The Latest Advancements in AI for Food & Beverage Production with Icicle Technologies

The Latest Advancements in AI for Food & Beverage Production with Icicle Technologies

Show Video

Hi everyone. Welcome to another webinar in the Food Industry Executive Series. My name is Amanda and we're excited to have you here with us today.

I wanted to go over a couple of housekeeping items with you to start off. First, if you have any questions, please drop them into the questions tab at the bottom of the webinar. We will also be sending out any recordings and handouts after the webinar. Steven Burton is the creator of the award winning food production management system Icicle.

Icicle offers a complete solution for smart automation, improving quality standards, production efficiency, and expanding growth opportunities for all types of food businesses. Now with AI-enabled features, Icicle help food and beverage manufacturers protect profitability and move their business forward into the future. Steve's background in manufacturing and consulting, combined with sophisticated software development expertise, gives him a unique perspective on today's topic.

And with that, I will pass it off to Steve. Thanks, Steve. Hi, everyone. thanks for the introduction. Appreciate that. I am Steve Burton, and today I'm here to talk to you about, artificial intelligence and specifically its applications in the food industry.

And we'll take a look at some of its inner workings, will shed some light on its capabilities, and look at how it's used in various domains within the context of the food industry itself. So first of all, we are Icicle Technologies. So, as I mentioned, we, we touch pretty much every aspect of the food production process all the way from product design through to, hazard analysis and, supply and generation. then to, you know, receiving, production, shipping and traceability, all of that. So we have a kind of a holistic view of the industry as a whole. So that's kind of useful when it comes to figuring out where I can be applied.

One of the things that I wanted to point out before we get started here is that it's a kind of a scary concept, embracing, I think it's time that, new technologies, disrupted. And in fact, many times in the past, it has. So the telephone, you can see this quote here from, William Morton, the president of Western Union, the union who said that the telephone, has too many shortcomings to be seriously considered as a means of communication, which, didn't work out that well.

So I think AI is the same kind of category where, you know, there's a lot of people now see that it's kind of coming down the pike. but, there's still people holding out. And I think the best thing to do at this point is really not to be too afraid of it, but to figure out how best to embrace it and to use it to really help you. So first of all, I just wanted to talk a little bit about what it is I so just as a heads up, what I'm going to try to do here is just give you a brief introduction to what AI is so that everybody who might not know, get some sense of it, and then we'll dive specifically into the food industry. AI is not really a finished, really, really is is just the the simulation of, human intelligence, by machines.

So that's it's really more of an objective or a goal as opposed to an actual specific technology. There's a lot of people are currently quite afraid of this. There's a number of reasons, you know, some of which are more critical and significant than others.

But, you know, data privacy and security is one of the problems that, people are worried about. And, there's absolutely, something to be said for that, you know, especially large language models that, assimilate, the entire internet, you know, they could come across proprietary information that can be disclosed. So we have to guard against that kind of thing. People may become quite depending on this technology over time and sort of abound in their own decision making skills. I would propose to you that AI is more of a prediction engine, as opposed to a decision making technology.

You know, I wouldn't rely on it to mix because retain your own faculties of decision making. But I think the dependency is also a risk that people might read it to a loss of control. people fear that they may have loss of control. Usually when new tools come in, people do fear that. But ultimately it turns out that it's, you know, actually allows more control.

AI malfunctions and safety concerns. So what happens if it goes wrong? right now we're in a fairly early stage in AI development. So that is definitely a concern that you have to, you know, take it with a grain of salt or if it's a device where it's, you know, it's ready to a specific device. You know, those devices are usually only doing one very small, specific task. So that kind of mitigates the risk of malfunction, loss of jobs through automation.

this is, probably the most significant issue that, we're going to currently face in the next decade or two. Bias and fairness. There's a little bit of a two edged sword to this one. So a lot of people are worried about bias in AI.

and they have the notion that bias isn't that AI is inherently biased. And it is true that there's trained on data sets usually. And if those data sets were generated by humans, you know, bias definitely could, could happen.

There's also another kind of bias that you have to watch out for, which is if you some people disagree with AI's results and, just because you disagree with it doesn't mean you lose clients. So if you look at, you know, think about, for example, Google, Gemini and what happened without regulatory compliance, there is going to be probably more regulatory compliance. even though we have today coming in as a result of AI. also regulators will use AI. So it may be, more difficult to kind of get around certain things that we're used to being able to get around, integration challenges. There's going to be a lot of integration required to connect everything together.

