently. Um, so the strategy for each of them is very different, and that requires a human and it requires collaboration. And so, again, it should be an input to help make the most informed decision that's best for your business. Awesome.

Jennifer Karpus-Romain 25:27

Well, thank you so much for coming on the show. If you want to hear more from Carly, join us in Denver June 7th through 9th, as she takes the stage as one of our opening keynote technology finalists. You can get more information on that and register at events.tmsa today.org. And I do have one last question for you, Carly.

Career Advice And Closing Invite

Jennifer Karpus-Romain 25:53

And that is if you can go back in time and advise a younger self, and this can be anything personally or professionally, when would you go back to and what would you say?

SPEAKER_00 26:05

Um, I would go back to my first year in brokerage when I came from the ocean side, and I thought I was I was wildly confident and what was happening. I'd go back to that first year in brokerage and I'd tell myself the confusion that you feel and not understanding the way brokerage operates as opposed, like a non-asset model as opposed to an asset model, is actually your own internal asset and not to rush past it. When I think when you're new and you're moving into a different part of the industry, or you're maybe going from brokerage to tech, you start to ask questions that the smartest people in the room have stopped asking. And those questions are often where real problems live. I spent a lot of time early in my career trying to sound like I knew more than I did. I felt like, you know, everyone, I needed a seat at the table. I needed to be informed. When the most valuable thing that I had then, which I didn't realize, was fresh eyes to the problems that were in front of us. Um I think the the operators that I've respected and looked up to the most for both tech and in the brokerage space are the ones who stayed curious longer than they felt comfortable. Um and I'll just call a spade a spade. This is a very confusing time. It's just a fact. Geopolitical factors are being thrown into the mix, left and right. Um, and so I don't think there's any better time to get curious and ask why and challenge the norm and be open to shifting your strategy. Um, I yesterday was on a call internally and I was like, okay, this may be a dumb question, but I need you to really break this down for me. I don't understand why we do this. Um, and from, you know, if you can't ask any of those questions, you're you're never going to be informed enough to be able to, you know, sell your product or work through it. So I get that it's like wanting to feel like you belong in the room and that you've earned your seat, but I I think you should always stay curious. And it's something that I try to remind myself today.

Jennifer Karpus-Romain 28:19

I think that's great advice. I have always said here at TMSA, if I ever say, oh, well, this is the way we've always done it. So this is how we're gonna do it, just fire me on the spot because that's not that's not what's gonna get you there. And that's I mean, I live and breathe that too. Like we we changed our entire model to be track-based. So, because I was like, we need to give more individual support for people at different stages of their career, saying sales and marketing isn't enough. So, like we have to change, we have to evolve with what people need. And it's scary, it's terrifying because you don't always know if you're doing it the right way. But um, and then I have a quote that sits over there in my office and it says, Whatever got you to where you are won't keep you there. And that is like the sentiment of what you're talking about. Like you might be rising in your career and you might be in that top seat, you might have a seat at the table, but then if you sit and you rest on your laurels and you don't stay curious and you don't stay interested, you'll lose that seat. So everyone's got to keep earning it, I guess, would be, you know. So I think that's great.

SPEAKER_00 29:24

Absolutely. Yeah. My quote, um, which sits on my desk over here, is it's not about what you want, it's about what you're willing to do to get it. And that might mean asking the dumb questions or finding a way um around it so that you better understand it.

Jennifer Karpus-Romain 29:40

Yeah. And I tend to learn that if you think it's a dumb question or something that you don't understand, you are not the only one.

SPEAKER_00 29:49

It is, I mean, I've asked some dumb questions where people look at me and I'm like, okay, now that I said it how loud I get it.

Jennifer Karpus-Romain 29:54

Okay, so but like for the most part, other people are questioning or thinking it too, but they don't want to say it. So well, thank you so much for coming on the show and for coming to Elevate. So excited to see you take the stage, and hopefully, we will see everybody in June as well. So, and if not, I will catch everybody on the move next week. See ya,