Episode Transcript
[00:00:00] Speaker A: Hello and welcome to the Lodestar podcast. I'm your host, Charlotte Goldstone. This episode is in collaboration with Raft AI.
[00:00:10] Speaker B: As this complexity continues to increase, it allows our customers to cope and cater to that complexity with something that's really scalable.
[00:00:18] Speaker A: Raft AI empowers freight forwarders and other stakeholders to automate workflows across the entirety of the shipment lifecycle. I'm going to be joined by its co founder and CEO James Coombs to chat through AI. AI's role in an enterprise setting and specific use cases, benefits, the importance of accuracy and also the role of humans in AI empowered processes. And excitingly, we are also going to be joined by one of raft's customers, customs technology consultant at international customs provider als, Michael Douglas. And obviously as an actual user of this software, Michael is able to really illuminate how ALS identified the need for AI tools, specific use cases within customs processes, what benefits they are able to pass to their customers. And also some advice for those of you who want to take on similar technology in your processes. So you'll have to stick around to the end if you want some good advice from Michael. And it is no surprise that obviously James and Michael are able to talk about AI much more fluently than I can. So without further ado, let's get on with the episode.
Hello James and Michael, thank you both so much for joining me today. How are you?
[00:01:37] Speaker B: Hey Charlotte, great to see you. Very much a pleasure.
[00:01:39] Speaker C: Hi Charlotte, looking forward to it, as am I.
[00:01:41] Speaker A: And I will come back to you in a second, Michael. But starting with James, can you please briefly introduce yourself, your company and your role?
[00:01:49] Speaker B: Sure. My name is James, James Coombs. I am the CEO of a company called Raft AI. We automate shipping executions in the freight marketplace.
We look across shipment lifecycle from bookings to customs applications, right through to AP audit. And to give you an idea of our scale, today we type about 5 million shipments a year on our platform. We have 10,000 users and we serve roughly half of the top 25 trade followers globally. We have a global presence, we have teams both in Asia, US and Europe. Really good to be here. I think it'll be a great conversation. Really good to see Michael here as well.
[00:02:27] Speaker A: Wonderful. And Michael, would you be able to give a brief introduction of yourself, your company and your role, please?
[00:02:32] Speaker C: Yeah, absolutely. AIs Custom Services are an international custom service provider. So consultancy customs clearances at ports and airports, over 20 countries in the world.
A lot of it is 24, seven operations. So it's real people in real offices.
Providing, you know, custom services to logistics companies, importers, exporters, traders.
I work as a senior consultant on the customs technology team. I mean, I've been in this business for over 45 years. We have our own technology team where we are developing our own in house innovative software for customs management.
We're doing great things. But some specialists functions such as artificial intelligence, intelligence document processing, such as that provided by raft, we tend to find good partners because we're not going to reinvent that particular technology ourselves. So we integrate that type of thing into our own technologies. And I work at the concept level on that to make sure that we're developing what the customers need.
[00:03:42] Speaker A: Well, I am looking forward to hearing lots more about that later. But before we get into the specifics of how this kind of special specialized technology can be used, I want to really clarify what we actually mean when we use the umbrella term AI. I mean, it is no secret that AI is a bit of a buzzword in logistics right now. I see it all the time in press releases that land in my inbox. You hear it all the time at trade shows. I understand AI to be your kind of typical day to day ChatGPT or the Siri or Alexa that you use on your phone.
But we are obviously here to talk about AI in an enterprise setting. So James, what is actually the difference here?