If you really want to use AI effectively, and there could be unforeseen consequences, you know, who knows what those are. So, in terms of the fundamentals of AI, there's right now there's actually three types of AI. We're actually forever. This will be called there are. So there's narrow AI, which is also called weak out. What that is is it's AI that does one task very specifically.

So think about Siri. Think about, you know, self-driving cars, very targeted, kind of limited, not general. And then general artificial intelligence is really summation of human intelligence.

So once I have capabilities that allow it to operate generally and not solely in specific domains, that's going to be general artificial intelligence. And we are not there yet. best guess is probably 25 to 50 years before will achieve that.

So we're quite a long ways out from that. And then the big thing that a lot of people are worried about, superintelligence, that is, where the AI is and passerbys will surpass human intelligence and probably at least, 50 years, more likely 100 and maybe never for that. but, you know, nobody really knows for sure, but it definitely isn't, tomorrow or the day after then. In terms of deployment types, there's two different ways that I can be deployed. Things like chat, GPT for that most people are familiar with. That is a cloud AI technology.

So basically, you know, there's some sort of a query that's sent up into the cloud. The processing is done in the cloud, and then the result is passed back to you through the your type, which is, also very germane to the food industry is AGI. And all that means is that, you know, you have AGI. It, AI running on a local device of some kind that, interacts with the outside world.

So, like this, image here is for a, an acoustical sensor, for example, where, you know, you might mount that it might have the ability to listen for specific sounds, and then, you know, react based on certain codons if those codons should occur. So think about when you say the, you know, hello, Google or something on your mobile phone. You know, there's actually a chip that listens for that. And that's all it does and that's all it can do.

And so the, you know, it's not like every conversation that you have is going up to the, you know, your cloud provider. It's only listening for those certain words based on that chip insight. Another thing that's, you know, people might not be aware of is that there's often a big time lag between when a new technology is initially introduced and when that technology kind of goes mainstream. AI is, just recently come on the radar for a lot of people, you know, people have heard about it for a long time, but never really, thought about it or the use of it themselves until quite recently. But it's actually quite an old technology already. People have been working on it aggressively since 1965, and it was 1922.

And it really, sort of 2022 when it actually hit the mainstream. So there's a very big time lag between those, things. So we're already quite far down the road in terms of the development of AI itself. There's quite a few limitations of AI itself.

And with ChatGPT, for example, just, just to have some fun, I decided to ask ChatGPT a question. You know, what are your limitations? what it told me is that my intelligence is different from human intelligence. I can process and analyze vast amounts of information, quickly, perform complex calculations, and generate responses based on patterns in the data that I've been trained on. However, my capabilities are limited to what I've been perform, programed, and trained to do.

I don't possess the consciousness, emotions, or the ability to understand context and nuance in the way that humans do. I also can't generate new ideas or experiences, learn in real time, or make judgments based on practical considerations or common sense without explicit instructions. So that's quite a telling, you know, long statement. I'm not going to quote lots of stuff like that for you today, but I thought that would was an interesting one. One of the main takeaways there that you should really understand is that if you cannot learn in real time, so you might have heard the term, for example, machine learning, or you know, and I think that or even conversed with ChatGPT and then thought that, you know, it's, it's assimilating and understanding what you're saying. It's pretty good at not predicting what you might want to hear, but it can't assimilate that information.

you know, if you give it feedback, it can only the next time the entire model is trained, then it can incorporate new knowledge. But it's do not, you know, every time it talks to a user. So it's quite it's kind of a train it and then, it's fixed in time after that. So it's really quite a lot more limited than a lot of people might think. And in terms of limitations, you know, here's an actually a real example from our production system. So this is, production data.

And it's, it's a graph of customer complaints over time. So the blue line there is representing the number of complaints that, have been received from customers, you know, on the daily basis and the, the, the kind of orangish colored, tone at the end is a code of probability that's extrapolating into the future. And it's kind of giving you an indication of what sort of range of, values you might expect to see going forward. So it's making a prediction that, and, you notice that the cone, you know, gets wider as it goes farther in the future, which makes sense because, you know, the farther out we go, the less certainty we have. But, you'll notice that it drops below the zero line.

