The SAF Podcast
Welcome to The SAF Podcast, the only podcast on the internet that exclusively covers sustainable aviation fuel (SAF). So if you want to find out the real issues and challenges are for commercialising and scaling SAF production, look no further.
Every week we will be hearing from senior industry leaders who are actively shaping the future of SAF and aviation.
Hosted by Oscar Henderson and brought to you by the team at SAF Investor. Connect with us at www.safinvestor.com
The SAF Podcast
The SAF Podcast with Dani Charles, Veriflux: AI as the co-pilot for compliance
In this episode of The SAF Podcast, we welcome back our first returning guest, Dani Charles, CEO and Co-Founder of Veriflux. Following his deep dive into U.S. policy in his first appearance, Dani returns to unpack the increasingly critical—and complex—world of feedstock traceability in sustainable aviation fuel (SAF) production and how AI can could be the secret to doing it accurately at scale.
Dani explains why relying on spreadsheets are no longer viable at scale and how best-of-breed technology is essential for managing dynamic trade flows, compliance uncertainty, and complex conversions across liquid, solid, and gaseous feedstocks. Does anyone actually know how many bushels make up a bale and how many bales to a ton?
AI offers practical solutions that address real industry pain points. Rather than futuristic robots making decisions, think of it as a "co-pilot for compliance" that can extract information from bills of lading in seconds rather than hours, reconcile data across systems, and ultimately compress days of manual work into automated processes that happen seamlessly in the background. For an industry drowning in paperwork and cross-border regulations, this technological evolution couldn't come at a better time.
We also explore recent developments in U.S. policy, where proposed extensions to the 45Z tax credit could provide regulatory certainty through 2031 while potentially reshaping feedstock preferences toward crop-based materials.
SAF Investor London 2025 is running on the 28th-29th May and is filled with key industry players discussing the critical challenges around scaling SAF investment globally. If this is of interest find out more here: https://www.safinvestor.com/event/145508/saf-investor-london-2025/
You can also listen to our previous episode with Kwame Bekoe, AfriSAF where we discuss growing investment in SAF production and feedstock procurement across Africa here: https://www.buzzsprout.com/2202964/episodes/17070781-the-saf-podcast-afrisaf-tapping-into-saf-s-continental-sleeping-giant.mp3?download=true
Hello and welcome to another episode of the SAF podcast. This week I believe we have our first return guest on the podcast in Donnie Charles from Variflux. In the first episode we spend a long time diving into all the fun details of US policy structures and how policy is implemented. This week we're going to do something slightly different and actually talk about another aspect that Donnie spends a lot of time doing in feedstock and SAF traceability and specifically the role AI can have or not have in streamlining that. But before we get into all that, donnie, how are you?
Speaker 2:I'm doing well. Thanks for having me back. I'm honored to be the first repeat guest. I'm going to add that to my LinkedIn as an honorarium, and it's probably the first time that I've spoken about the sausage making that is US government policy and been asked to come back and talk more, so it's a first for me too. So thanks for having me.
Speaker 1:Well, a lot of firsts and a second. So, as I mentioned, we're going to talk about traceability today and get on to what AI can do within that sphere. But before that, before we get into that, do you just want to map out, you know, what traceability for feedstocks and SAF looks like now and sort of remind everyone of the requirements, who the players are, what the sort of deal with traceability as it stands in right now?
Speaker 2:Sure, yeah. So to take a step back and I mentioned this in the first episode that we did together I wrote a piece at the end of 23, beginning of 24, where I said 2024 was going to be the year that traceability goes global, and it certainly proved to be the case. So let's talk about how it went global and what some of those programs are Starting. In the US, we have the EPA's Renewable Fuel Standard, rfs, which has record-keeping requirements to be able to demonstrate where the feedstock originated. That ultimately led to the production of fuel in the subsequent generation of a REN or renewable identification number.
Speaker 2:We also have, at a federal level, 45 Z or 45 Zed, as, as I know you all like to refer to it.
Speaker 1:I keep saying it properly. Thank you for your time.
Speaker 2:Yeah, I wanted to make sure that I'm addressing all possible audiences the right way. So in the context of 45Z, you know we can talk a little more about that later, but just to set the ground, in January of this year it went into effect replacing what was previously a Blender tax credit 40A 40B. The reason why there's these numbers 40A, 40b, 45z is it corresponds to the part in tax code where this is found and these are fundamentally tax credits.