[00:04:22] Speaker B: Yeah, so I mean, AI is one of those really interesting phenomenons, I guess that has really started slightly differently from many of the other ones that have essentially defined technology and technology landscape over the last 20, 25 years and where Internet and the Internet started more of a kind of enterprise and large, large institution level and then kind of was brought down to the consumer level, we're almost in this interesting position in AI where really consumers have led the charge. And so you have large kind of foundation models and large businesses like OpenAI who are out there and who have really released this quite fascinating technology directly to consumers. And it's really interesting to see. And actually consumers have in many cases kind of more power and a better experience at the consumer level than they might have in their day to day work. And that's a really interesting phenomenon because actually it means that a lot of our customers are actually really well informed. And that's an interesting position because a lot of our history, we've been going for now almost seven years, we've had to do a lot of education, we've had to explain what the benefits are, what the merits of AI are to our customer base. But in a world where consumers are increasingly well educated understand what is possible and to a lesser extent, what's not possible. It means that we're not having to educate as much as we used to.
Now then, the difference between the consumer and enterprise applications is there's clearly a need for an amount of reliability when it comes to data in an enterprise setting. And there means that the bar is slightly, it's not even higher, it's just slightly different. Whereas a user you might be familiar with asking a question to a ChatGPT or whatever and getting an answer that you think is relatively accurate, you're also a little bit more tolerant of mistakes because it's not quite as mission critical as it might be in the enterprise setting. In enterprise, especially in our industry, in freight execution, freight forwarding, et cetera, there's a real high cost to error. And so it really is important to be able to kind of take all the benefits that you can get from these models, but then also have a consistency that allows you to run your business with the kind of reliability and lack of error that is really critical in the enterprise setting. And I think that's the biggest distinction between the kind of consumer experience and enterprise experience.
And that's really where we invest a lot of our time and efforts in terms of our platform, in terms of our user experience. A lot of what we do is informed by all of the same technology that met familiarly with a consumer level. But everything has an override, everything has an ability for a user to validate what's going on in order to drive that kind of data reliability that we all know is this incredibly important. And I'm sure Michael will help to provide some of that context on his site too.
[00:06:59] Speaker A: Yeah, I mean, I think we've probably all got something back from ChatGPT at one point or another that has just been utter nonsense. So it is really important to make that distinction and as you say, have AI technology that's actually going to give you that reliability in your data. I'm curious, is this kind of push to extract reliable data, as you mentioned, one of the main use cases for AI? I do often hear it kind of discussed as a tool to get relevant points extracted from overloaded inboxes or to automate bookings. But James, as someone who has these kind of perplexed stakeholders come to you every day looking for solutions, what are you seeing as popular uses for this technology?
[00:07:38] Speaker B: I mean, for us, I mean, we only have to look at some of our kind of core product offerings. Customs is clearly one that comes to mind. It's something that all of our customers struggle with in some shape or form. It's a really interesting space where in my mind, you have a kind of, you have to evaluate what you have and what you've been provided in terms of documentation. But that data can be in any form, whether it's documents, emails, APIs, EDIs, tribal knowledge, all of the above, really.
So it's part of partly understanding what you have, but then also partly understanding what you need. So if you are importing electronics from Korea, very different to importing bananas from Panama, and what you have to do in both of those scenarios is quite different. And so you kind of have to really grasp and kind of get your head around all of those pieces.
And that's also where AI is great. It helps define what you have, but it can also help to increasingly build an understanding of what you need.
And again, millions and millions and millions of customs decorations out there, in many cases being done manually, there has to be some kind of oversight. Clearly, it's an incredibly regulated environment that does need human oversight. But a lot of the legwork that leads up to it, what we might call customs prep, is definitely something that AI is very well suited to and something that I think is going to get a lot more attention in the, I guess, the months and the years to come. And a large part of our business, another product that we have that is particularly relevant is on the accounts payable side. A lot of work and effort goes into understanding that. And what our customers are being asked to pay accurately reflects the rates that they've been quoted, all the other kind of details that are part of that AP process.
And there we're finding tremendous use cases for AI as well. I could go on. Like you mentioned, there's bookings, there's all the other type of orchestration elements, but there's some of the kind of two real tangible use cases that we deliver.
But again, I think it's a growth area across the board across that kind of shipment, that whole shipment lifecycle.