So, you know what? What it's really saying is that there's a non-zero there's a probability that there could be negative complaints, which doesn't really make sense. So this is a pretty good demonstration of how, AI doesn't necessarily have common sense. Right.

so something to keep in mind when you're using AI. So let's go now a little bit more specifically into the food industry and talk about, you know, why you might want to use the AI in the food industry. So the meat really are that you can you can achieve higher accuracy in terms of predictions. So that's that's a big one. And that's probably one of the main words is that because you can the systems can assimilate vast quantities of information.

You know, trillions of records. And in the case of of ChatGPT type models, or or, you know, all of the existing data that you have available historically and then it can, it can, use machine learning to kind of anticipate, identify patterns. And then later on, if you input can use all of that knowledge that's been gained to generate outputs. So the prediction accuracy of the predictions, is significantly higher. You know, if it's properly trained with the right data, you can improve your production, velocity. so you may be able to produce a product faster.

it's constantly attentive. So if you have been thinking here primarily about quality assurance tasks. So if you have an employee who's, you know, doing some quality control checks and, you know, they get distracted for a few minutes, you know, you can have product running by that's not necessarily checked. whereas it's it's always going to be checked if there's already, and you're done with your, resource utilization. But these are, these are, there's other areas as well.

But this gives you a kind of a high level taste of, you know, four major areas where you can really see significant ROI from, using AI. So I now I'm going to talk about some specific use cases, and then we'll talk about, we'll do some case study. So you talk about some, cases where AI has been applied okay. So demand forecasting. So this is one area you can, currently like our production system prior does set state of observatory for all existing temporary that sort of thing to come up with with some sort of forecasting, that can be used to forecast production.

you can add additional information even from non structured data or, you know, unexpected shocks to the economy, things like that. that otherwise you might not be able to, incorporate, supply chain optimization. also a little bit of a to which sort of, you know, with better than before because you can also, have that are just in time inventory, you know, procurement, for example.

So you can order only what you need from your suppliers, you know, in a more sort of targeted way. But, you know, we also have to consider kind of what the rest of the supply chain looks like. Like, for example, we have customers who make prepared foods that includes salads and, you know, to have very targeted, forecasting for their lettuce requirements is really great. But we also have vertical farming operations that make lettuce, and it takes them 18 weeks to produce the lettuce. They have to grow it. Right.

So there's limits to flexibility of the supply chain. So it's not always quite as good as that might look like. energy consumption prediction. One of the case studies I'll talk about, shows how that kind of works out, price trend forecasting. So that's, kind of an interesting one as well, where you can, you know, perhaps get a better indication of what the competitive prices and the consumer market might be looking like.

And then, product development forecasting, too. So one of the case study this I wanted to talk about is a company called snowcat, who's one of our clients. And, one of the things this is an ongoing project, very recently started project, really, that we're doing for them, which is, pallet design. So this is a major, Canadian retailer. sorry, not retailer, but distributor who, shipped out about 1500 shipments a day, and they ship out trucks, Todd.

Sorts of different routes. And the the trucks have mixed orders for different customers, and they have multiple pallets on also, which may be mixed for different customers. So the currently what they, what they do is that they have a, they get other orders, they lay them out in a big table, and they have this guy that's like a savant that comes into the room, spends 3 to 6 hours kind of rearranging all the orders until he gets to the right kind of sequence magically. And then, they, they load the pallets, based on his recommendations and, that's obviously not very sustainable, not very scalable.

so they're they've got a brooding software, a system, one of our partners that we're working with, we've had an integration with who's doing the rooting for them. And then we're doing the pallet design based on AI. So we'll determine, look at various parameters, like the weight of the item, the for for gelati of the item delivery order. You know, how high can the pallets be? What are the dimensions constraints of the truck? What is the weight limits, temperature constraints because you've got frozen and refrigerated and dry products all together, you have to consider all of those and then come up with a pallet design. So you tell the pickers to load the pallet exactly.

And then, you know, maybe someday we'll have robots loading the pallets. So another, kind of use case we want to talk about is kind of profitability. So, you know, how can you make more money for me, which is obviously everybody's, goal. And, one of the case studies we came across.

And by the way, just as a side note, like finding case studies that are not just pilot projects are not opposed, but actually in production, it's not quite as easy as one might think. There's, there's a lot of, pieces of the puzzle that are being put together, but it's not always a complete enough picture to go into production. And I'll explain why I think that might be a little bit later on. But this company, Corolla, actually came up with a system that they think they have a partner for on the AI side. And Tello Labs in the, in the company.