Speaker 2:They're credits that are provided to previously blenders and now, on a go-forward basis, producers domestic producers within the United States of SAF and of renewable diesel, hvo and things like that. So that's at a federal level. Then, if you go on a state level, we have a number of different state programs. We have California's low carbon fuel standard. We have similar LCFS programs in Oregon and Washington state, a new one in New Mexico, oregon and Washington State, a new one in New Mexico. We've got SAF credits that exist across different states Washington State, illinois, a couple others, and then, if we look to our neighbor to the north, they have their federal program, cfr, the Clean Fuel Regulation, and then they also have a similar LCFS within British Columbia, and each of these programs have slightly different record keeping requirements.
Speaker 2:California focuses on full chain of custody and verification of the chain of custody of feedstock back to the point of origin. Canada is focused on declarations and the declaration of material to prove that it's in alignment with the program, and they've got formulas that you have to be able to account by. In particular, they have what they call their material balancing formula, which is queue incoming plus queue inventory is kind of how you manage, how you can balance material through different inventory locations. So there's all these subtleties that exist across these programs. California is in the process of potentially updating to 4.0.
Speaker 2:There was a challenge around. They passed that, but then there was a challenge around it, so it's still in further review and adjudication. And then, if we shift to Europe, europe's in the transition from REDD2 to REDD3. As part of that, they are rolling out the Union Database for Biofuels, or UDB. That's coming at a time when they're trying to increase their ability to kind of track product through the supply chain. There's been notable fraud investigations in Europe that have been publicized over the last couple months and so they're making it even more intensive from a traceability perspective.
Speaker 2:So when I said, 2024 is the year traceability was going to go global. Maybe I amended it and say 24 slash 25 is the year traceability went global. But I did write a piece at the beginning of this year and that piece the beginning of this year, and that piece in particular what I said is that the only certainty in biofuels is uncertainty and that was kind of my prediction for 25. And I wasn't necessarily forecasting the tariff situation though that certainly plays into it but I was forecasting some of these movements around the union database, lcfs 4.0, policy uncertainty around 45Z, and so all those things have really come into effect in 2025, which has made traceability all the more complicated because requirements are dynamic, trade flows are dynamic, so that's kind of just the state of affairs for traceability writ large.
Speaker 1:So basically, traceability in a nutshell is a load of acronyms, lots of numbers and letters put together and lots of fuel with a book of paperwork to accompany it. In a nutshell, yeah.
Speaker 2:Yeah, it's. You know, for companies that aren't using best of breed technology subtle reference to Veriflux there For companies that aren't using best of breed technology, it's a lot of spreadsheets, it's a lot of manual work, it's a lot of unknowns, both in terms of what policy is going to be, but also in terms of what you're doing to product might impact its eligibility across different programs. It's a lot of headaches. To kind of sum it up, a lot of people are taking ibuprofen and aspirin.
Speaker 1:You said you made your initial prediction in 2023, and you've got your new prediction for this year, which so far seems to have come true, hopefully, I mean, I hope for the industry's sake.
Speaker 2:you're wrong, but maybe get a bit more clarity. Yeah, I like to joke. I do think we'll get more policy clarity, and we're starting to see that in the US. So I do think we will get more policy clarity. What I don't think we'll get is alignment.
Speaker 1:I would love alignment.
Speaker 2:It would help further streamline things and, as someone who builds technology where you have to be able to write in rules and be able to have binary outcomes for math and things like that, it's hard to have ambiguous definitions or ambiguous guidances.
Speaker 2:We find ways to navigate that within the system by building out things like what we call bio bundles or things like that, where you can dynamically create those inputs, but it makes it difficult technologically. So I fully support alignment. As I say that we don't build with alignment in mind, because I don't think it's going to happen. I don't think you're ever going to end up in a situation where all of federal you know the federal record keeping and other requirements and California's response and they meant it sincerely was we're not the federal government, we have our own objectives and our own goals. So you already have that. You know, for lack of a better term tension that exists within the United States alone. And then, if you look at it globally, I mean the best example of of not aligning standards is we use pounds and I don't mean the currency. So that, just that's just a fine point on on the fact that I don't think we're ever going to reach alignment across all these different programs.
Speaker 1:Is that fundamentally quite challenging. Although it seems trivial the the metric and imperial system differences. Although it seems trivial, the metric and imperial system differences, Is that actually one of those things that you don't necessarily think about? This actually can be quite a headache.
Speaker 2:Well, yeah, I mean think about it this way we have to manage metric and imperial, we different states with different conversion factors, both converting from one state to another, but also just converting the product in terms of yield loss and things like that. So, yeah, it's incredibly complicated. And add to that that some folks use especially when you're talking about the upstream side of traceability some folks use metrics that don't tie to either standards. So someone might say I collected five bins. Okay, but what's a bin? Right? In the US we use 55 gallon drums predominantly.