[00:09:32] Speaker A: Well, talking about use cases seems like the perfect time to bring in Michael to the discussion. So, as I previously mentioned, ALS is an international customs provider and is also one of Raft's customers. Now, customs is such a hot topic right now. So, Michael, I'm really interested in hearing, where have you found it useful in ALS's processes to kind of use this AI technology?
[00:09:58] Speaker C: Well, obviously, I mean, the core functionality is intelligent document processing. So Rafter using AI to extract data from document, which is not as simple as some people believe. Many people believe that a PDF is an electronic structured document. And actually it's not, it's just an electronic readable document.
And so extracting that data has just come on enormously with the advances in AI recently. So the core element is we use RAFT to extract data from documents. But because we're custom focused, it's absolutely crucial that the data we pass through to systems is 100% accurate. And of course, IDP is sometimes 100% accurate, sometimes it's not, and it's less consistent. So one of the really important functions is what's called a human in the loop review. So IDP will extract data and then it will show you all that data and it'll show you the documents and you have to confirm it's correct. Now, if it's missed some things, the human in the loop is so important for Customs because we can sometimes have dozens or hundreds of pages with granular level detailed commodity codes, countries of origin. You can't have someone just looking at a normal IDP technology and sitting there trying to work out what's missing. So what RAF do is focus on human in the loop review functionality that is focused on customs. We can look at totals and summaries, we can look at missing granular level detail.
And that's important for us more so than accounts payable or something else where there's a lot less detail in there. And the other thing is integration to our system.
We develop our own customers management system. So what do we do to get that data out of RAFT into our system? You know, we're using seamless APIs, we're getting instant transfer of data, and for whatever reason that data has been blocked. And RAFT work with us really closely to make sure that that is absolutely seamless. And that means that we've now got data that we can produce decorations from. And it's been important to have RAFT because Customs is so focused on compliant that that human in the loop review is much more important because if something is wrong, then fines and penalties and delays can happen.
So that's why we were very determined to find a focused provider rather than just someone who says we extract data. And here it is.
[00:12:21] Speaker A: And so you've got this ability to extract accurate data and then also integrate it into your systems. But what does this actually mean for tangible benefits? Like how do you pass this onto your customers? I'm always hearing time saving as a big benefit of AI, but what benefits do you get from this in your processes?
[00:12:41] Speaker C: Absolutely, absolutely staggering benefits, I have to say. Obviously the first thing people think is it's Going to save time on typing, you know, and that was our number one idea when we first started. Where is that on the list now? It's probably number five or six or even more. The most important thing for us is to get data at the start of the process, not just somewhere down the line to create a declaration. When we get that data, it means we can provide end to end services. So we can do an export declaration in the uk, an import declaration in the Netherlands or Germany, all with one data set. I mean, that makes it cheaper and really efficient for the customer. We get better compliance. As we've said, there's less errors. If you're looking through 95 commodity codes, AI IDP is going to make less mistakes than a human. Even our very best, it absolutely is. Returning data upstream and downstream. I mean, this is something we'd never really understood, is that lots of people in the supply chain are putting that data into the systems. Why is multiple people doing it? Why doesn't just one, one party extract the data and give it to everybody else? Now, as a customs broker, we need more data than anyone else because we are doing the decoration. So one of the things we quickly did with a very large logistics company is they put the data into their system to book fairies. But we agreed, look, we're extracting data to do a customs decoration. We can summarize your packages, your pallets and your weights and give it back to your system automatically to automatically boot the ferry. So all of a sudden we are doing what we do very efficiently to create declarations, but we can pass that data back upstream and downstream. Probably more importantly for the future is digitalization of data and logistics. And customs is becoming a critical focus because of things like EFTI, which is the EU's electronic freight transportation and the UK's ETDA electronic trade documents Act. They're only just on the horizon now.
The authorities are starting to say, you've got data, we can now use that data.
And so because we can get that data early in the process, we've actually started running a pilot with HMRC to prove that that data in advance of arrival and declaration can revolutionize border crossings. We can stop the delays cost by Brexit in the UK and other places. I mean, it's been really important. All of this comes from getting data at the start of the process. And I mean, watch this space. It'll be something huge and then they save time on typing. So that shows you that you know the kind of where that suddenly came it important. But it's now down the list.