Previously, their employees were going out to, farmers and then, analyzing the, the car demand manually and then grading, you know, each bin and coming up with some sort of classification, you know, which, leads to a certain price. And, instead of that, now they can just snap a picture with their mobile phone, they send it up to the cloud. So that's, that's cloud AI as opposed to AG, and, in 45, 55 seconds, they get a result, doc, that tells them exactly what the quality of that particular grade was, which saves them an enormous amount of time and money.

So there's quite a lot of opportunity also in food safety and quality control. The main areas there where you may want to use AI is in the, contamination prevention piece. So you know how to stop contamination from occurring. So this would be, you know, biological, chemical, physical contamination, radiological, automated visual inspections. So it was just study, the United States last week where I saw a, five people working on the same packaging machine, and, one was operating the machine, and the other four were doing visual inspections, all of which, could be replaced by AGI devices.

Predictive hazard identification. So, being able to kind of advanced, determine in advance with what hazards might, there might be. And this is actually something that we do already. So you can enter into raw material. And this then will suggest to you what sort of biological, chemical, physical, radiological hazards might be associated with that. And then once you identify them and you know, you need to again, humans need to AI at this point in time.

So, once you identify what those hazards are, you can enter them as hazards and use their hazard control mechanisms to come up with your house controls. automated reporting. So this is, this is a good one because, you know, typically what we do is we will send out, automated reports on a scheduled kind of basis. And we also send out alerts when events occur outside of some sort of boundary. But, having, reports that go out to you specifically when you need to see them, you know, based on AI is something that could be very useful for people, defect detection. So that, packaging line I was talking about before, they were looking at the ceiling of the package, you know, that was, a product where there needs to be a hot dog sticking outside of a pastry and to be able to measure exactly how far out that, hot dog is.

Not too far, not too little, that the kind of, defect detection that can be easily hardcoded into an edge AI device, then replace that individual. so, a case study kind of on the QA side, this is an interesting one where right now, if you're going to do swab tests, everybody needs to do swab tests. So what you do is you swab the, through contact or adjacent surfaces, you put it in a bag, you may have your own lab. you may not be. You may not have a lab.

So you have to send it out to a third party. And then 2 or 3 days later, you hear kind of what the result is. with this, what you're doing is using the light scattering, through a laser and, essentially being able to detect causative genic bacteria on the work surface without a swab required. And, there, has been a successful implementation of this where they've achieved accuracies between 80 and 100%.

anybody interested in finding the study is welcome to reach out and we'll give them the reference. but this can lead to a significant savings in time and cost compared to traditional methods. It's also an interesting one, because we actually did this same project in collaboration with the University of British Columbia, or maybe the other way around the University of British Columbia did it in collaboration with us and also, the research, Industrial Research Council of Canada. also around the same time frame, 2019, our, our project, did not succeed. Unfortunately, we were actually able to get it to the point where we can identify the bacteria.

but we didn't have enough resolution to be able to determine the species. so that was, the showstopper for us, but apparently somebody else has succeeded. So. Good on that. a couple of places where we use, AI is, in the development of SOPs. So you when you're creating a copy for each section, gossip, you can use the chat model to, suggest to you what the text should be.

And that saves you an enormous amount of time in it. And I'm going to caution you one last time here that, you know, you need to vet whatever the AI tells you because they do hallucinate. But, it definitely saves an enormous amount of time because you just now you're just doing editing as opposed to try to figure out how to do things from scratch. Another thing that we do is hazard identification. So as I mentioned, you can put in that raw material. it'll give you a description of exactly what hazards, you know, are typically found in that type of product.

And, and then you can choose whether or not you want to incorporate them into the, hazard section, which is on the right hand side. It takes into the, AI systems that. optimizing, production, is, is another, use case. And, you can, you know, the main one that people think about, during, in terms of optimizing production is, equipment automation and robotics, of course, which is advancing in leaps and bounds. I saw a very cool video yesterday of a, of a, food related, robot, essentially that, you can speak to because it's, it's, enabled and it can respond in quite, but, convincing natural language.