Speaker 1:Isn't it a trash can in the US as well?
Speaker 2:Yeah, yeah. So my point being or bunches, right, when you're talking about agriculture, when you're talking about different types of fruits, right, they come in bunches. We have bushels for agricultural stuff here in the US.
Speaker 1:So, to those that aren't aware, what's the difference between a bush and a bushel, a bunch and a bushel a bunch in a bushel.
Speaker 2:So I feel like I feel like this is the setup for, uh, for some sort of, for some sort of joke we can make, but um, but no, but like a bunch. So when you talk about palm right and, and there's big movements around sustainable palm and things like that, the first conversion you have to do from the plantation to the mill is bunches to, ultimately, to palm oil and all the derivatives. And again, a bunch is just. There's not a particular weight or volume to a bunch, same for a bushel, right. So a bushel of wheat, a bushel of corn, a bushel of soy, whatever it may be. So my point is like you have to layer on all these different intricacies. It's why, fundamentally, paper spreadsheets, et cetera, don't work at scale for this. At some point you reach a breaking point, and we're seeing that across a lot of companies as they're saying, okay, I can't manage this in the old way. And that's where technology really comes into play.
Speaker 1:Is this the sort of the fundamental challenge, the sort of the fragmented nature of it all? Does it actually become simpler if everyone's using the same metrics? Or are there sort of other challenges on top of that that you're constantly sort of fighting and sort of contesting against? Or is that sort of the fundamental issue just that fragmented?
Speaker 2:against, or is that sort of the fundamental issue just that fragmented? No, I don't. Yeah, I don't think that's the fundamental issue. I think it's just an exacerbating factor, like if, if those issues didn't exist, um, my job would would be well, I should say my development team's job would probably be easier because we could, um, we could update the product with fewer dependencies, more frequently and, by extension, customers would be able to have more streamlined operations. But, yeah, I mean it's not going to happen.
Speaker 2:I think the best example there is tariffs. Right, we don't know what's going to happen with tariffs. Is 10% going to stay a benchmark? Is there going to be an escalation again, or are we done with those escalations? Are we going to see reciprocal? And what I think is constant in all of that is the uncertainty, which is where I started, but it's also why you can't write in rules. If we treated assumptions based off of like, if we assumed that South American feedstock is going to flow predominantly to the United States and therefore we want to make sure that South American feedstock that's coming into our system is aligned with North American standards exclusively, we'd be missing the fact that some of that feedstock already flows to Europe and, with tariffs being implemented, more and more of that feedstock is going to flow to Europe, and so we need to be able to build in that optionality, or as we would call it, bio-optionality, into the system.
Speaker 1:So yeah, even if we solved one problem, there's 15 other problems that still exist, so I I think they're more exacerbating factors than they are solutions um, in terms of sort of dealing with domestic versus sort of international products, you mentioned the just the challenge that tariffs represent and that that uncertainty is it. Is it much easier dealing with sort of feedstock within the US and staying in the US, partly because you know there's the metric, or is there sort of similar amounts of challenges that are just slightly different depending on dealing with sort of international versus domestic?
Speaker 2:You could argue it's easier.
Speaker 2:Here I'm talking just about the technological side. I'm not talking about the policy side, you could's it's easier here. I'm talking just about the technological side, I'm not talking about the policy side. You could argue that it's easier. The reality is you still have these conversion factors. You still have bushels to to to, you know, to liquid or whatever, um and plus. Even even among liquid, like when we talk about upstream, we're talking about pounds predominantly, so, so at that. Or it starts probably in gallons. About upstream, we're talking about pounds predominantly so, so at that, or it starts probably in gallons, then we're talking about pounds, so now we're in in more of a weight um, and then invariably it's in it's in barrels or cubic meters, or you know who knows. So we're still at the. You know, at the end of the day we're still dealing with all these different um conversions, whether it's actual conversion of product or just conversion of the day.
Speaker 2:We're still dealing with all these different conversions, whether it's actual conversion of product, or just conversion of the unit of measurement from one to the other. So, in theory it's easier, but in practice it's not.
Speaker 1:Bringing AI into the equation. It's sort of the buzzword. If you're talking about uncertainty, the buzzword of the world for the last 18 months has been AI and everyone seems to be finding a use case. And where do you think AI can simplify? Streamline help in the traceability of SAF, streamline help in the traceability of of SAF? Or do you think there's an element of people are just you know, you're just trying to find a, a way to jump on the bandwagon, a bit like sort of? You know, blockchain was the buzzword just before AI came out and everyone was trying to get everything on a blockchain. Um, or do you think there are real practical applications that can make a real difference? You know what you, what you guys are doing.