[00:15:12] Speaker A: Yeah. That point about data is so important. I'm talking to freight forwarders and shippers every day to get their kind of take on the current supply chain environment as part of my job. And one of the things I hear all the time is that increasing requirements for data, especially with cross border shipments, is becoming such a pain. And obviously we've seen the US impose stricter requirements on shipment information, what with all the tariff codes and the removal of de minimis. We've had Brexit here in the uk as you mentioned, Michael, there are stricter reporting requirements for environmental reasons. I mean, just so much that comes back to the need for data and to have it in this standardized format so it can be shared across parties. Obviously this is such a fragmented industry.
James, is this something that you've noticed an uptick in like more and more people coming to Raft to help with this kind of data side of things?
[00:16:00] Speaker B: Yeah, I mean there's a ton of great insights there from, from Michael and I think most of our customers would align with that. Is that that concept of this kind of data and the importance of this kind of clean, standardized data, but it's also like fungible with other people in the ecosystem, but equally for any given shipment. Right. So the data that you need and the customer separation can equally inform other parts. Right. It can be part of order management, it can be part of a variety of other processes that adds a huge amount of value to the end customer, I. E. The end customer of our customers. And what's important about that is that increasingly a lot of our customers are coming to us because their customers are coming to them and saying, look, what are you doing in AI? What can you give us in terms of data? There's this ever increasing demand for insights on data and the insights that allow the end customer to make better decisions. And that starts with getting the data right, getting it up and running quickly, getting it standardized, getting it clean. And I think that's increasingly where we see this industry go. Now on your question, in terms of the kind of global outlook and the increasing complexity both in the US and elsewhere, ultimately that just exponentially increases the amount of complexity that is required at the customer level, but also just generally at the operational level. A lot of our customers are having to diversify a lot of their trade lanes and their trade routes as well. And so what that means is that maybe when an end customer might have had one or two core relationships that were driving a lot of their business, but all of a sudden those couple of relationships are now 10 plus. And so when you have all those different relationships, then you have a far greater variability in this kind of data, the type of partners and other stakeholders in the business that you have to work with, which means just if anything, it's just more and more complexity.
And so again, it's a huge benefit where we talk to our customers and say, look, the goal is to scale with tech, not people. And as this complexity continues to increase, it allows our customers to cope and cater to that complexity with something that's really scalable. It's really easy to scale up bits and bytes of technology, far easier to scale it up than it is for people. It just helps build more resilience, it helps our customers deal with all this complexity better. And ultimately it's been a massive positive for our business.
[00:18:15] Speaker A: Well, especially when we talk about data at the customs level, but also, I mean, just data across the whole supply chain. It is so important that it is accurate. And I think this is one of the main concerns with AI usage. I mean, probably stemming back to the inaccuracies we mentioned with consumer based AI products like ChatGPT at the beginning. So James, how can RAFT actually ensure accuracy?
[00:18:39] Speaker B: Yeah, accuracy incredibly important. I mean, the human loop piece that Michael mentioned is a big part of that. There's so much more as well around all of the context about our work exists.
So like Michael said, I mean, nothing really exists in isolation. So what you have to do in customs is equally relevant to order management or to warehousing or whatever the case might be. And when you have this kind of data that in many ways overlaps, there's a lot of things you can do there as well. In terms of things like validations, you can look at a lot of historical stuff as well. So there's a lot of approaches that you can take to make that accuracy better, make validation better, and ultimately provide that confidence. A lot of our customers need. There is still today a need and a requirement for a user to be in and around this. But I think that there's a future where that happens less and less and really a future where you have humans essentially overseeing and overseeing the exceptions as opposed to actually doing the kind of the day to day. And it's not a huge stretch to imagine teams of bots, if you will, working here with a kind of broader human oversight. But again, I think you have to carefully manage the kind of the enterprise requirements against that. It's almost like it should be a co pilot rather than autopilot.