And it has significant dexterity as well. five fingers with three joints. So, really, quite a functional piece of equipment. I wouldn't say quite ready for prime time, but, getting pretty darn close.

product customization. this is a hyped quite a lot in AI, you know, being able to design products for particular individuals. Not quite as big of a believer in that one. I'm not sure how much people are willing to pay for a product customization, you know, do you want to do you really want to have cookies with you know, exactly your own logo on them? You know? Yeah, maybe once in a while if you it's a company trade show or something, but it's not something that you're going to decide to, have a different foot, colorful cookie decoration every, every Friday night.

So I don't think it has mass appeal, but I could be wrong. Probably just for, real time monitoring with sensors. so this is a huge one, because a lot of times, you know, when your production line is running, all of a sudden the registration of your film is off or something like that, and then pretty soon you got a whole batch of products that need to be repackaged or remade or discarded. So it's a massive one, automated material handling.

It's kind of a hard problem to solve. There's a lot of people working on that. Hugo is most notable for that.

They have all sorts of logistics and material handling. I, I have seen it, it's a successful implementation. interestingly, not quite a big tech company, but by a small company that decided to build their own robots on rails. And, they used it and it works quite well. So I think there's a lot of opportunity for that.

And in terms of optimizing production, you know, one of the case studies that, we came across that, worked out really well was a US based cheese manufacturer who used machine learning models, and it took in about six months. They, the AI, learned, you know, kind of all of the historical data. And then they it was able to process 29 different variables, looking at the amount of starter culture, mixing times, the raw milk composition. And it, it came up with a suggestion for how to optimize the final moisture content.

And as a result, the manufacturer was able to reliably increase the average moisture content by a whopping 0.6%, which is, obviously did see very impressive when I first read that number being so small. however, that corresponds in their case to $1 million a year in extra revenue because they sell their cheese, by the pound. So good on them.

another very major area, where AI is going to be applied is in maintenance, so you can do losses with maintenance, you can do predictive maintenance for equipment. you can and that's a big one because, you know, within cities, for example, when they need to do maintenance on late, you know, light bulbs on the, on the street lights, they, typically just have a regular schedule and they'll just go down the street and they replace every light. whereas if you could be very, you know, with smart city initiatives already, people, people are using AI to kind of predict which lights are going to fail when it's still go out and do a big round. But they don't have to change every light, you know, because if they if they a reasonable degree of certainty that those, you know, the, the majority of the lights won't fail within a month if they run around once a month and, they don't have to replace all the bulbs so they can save an enormous amount of money.

Same is true in a production facility. If you know when your parts are going to fail, you can you can both. You know, make sure that you replace only the and, also before they fail so that you don't interrupt production. training and support. So maintenance staff. So think of, chat, chatting to a maintenance, air duct and, give you answers to your maintenance questions, root cause analysis.

So to be able to analyze data to figure out where the problems were caused. And, you know, the first place that, resulted in some sort of a deviation non-conformity, equipment failure, optimizing maintenance schedules. in the past life, I did software for the medical industry and, to try to get, you know, all the doctors and nurses and technicians and equipment and patients into the right room at the right time was a major, major, major problem. so, being able to have a AI that helps you optimize the data schedules, you know, based on your available workers and their of skill sets is a huge advantage, enhanced safety monitoring. So being able to, you know, can have sensors, for example, that can detect when employees maybe enter an area that's, unsafe, and send off some kind of warning, is something that could be quite helpful. so another case study in this is, a company called, Tetra Pack Services that, published a little peer reviewed journal about a study that they did, where they actually, were able to predict specific equipment issues and, downtimes within their case, near, 100% success rate using AI and cloud based training, with some collaboration, from human apps and, that enabled them to fix maintenance intervals partially with data based predictions from sensors by measuring temperature and vibration profile, as in another past life, I had a production facility and we had a machine that was malfunctioning.

it was related. We didn't know what part of the machine was malfunctioning. So we hired a company to come in who set up vibration probes on our machine and, managed to use the frequency of the noise that it was making to figure out which component had failed.

And then we replaced it. just cost us a couple thousand dollars. You know, you could probably put a $20 sensor on the machine, and you know, accomplish the same thing today. another place where you can use AI is in, product design.

So, taste and flavor optimization. texture and appearance analysis, to give you a little bit of a capability of some of the AI in terms of product design and kind of anticipating what things might be, might be like. I mean, there's a lot of, talk about protein advancements in the medical industry. in the pharmaceutical industry, there was a company that, spent about six hours looking for harmful toxins, and they identified multiple different, chemical warfare agents.