Speaker 2:Yeah. So let me start with kind of the headline, which is AI is going to be transformational for compliance and, in particular, for traceability within our space. We're building AI into our products. We call it Flux AI. We say it's kind of your co-pilot for compliance. That's kind of how we're thinking about it. I'm going to come back to that and explain practically how we're doing that, but I think your blockchain point is a good one and it's kind of a cautionary tale which is we can't.
Speaker 2:First off, if we treat these things as just buzzwords and we don't look at the practical application of it, then we're not actually doing a service to our customers or to the industry writ large. The problem with blockchain was exactly that. There was a lot of systems that came out that said, oh, we're going to be blockchain for supply chains and things like that, and it's great in theory supply chains and things like that and it's great in theory, but fundamentally, a blockchain is meant to be a distributed, decentralized ledger where data gets recorded, and that's not the problem the industry had. It wasn't. We had an issue with the distribution and decentralization of data. The issue the industry had years ago when blockchain was in vogue was there wasn't good data, and so if you're putting bad data in, you were just going to get bad blockchain validated data out, and so it was the wrong focus on the problem set.
Speaker 2:Now, what's novel about blockchain more broadly is the cryptographic hashing and validation, and some of that's new or was new as blockchain was becoming popularized. Some of it wasn't. If you look at law enforcement and kind of how law enforcement was tracking digital custody of evidence, they were using SHA hashes, cryptographic hashes, to be able to validate that data hadn't been altered from when it came into custody of law enforcement, and so those same standards are good standards and are standards that can and should apply to data that we have in the industry. We do that. We actually launched a product earlier this year called Fluxchain, where we do exactly that, but the reality is most of the market doesn't care, doesn't need that, doesn't really want that, so we've only launched it for a select few customers for which that extra uh level of of cryptographic confidence is really important.
Speaker 2:As we look to ai, what's going to be different about ai is customers are not going to know they want it and, in a lot of cases, they're not even going to know that they're using it. Because if we do our job right with our AI, with Flux AI, we're going to incorporate it into a lot of the existing manual workflows to allow them to be automated or semi-automated in a way that has a meaningful practical impact, in a way that has a meaningful practical impact to, but one where the customer doesn't necessarily know that that impact is as a result of ai, and so I think it's. I think it's significantly different. And the other the other thing is today there's a lot of fear around ai as kind of this.
Speaker 2:You know, for lack of a term, this scary boogeyman that's going to replace a whole bunch of human jobs and there's going to be, you know, global consequences as a result of that, et cetera. That certainly may be the case. I'm not here to you know. Try to pontificate on that as a topic outside of my wheelhouse, but in our space. The way I see AI again from a practical application perspective is AI is going to be an enhancer, so your compliance person that's in your enterprise is they're not going to have to hire four more people as the scale of their operations grows because AI is going to enable them to do 4x the amount of work. And, more importantly, they're not, in some cases, they're not even going to realize that. It's not as if they're going to say, oh, I need AI in order to be able to do 4x of that work.
Speaker 2:We as a service provider are going to embed AI in our product, which is going to unlock 4x potential from them and it's just going to be seamless and natural and part of the evolution. And we're at the very cusp, the real beginning of that overall trajectory. So I'm very excited in a way that you know I never was equally excited about blockchain, not because I don't think blockchain is cool or a neat, a neat use of technology, but just because the application to traceability, sustainability management staff, all that kind of stuff, it wasn't there, whereas for it with ai, it's definitely there and we're in the early innings, as we would say in the us and where specifically do you see it?
Speaker 1:you said it could be implemented some places very easily and people wouldn't necessarily notice. What are those sort of first things that you can utilize it for initially, and then look to sort of go from there, sort of as a stepping stone?
Speaker 2:And we're building this right now, which is bills of lading. So you know, any transportation has an associated BL or BOL with it. That includes to from. That includes an amount that includes a product category. You know information like that. If it's a ship, it includes a vessel number. If it's a rail car, it includes a rail car number, whatever it may be. That's all super rich, valuable data that being able to auto ingest and make sense of it is something that AI can do today. It's a low hanging fruit. It's an easy use case. Now, this is where you know. I think it's important to talk about the practical side of things.
Speaker 1:You could.