[00:19:58] Speaker A: Yeah, this balance of automation and human oversight is another topic I want to explore a bit further because as well as accuracy, when we talk about AI, another very real fear for users is that AI is going to take their jobs. And I mean, that is something that I'm quite scared about as a, as a writer and a podcaster.
But I mean, also this is one of the main big barriers to adoption is if a company's been doing something one way for years, it can be a very real challenge to introduce a change in processes. You obviously need to get users on board with it if it's going to work as part of your team. So Michael, maybe this is something that you can speak on a bit more as a user of this technology.
I mean, how did you get your team on board with the idea of using this and then also kind of make sure that they understand the technology?
[00:20:44] Speaker C: That's really interesting because we've gone through a path on that.
I mean, ultimately what we want to do for us is provide our staff, the decorates, the knowledgeable people. We want to provide them with more reliable data rather than them having to type it because that means there's less stressful outcomes.
We want them to utilize their skills, which mostly are finding the correct customs, processes and following them, looking at port regulations, dealing with that minutiae, rather than worrying about typing data from this document into a system.
So what we've started to do is people who do like using that technology and transferring data, they get the option to become the data processors and they do have customers knowledge and the ones that don't, then data is just arriving almost out of a black box. It's not a black box. One day, as James said, it will be a bot, fully functional and 100% accurate. But at the moment that's not necessary. And even when that does happen, it will give more jobs. I mean, we want our staff to spend time on important things such as customer liaison, process efficiency, dealing with the exceptions, you know, when things go wrong.
So I think what we've managed to do by doing this type of thing is help them to understand that AI won't mean fewer jobs. I mean, that was a fear in the Industrial Revolution. Machines would mean less jobs. It actually created more jobs and better jobs. When computers exploded, not literally the same fear surfaced that there'd be less jobs, but there were more jobs, higher wages, better conditions, and it'll be the same with AI.
It's not about robots taking over the world. I mean, let's be honest, they're not even very good at tying shoelaces.
It's about AI helping everyone to do their jobs better and with more satisfaction and improve their lives. And, I mean, I think it's better described as assisted intelligence rather than artificial intelligence, because in the mindset of people, they suddenly realize, actually, I'm getting something to help me do my job better rather than something that's replacing me. Of course there will be different jobs arrive, and that's always happened. There'll be different jobs, but it will work. And we have started to get our staff to understand that rather than just dumping a technology on them and saying, you have now got to use this. So for us, it's been an interesting journey and I think it's going to prove really fruitful.
[00:23:06] Speaker A: I mean, as a user of this technology and someone who was once at the start of this journey, what would your advice be to someone who might want to get involved?
[00:23:16] Speaker C: Well, the first thing, you've got to find the right AI provider. I mean, AI is just a global term. You find the right AI provider. We wanted intelligent document processing to extract data from documents.
Pick the right one. We picked Raft because we wanted someone focused on our domain sector, logistics and customs who could meet our special requirements.
So find the right fit and the right provider. Don't just go and find anyone and raf with the right fit for our needs.
And if you're an importer, exporter or logistics company, then partner with a customs broker like als. Because we are using AI to the maximum benefit, you know, we can make people's businesses more efficient, really, that that bar of their expectations would be absolutely blown apart when they realize what you can do once you have accurate data at the start of your processes.
[00:24:10] Speaker A: Some wonderful advice there, Michael. Thank you so much. And James as well, thank you so much for your insights earlier on in the episode. It was wonderful to speak to you both.
[00:24:19] Speaker C: Brilliant. Thanks very much, Charlotte. Thanks all.
[00:24:26] Speaker A: So there we have it. Thank you so much for listening in to this episode in collaboration with Raft.
I hope you've all learned something useful about AI. And if you want to speak to James or Michael about anything you heard in this episode, their names and contact details will be down in the description below. So do get in touch if you're interested in hearing more. Thank you so much. And I'll see you next time on the Lodestar podcast.