So in just a six hour period, based on their AI model, which is a good example of where I, should not be used, but, it was an interesting experiment and, you know, did highlight the risks and how we need to be able to control that, but using it in a positive way to enhance our taste and flavor and textures and appearance of our product is, thought about it idea and also nutrition. So we might be able to, optimize the nutritional content of a product. think about, you know, nutritious, fantastically nutritious junk food, you know, for kids when, like, consumer, consumer insights, and, and trend predictions, are another area that, you may be able to use. I so a case study there is, a beverage company. Funny, I looked quite extensively and couldn't find the name of it, but it was published in a peer reviewed journal.

Same idea if you want to send it over to. You didn't name the company, but they, essentially developed, a tasty lemonade, product. And, their objective was to use less sugar than traditional lemonade. So they were able to do, come up with something that had only five grams of sugar per 100ml, which didn't seem to me to be that, low in sugar. I think five grams is about a teaspoon, but, apparently that's quite good in that industry. And, the the interesting thing from my perspective was that they saved, about, roughly one third on their development costs.

So their product development, is we're seeing a one third savings on that from introducing AI. And the reason was because the AI can analyze different ingredient interactions and process processing conditions and, and, and similar consumer taste preferences in order to optimize the, the recipes. so a case study, here's so, three use cases in terms of sustainability. So sustainability is a big buzzword into those systems that are trade show last month where everybody was talking about trade shows, government fakes and all the, of the, food companies that were there, that one was in Australia and they, sustainability was the, definitely the flavor of the day.

so efficient resource, optimization is, is one of the things that you can do, reducing waste. and waste management, sustainable packaging solutions. And I actually saw an example of that while I was there where they had a, biodegradable plastic that was quite impressive, really. Almost.

It looked like real plastic, both in terms of shrink wrap and also things like, you plastic utensils that were, degradable, reduced energy consumption. So if you can track your energy as an input and you can forecast it, you know, just being able to do things as simple as looking at the very between the actual consumption and your theoretical consumption. with an AI assist can help you, conserve energy and greenhouse gas emissions as well. So if you know what your incoming, carbon footprint is or greenhouse gas, profiles of your raw materials, you can forecast what the, your finished products are going to be.

And an example, there is a, a refrigeration company that used AI, and they applied it across 39 locations, about 100 different systems, around to a quarter, ten horsepower. They were able to generate, energy savings of 13 to 27%, roughly. And, basically showed really that they could reduce their maintenance costs and at the same time improve their energy efficiency, especially compared to Hvac systems, which I thought was quite interesting because I, I have also had some experience with Hvac, savings and, that, that kind of, you know, 26 senses is pretty good compared to traditional truck, approaches. and their, their return on investment was, over 40%. So not too bad. So, kind of brings me to the end of the period where we're talking about, the various use cases and the case studies.

And now I want to turn a little bit kind of, you know, to the future and how to implement it. So, first of all, I want to talk a little bit about the negatives. So what's going to happen, in terms of the downsides and what do you need to be careful and watch out for? So I think, you know, the number one thing that people are, you know, I think justifiably concerned about is job displacement. I think as employers, you know, we're going to have to be sensitive to. So what's going to happen to workers? We're going to have to definitely help them shift skills.

And it's I don't think it should be all on the, you know, food manufacturers. I think, government and others are going to definitely have educational systems are going to have to step up to the plate in this regard as well, because there will be quite a few changes. Know we do in many parts of the world, see, demographics where you have a declining workforce. so we may not be replacing as many people as, in other areas.

And I think the skill shifts, it's obviously splits, skill shifts and probably North America are going to be, you know, higher maybe than in other parts of the world. just due to the fact that we have a little bit healthier demographics, both in terms of fertility rates, particularly in Mexico related states, combined with, immigration, whereas in places like Europe, or going to others like that where they have massive, declines in demographics, you're going to see, that's going to be less of an issue. And in places like Japan, they're already using the using, you know, automation to do things even like work in hospitals to flip patients over in beds and things like that. So it's going to be, a lifesaver for those types of, you know, economies that are really aging quite rapidly with very few young people replacing them. And these are all going to cause lots of social, political implications.

And, you know, if it's not managed, it could be social unrest as well that would result from that. So we really need to be careful and sensitive to what's going to happen to workers. But on the bright side, there's a lot of things that come out of this as well. And I think on balance, you know, as long as it's properly managed, it's going to be, you know, more good than bad.