Speaker 2:OCR, and OCR is kind of the you know, the pre-AI ability to kind of machine read a document and pull it and extract the text from it. So you could OCR, bol wasn't, or BL as it's internationally referred. If the bill of lading wasn't structured in a very uniform format, then the success rate on the OCR wasn't all that great. It would mess up the OCR and so you had to train and train and train and retrain in order to get that kind of success rate, and all you were doing, while it was important, all you were doing was effectively just extracting the data and then you had to do something.
Speaker 2:You would put it into a structured database and you would do something with the data the difference today is what ai can do, is ai can pull that data at a much higher success rate even if it's in different forms, if you know if the chipper name is located in a different place than it was previously, the, the AI can pick up on that nuance and, as it's ingesting that data rather than just putting it in a structured database, it can also make sense of it.
Speaker 2:It can say oh, I see a transaction between these two counterparties. I know in this other relational database that this transaction exists because I pulled it from an ERP or from some other system. I'm going to reconcile those two. I'm going to attach this BOL or this BL to that transaction and then, based off of automation logic that the user has written, I'm going to do X, Y, Z with that particular data. All of that is feasible today. It's what we're building into the system today and it's a very, very practical and impactful application of AI that'll save on a per BL BOL level. That'll save minutes worth of work In aggregate, will save hours and potentially days worth of work. And that's just one example.
Speaker 1:And when you say it has minutes, collectively days saves that amount of work. Is that transferred into the speed that people see actually buying stuff? Will this have tangible differences in the speed of the supply chain? Is there other occasions where things are slowed because you're trying to get all the data aligned and get all the documentation for for things in order, or is it more a case of verifying and making sure the security of what you're actually transporting is what you say it is, or is that sort of?
Speaker 2:both of those sort of aspects come into play it's both, and the reason why it's both is there's already a lag effect around compliance data. Um, you know, in a perfect world, everything would be real time. And 15 years from now, when we have endless sensors and endless AI, maybe we'll get there.
Speaker 1:Certainly my dream.
Speaker 2:But we're not there today and so you have a lag effect on data. It's why you know you typically see verifications or audits that are a quarter or a year behind. You know today as a point in time, because it takes time for people to compile all the data, get it ready, make sure there's not any gaps, things like that.
Speaker 1:So what I'm talking?
Speaker 2:about speeds up that flow. The commercial implication of that is a lot of times you're taking on you as an economic operator. You're taking on more risk with product by not knowing with 100% confidence its eligibility or that you have all the data to meet certain markets. I mean you and your gut know right, I'm buying from supplier A. I trust supplier A. I've got a long relationship with supplier A. I know they'll get me that data. I know they'll get me the ISCC proof of sustainability or whatever it may be. So I know in my gut I'm going to get it.
Speaker 2:But that's not zero risk. There's a chance they don't. We've seen situations where companies go out of business and so they were well-intentioned in being able to provide that data, but they don't exist anymore and so, as a result of that, any lag becomes increased risk and, by extension, if you apply that risk forward, becomes decreased optionality, because you don't have 100% confidence about where the product can go because you don't have all the data associated with it. So as we close that lag factor and reduce those inefficiencies and reduce that time using AI, it'll lead to more optimized commercial decision-making.
Speaker 1:Is there any way you wouldn't want to use AI where you think I'm not sure, actually, that actually either has much benefit or you're just slightly wary of because you're not necessarily sure you can trust it at this stage. Maybe in five years' time maybe. But are there any where do you think I'm not, I'm not sure it's that's a good use case for AI. So yeah.
Speaker 2:So I think the way to think about AI because there are genuine concerns around hallucinations or around the AI doing things that aren't um 100 accurate and in our industry right you certainly can't afford that. So I think the way to think about AI is taking a compartmentalized approach. It's why the bill of lading is such a good example you can test it. You can back test it, you can train it, you can further refine it.
Speaker 1:And you can get to a point where you say, I've got 99.999% confidence that this is going to yield an accurate result.
Speaker 2:And when you get to a certain threshold, especially for a segmented use case, then you feel comfortable and I think that's the right approach.
Speaker 2:The way I would think about it in terms of where we are with AI is right now, if AI is a house that you're building, like we just dug the, you know the basement and haven't even laid the foundation yet, and so identifying like very specific use cases to kind of start with, are the equivalent of starting to put in the foundation. There's also an important side of like making sure that your data is good, that it's consistent, that there's. You know that you're bringing your data in in ways that can all be managed properly, which, as a system provider, we do that kind of for you. So it's kind of built in, but that's an important part. Once that foundation's in place and once you've kind of laid that groundwork, then you can start to build, to use the house analogy, then you can start to build the living room and the awesome master bedroom and the really cool bathroom and all that stuff, and so that's that's where that's where ultimately AI will be five, six, seven years from now, where we're just laying the foundation today.
Speaker 2:So anyone that comes to you and says oh, I'm going to have AI run wild and it's going to be executing your trades and it's going to be doing this. I'd have a high degree of skepticism, not because we can't get there, but because we don't want to be there today, because a lot of this stuff is still in its infancy.
Speaker 1:Are you concerned to sort of build on your house analogy? Everyone's had a contractor or a builder that's come in and cut a corner or so, or an electrician or a plumber. Are you worried about the sort of AI potential to become lazy? Because it's one of the sort of most interesting and strange things about AI that has a tendency to, over time, develop certain lazinesses and not be as thorough as it potentially was before. Is that something you sort of also need to be aware of, that you actually just can't totally leave it because you know at the stage we're at, it still actually needs to be sort of monitored, albeit from a distance?
Speaker 2:yeah it's, it's a good um. I really like the contractor example. The way I would think about it is, when you're laying the foundation, it's as critical of a point as any to make sure that you have a contractor and, probably take it a step further, an overall project manager. When it comes to building a home, that's making sure that you're using the right vendors that are upholding the right standards. I think, unfortunately, in the commodity space, there's often a piecemeal approach. Folks have an immediate problem and so they try to fix that particular problem, like a plumbing leak kind of analogy. So I think a lot of where we're in today is folks have a problem with EPA record keeping. Where we're in today is folks have a problem with EPA record keeping, so they're like I'm going to adopt whatever. The simplest, cheapest option is for EPA record keeping. That's great.
Speaker 2:Do that. But if you're not doing that, with also the foresight of okay, here's how this solution is also adapting to across these other markets, but also, but in addition to that, here's how this solution is laying the framework for an AI first world that we're headed into then all you're really doing is setting yourself up to switch vendors in two to three years, and I like to joke that the only thing that people dislike more than you know some of their software programs is the thought of having to switch away from them, and so that's why I would.
Speaker 2:that's why I would. You know some of their software programs is the thought of having to switch away from them, and so that's why I would. That's why I would, you know, I would caution folks as they're looking at the realm of possible, whether it's in the traceability and compliance management space as it relates to us, or whether it's others in similar spaces that are looking to apply AI. I'd be very thoughtful and do a lot of due diligence about who that provider is, because they're going to be more than a provider, they're going to be a partner, and if they're not, then they're going to be the shady contractor that screws you and leaves you with a half-built house at the exact time when you were expecting it to be completed.
Speaker 1:And that's why it's really important to be polite. So do you say please, and thank you.
Speaker 2:When you use chat, GPT or other AI softwares the large language models that are available I always say please and thank you. One of my closest friends growing up said please and thank you all the time and it got him very far, and so I model that, that behavior, including with, uh, with AI, because it knows everything about me at this point and um, and so I don't expect when, when the AI takes over, um, I don't expect that I'll live long, but I should live longer than others at this point.
Speaker 1:So it's a worthwhile investment in saying please and thank you to your, to your AIs Um. That's the public service announcement for this.
Speaker 2:Yeah, psa to everyone If all this goes south. It's the old joke, right, you don't have to outrun the bear, you just have to outrun the other people.
Speaker 1:If my.
Speaker 2:AI likes me more than your AI likes you, then I think I'll be okay for a bit.
Speaker 1:Yeah, I feel like I need to sort of send an industry caution at this point just to say everyone, you know it's early stages, it's not everyone will panic, but yeah, yeah, I mean listen, the reality is, this is, this is the new world that we're in.
Speaker 2:And you know, if we want to, if we want to take a step back and think about it at a macro level, like I think the benefits outweigh the costs in terms of just using ai in general, um, I use it for a lot of things, both personally and professionally, um, and I think in the sustainability industry in particular, uh, it's gonna have tremendous impact. I joke, you can't spell sustainability without ai, but you also can't spell without a bunch of other letters.
Speaker 1:so I'm also going to say that, at this point, the questions of this podcast are courtesy of AI. This whole conversation. You and I aren't even talking.
Speaker 2:Yeah, you think I'm kidding, but the Sims thing is is real, my sister is a little bit tangent. But my sister works in sales and they train all of their folks using using Sims that are based off of kind of AI understanding of of what their sales strategy is. So I mean this whole podcast could be fully manufactured as far as the audience is concerned.
Speaker 1:Yeah, but I'm going to assure everyone now we are totally real. You're not listening to me. We are real people.
Speaker 2:Absolutely, you can't you can't manufacture this type of beautiful voice.