We're going to, first of all, have the creation of entirely new categories of jobs, some of which we probably can't even imagine today. But we're definitely going to reduce a lot of, we're going to eliminate a lot of jobs, but there's going to be, you know, sort of new jobs that are going to be made possible because of that. interestingly, they thought, originally that it was going to be more of the manual labor jobs and not so many creative jobs that are going to be displaced.

But I think, technologies like Dall-E have, sort of possibilities disabused us of that notion. There's going to be a lot more economic efficiencies, which could be more efficient utilization, more abundance for. All right. and we're going to have an opportunity for collaborative roles.

I think a lot of us that are using AI, even at, you know, just the ChatGPT kind of level, at the moment or Microsoft Card, you find that, you know, working with and I actually, is a kind of a satisfying experience because it increases your productivity and everybody likes to be more productive. you can also do more work. And hopefully at the end of the rainbow here, we'll have more leisure time. Everybody will be able to take some more time off. there's not a lot of things that I agree with Bernie Sanders on. but it is a notion that he recently proposed about a shorter week work, work week might not be such a bad idea, you know, as I really takes hold and we see the benefits flowing from that.

So, in terms of managing the, the transition from I, there's going to be new skills required. So the question there is, you know, do you want to, develop your own in-house or do you want to outsource it? And, some of the skills that are going to be required are going to be like data science and analytics, AI and machine learning, engineering, robotics and automation engineering, operational technologies and security. you know, software development is still going to be required.

system integrations are going to be huge. As I mentioned before, I think we're going to see a lot more integrations, than we've had up to this point. change management and change training.

I think about 60% of the of any technology change is actually not related to the cost of the technology. It's related to the cost of the change management with respect to moving the existing workforce from where they were to where they need to be. And I don't see that changing. It's anything. It may go up from there, but 60% I think is a pretty good benchmark. There. regulatory compliance and ethics.

so there's going to have to be more regulatory compliance. you know, it's going to be probably ethical. constraints put on you, you know, maybe about, laying off workers and that sort of thing, strategic planning and, project development.

So this is going to have to be rolled out carefully and, in any kind of, in a cohesive way, you know, if possible, if you're going to maximize your benefit from it, you got a really good plan it carefully, and there's going to have to be a lot of interdisciplinary communication, maybe between departments that really have, you know, have sort of worked tirelessly in the past. So there's also the possibility, like some may say, okay, look, you know, we're just okay, we don't need any AI. And, you know, you may be right, depending on your particular industry. but you do have to be aware of the loss of market share. So a lot of companies today, in the food industry, have very skinny margins.

The average in North America, margin for a food company is under 3%. so that's that's pretty low. and as long as everybody else is also, operating on a level playing field, that's not so bad. But if you know, your competitors end up becoming, much more efficient than you, it's pretty, pretty short step from that 3% margin to going underwater. so you have to definitely, you know, consider whether your industry and your competitors are, you know, kind of overtaking you in that regard or not, because you don't want to lose the market share and you don't want to lose your margins. Right? so there's different risks for different sectors.

So the main ones that I think are at risk are going to be processed food manufacturers, agriculture and raw material production, meat and alternative proteins, and fast food, there's some remarkable, work on the fast food side. I think McDonald's had a, an, a pilot project where they had no humans inside the facility at all in a demonstration. store that they set up. So, you know, not only were their stokers, entirely automated, but they had some big advantages in terms of food safety and product quality as well, because, you know, human hands touch anything. It's a no risk of that type of contamination.

And, you know, your onions get sliced, you know, 12 milliseconds before it goes up the Drive-Thru window. so, you know, you could you could have very high levels of freshness. some sectors are going to be fairly, fairly, insulated from the effects of AI, artisan and Kraft Foods. I would say things like, you know, breweries and distilleries, there's different types.

So if they're, you know, mass market stuff, that's true. But, you know, the, the highly sought after, you know, artisan, beers and that type of thing, cheeses, you know, these Kraft Foods are going to still happen. very high end consumer products are probably going to, you know, maintain their cachet and, are not going to suffer too much. The final one that's probably, sort of insulated is like, you know, think about, farmers markets. So if you're making, you know, artisan products and selling at Farmers Market, I think you're going to be okay.

Anybody that's in the masks market, space is really going to be the suffer a bit. So what are your next steps? So if you take only one thing away from my talk today is start collecting your data digitally now. And think about it this way. Every time you write a letter on a piece of paper, that is an opportunity lost. So it doesn't matter how you really collect it.