Speaker 1:Oh, absolutely not. Widening sort of AI beyond sort of the traceability angle. Are there any other sort of? When you think about the SAF industry more widely, are there any sort of prime use cases where you think this is screaming for stream being streamlined for AI? Sort of my immediate sort of? The one that comes to mind is feedstock identification when you're going to a project. Surely that is something that could be massively streamlined. When you're sort of looking at doing feed studies by sort of using AI, because all that information somewhere for sure, I think.
Speaker 2:I think feasibility studies, you know, I think we're going to get to a point where a lot of the FID stuff gets shortened because of the ability to do really quality research and assessments a lot quicker.
Speaker 2:You know there's already steps when you talk about the deforestation regulation, eudr that you know there's already ways that imagery is being analyzed by AI to be able to prove or either or disprove, claims around sustainability. Google did some interesting work with contrails for airplanes um using using its ai and analyzing um images and videos of of airplanes uh to be able to determine how to reduce emission factors based off of that. So there's going to be endless applications around all of this.
Speaker 1:It's going to be really interesting and now again, sort of I'm asking you to bring out your ability to be Nostradamus Five years in the future. Where do you see? Where do you see, or where would you know, yuko or a bushel of wheat and going boot? This is that, and it's got all the data on it, or it's you know? Where do you, where would you like it to be, so that you can sort of spend more time mowing your lawn?
Speaker 2:yeah, I, I think that, um, I think that between trends in ai and trends in iot, which is an acronym for Internet of Things, but really sensors, I think between those two developments, five years from now we're going to live in a world where there are really really efficient and advanced supply chains from a traceability and compliance management perspective, there's still going to be some that don't.
Speaker 2:I mean, just like there's inequity globally, there's inequity in terms of technical advancement, and so we're going to see certain regions that are going to still rely on paper, still rely on spreadsheets.
Speaker 2:But in some ways they might have the most opportunity in all of this. If you look at mobile Africa, you know to use the continent as just kind of a broad stroke, but Africa skipped desktop, like they went straight to mobile, which was a huge enabler for the continent. And so I think, in the same way, like AI, is going to be an enabler for a lot of folks that want to have access to markets in Europe, in the US, in Canada, in Southeast Asia obviously depending on what the regulations are but are going to want to have that access and it's going to be a force multiplier for them to be able to do it. The counter to that, a force multiplier for them to be able to do it, the counter to that and I think the concern in all this is, ai is also an enabler for fraud because it allows you to generate data, and data at scale.
Speaker 2:So think about if you're trying to be able to prove points of origin. You can go into ChatGPT and say, hey, give me a list of you know a thousand restaurants in X country that I can claim that I collected from and put it in a spreadsheet, and so what was previously five hours, six hours, of manual fraudulent work, is now 30 seconds 45 seconds of automated fraudulent work. Seconds of automated fraudulent work. Which means from a regulator perspective, you're going to have to make sure that you're using technology first detection mechanisms to try to prevent that. It means, from a service provider perspective, we're going to have to up our game in terms of kind of the baseline data that we're allowing into the system and how we validate that. So it goes both ways. But net, net. I'm bullish on AI and I think five years from now, between that and the trends in IoT and Internet of Things, we're going to have some pretty cool data-rich, high-fidelity data supply chains.
Speaker 1:It's undoubtedly sort of going to be a fascinating and revolutionary five years in so many different aspects. Now it's going to be a fascinating and, like, revolutionary five years in so many different aspects. Now is going to be front and center. I think everyone's in in full agreement on that. And speaking of things that are sort of going five, six years in advance, you mentioned 45z or 45z um earlier. Uh, right at the very beginning and we we're recording this on the 16th and earlier this week, there was an announcement about it being reinstated until 2031. And there being some very favourable signs, me and you spoke offline earlier this year and you predicted that it was going to come back and that it was going to be till 2035. So you're four years too ambitious, but nevertheless it's a very sort of encouraging sign and a bit more policy certainty, whereas before there was a bit more uncertainty till now.
Speaker 2:Yeah, and so you know. Let's kind of break down where we're at, and again, as you mentioned, it's the 16th of May and we're early on in what we would call the reconciliation process for passed in the coming weeks and months. And what you're referring to for the 45Z or 45Z adjustment modification, is in a version of the tax bill that passed through the Ways and Means Committee in the House of Representatives, which is one of our two chambers, and so it's the first indication of what the future of 45Z could look like. It still needs to make it through the full House of Representatives it's just one committee. It still needs to go through the Senate. It still needs to go through reconciliation between the House and the Senate versions and that ultimately needs to be signed from the president.