You know, if you if you use a sophisticated system like ours or if you even just use Excel, you know, for the time being or whatever, you know, start collecting all your data because you're going to need to data mine that data and use that to train your AI going forward. So start collecting data, understand the challenges that you're going to be faced in. So, you know, kind of need to give it some education or sensors like this one, clear objectives. So, you know, get an idea about kind of where you want to start, what the low hanging fruit might be, what the opportunities are, and maybe get an assessment of your facility and and you processes, and then set objectives based on those. I started sort of telling you before that there's different skills required.

So, you know, if you're a big company and you have a 2 billion R&D, budget, you know, you may want to do it all in-house if you're, if not, you might want to partner with somebody else who already has AI expertise. and start thinking about how to manage your workforce forward in this regard so they can get the culture ready, in advance, if possible, for AI in terms of what we're looking for. Ultimately, fully autonomous food production is, something that I think is possible someday. Seems a bit far fetched, but let me point out that 25 years ago, I saw a construction project, a manufacturing facility where there were only three people working on a shift, no one shoveling raw material in one end, one with a forklift, taking a finished product at the other end, and one guy up in the control room controlling, you know, all the processes. So, products, there's a lot more complexity involved.

The technical obstacles are still substantial. We're still a long ways out. But, I think ultimately we may get there. So, you know, I'm just inviting everybody who may want to collaborate with us, to, you know, reach out if you'd like to work with us in the future.

We you can, you can kind of get into the ground floor and help us. we usually collaborate with our clients to sort of advance our technology in collaboration with our customers, which those that's, become available to the rest of the community. So it's kind of a nice community sort of environment in that regard. you can, you know, see your ROI, in one year. our solution is specifically designed for food manufacturers.

it's extremely complete. Almost every aspect of food production is, run through our system. And, you know, you're going to be able to digitize everything, proc everything and everything is unified in really one central database, one source source of the truth.

And that's, I think, why we've been successful in implementing AI, because we have all of this data and it's all connected. So for us to, you know, to use AI as we've done so far already is a very short step because we don't have to build all the other underlying infrastructure to gather the data, assemble the data, user interfaces and all of that. It's already there. So very easy for us to, to move forward on that. Anyway, how are we doing for time? Amanda? We have about five minutes, so we have time for a couple of questions.

If that works for you. Perfect. Let's do that. Wonderful. So the first one would be, for a food and beverage company just getting started, getting started with AI. Where would the best place to start be? I would say, start with an assessment.

So, you know, have somebody knowledgeable in the, in the art come and have a look at the facility and, give you some pointers in terms of where you could see the biggest bang for your buck in terms of, the application of AI. the other thing that I would do is I would definitely, as I mentioned before, start collecting your digital data yesterday as possible. Absolutely. and next one we have is in what areas can I contribute the most to the company's bottom line? I think, definitely on the quality control side and production. So those are going to be the two major ones, the production side, because you can just you can reduce so much labor, you know, at the same time improving your quality and your your throughput and in many cases, even your velocity.

so that's a low hanging fruit on the quality control side, you know, again, you're you're improving your quality, but you're also mitigating the risk of an outbreak of foodborne illness. And I know a lot of people don't really worry or to, don't believe that that's going to happen to them. But we've seen that happen to other companies. We have, I know I was brought in after the fact for a company where they killed people. You know, that's an extinction level event for the company.

So, definitely don't let it get there. And having this, the ten of the, you know, very, high level of quality assurance, is helpful, for sure. Wonderful.

looks like we have time for another one. How can I be used for inventory management? Inventory management itself, like tracking your your inventory transactions are, is more of a ledger kind of, software problem than a problem. But, other aspects like the procurement side, like getting the inventory in the first place.

definitely. I can play a lot there determining how to optimize this storage ability. I would like, for example, we have a client who we we refer face for where they were able to reduce the picking, time that was required by six seconds per pick, because we could provide optimal routine for each order through the warehouse. And, that saves them $85,000 a year, just about six seconds per pick. So let's, you know, routine, and, and optimization in general. Wonderful.

And, it looks like that's all the time we have for today. So thank you, Steven, for your wonderful presentation. And please keep an eye on your inbox for a copy of the recording. And everyone have a wonderful rest of your day.

2024-04-21 04:02

Show Video

Other news