Speaker 2:So while this is a very, very good indication of what's to come, it's by no means the final thing. But what's included in it? What's included in it is an extension through 2031. The current one was set to expire in 2027. It makes the producer's credit. Today is domestic only on the fuel side. It reinforces that. It extends that to feedstock from a US, canada or Mexico origin. So that's a meaningful change. The initial guidance on 45Z in January restricted foreign-use cooking oil but allowed Corsia material to qualify as well as tallow. Today, with this modification, this proposed modification, none of that would qualify outside of mexico, canada and the us it also changes um some of the indirect land use criteria values, iluck.
Speaker 2:it removes it, which allows other feedstocks, like canola, that previously didn't qualify or had a hard time qualifying unless you added in climate smart agriculture, carbon sequestration additions to it. It now makes it so those qualifying have, along with soy, have a much lower carbon intensity score. So overall that's kind of the adjustments that have been that have been proposed thus far and I think it gives. Put aside the CI score, put aside domestic feedstock. Just talk about the extension. I think the extension gives a lot of policy certainty and the fact that they're laying down rules of the road hopefully will give a lot of policy certainty to the business community writ large, and so credit to the Congress and ultimately to the administration if this does get passed and that we do have a better idea of what the future is, whereas previously we were kind of in this ambiguous state of initial guidance in January lack of clarity on what it meant, et cetera.
Speaker 1:Potentially the biggest, I think the most interesting and sort of biggest change that we're seeing, apart, you could argue, the extensions, the biggest one is this indirect land usage adjustment because it opens the door for so many new feedstocks and makes previously what were very high value feedstocks slightly less value, slightly evens the playing field, opens the door to a lot more different feedstocks that can get much more benefit from these tax credits or a bit less benefit. So it does change the the landscape slightly in terms of those feedstocks yeah, I mean it changes it monumentally.
Speaker 2:Um it, it makes uh, it makes a lot of the future of renewable diesel and SAF in the United States crop-based. There'll still be meaningful room for waste-based feedstocks, but it puts a heavy emphasis on crop when you remove the indirect land use criteria. Now that said that does run in conflict with California, which has, among other things, a proposal for a 20% crop cap per facility, per production facility.
Speaker 2:It does run in conflict with some of the European regulations, not that the federal government necessarily cares, but it just speaks to our earlier point that there aren't going to be alignment between these programs, and so both as it relates to how you make decisions around what your feedstock sourcing is going to be alignment between these programs, and so both as it relates to how you make decisions around what your feedstock sourcing is going to be, but also as it relates to how you collect data and quantify that data is still going to be, impacted by trying to have optionality across these different marketplaces.
Speaker 1:It seems like it's suggesting a slightly longer, a slightly bigger shift away from sort of the European perspective on, you know, the the value of crops as a feedstock, and a slight sort of you could see the alignment with sort of Latin America and that becoming a lot stronger because of this yeah, for sure, and I would add to that that I think it.
Speaker 2:you know, I think the core proposition in a 45Z that is similar or fully aligned with the current proposed version. I think the core proposition that's being communicated by the U president's executive order on this, which included biofuels, among other types of fuels that are looking to be made and ultimately exported from America, so to unleashing American energy and and to is supporting domestic industry, whether that's the waste based side of it and the recycling side of it, or whether that's the waste-based side of it and the recycling side of it, or whether that's the crop-based side, but basically leaning into American-made American production, et cetera, which is why it's called the producer's tax credit, and it's why I'll just say I do think you're going to see similar trends on the technology side. I think there's going to be an expectation that, as you're trying to manage your compliance with all these programs, that you're using American-made technology to do it as well.
Speaker 2:There's already labor requirements built into 45Z, and I fully expect those same types of made-in-the-USA requirements to apply to technology too.
Speaker 1:And, on that note, we've covered traceability with a large sprinkling of AI and then finished off with a little bit of policy discussion. Donnie, thanks so much for coming back on and giving us a very interesting discussion and a very interesting discussion.
Speaker 2:Yeah, I appreciate it, oscar. And to use one more American-centric metaphor in the US, winning back-to-back Super Bowls, winning back-to-back NBA championships it's impressive, but it's not everything. Until you do the three-peat, you're really not a champion. So until I'm invited back for a third time, I'm going to hold out on putting the honorarium on LinkedIn. So just throwing that out there.
Speaker 1:Well, everyone watch this space and there's far more discussions like this to come next week when we have our SAF Investor London Conference, which again we encourage everyone to come because you get to meet people like donnie and hear a lot of the latest industry insight about how to get investment into saf and all the other critical components that need to be in place. So find out more information about that in the description. Thanks so much, don, tony. Thanks, oscar.