Artificial Intelligence in the Supply Chain

Patrick Daly:

Hello, this is Patrick Daly and welcome to Interlinks. Interlinks is a programme about connections, international business, supply chains and globalisation and the effects these have on our life, our work and our travel in today’s world. In today’s programme, I’m joined again by one of my colleagues from the global supply chain think tank, David Ogilvie, who is principal at David Ogilvie Consulting in Brisbane, Australia. And today we’re going to attempt at least to discuss some of the pragmatic deployment of artificial intelligence in the workplace right now. So maybe just to explain our businesses Dave, so we both in consultancy business, mine in supply chain, and yours in, well, you say, what is your day business involved in?

David Ogilvie:

I have two pillars to my business, Patrick, and that is supply chain resiliency. So I help companies become more productive and build a resilient supply chain. And in doing so, I also work with ERP systems because for me, supply chain and ERP are hand in glove. When I’m making those supply chain improvements more often than not that we are playing with the ERP system. So I help companies in the lifecycle of ERP, whether that be a selection, implementation or what I call restoration. So I restore it back to the way it could have been or should have been in the first place.

Patrick Daly:

Excellent. So that’s the overlap. So we overlap in that we are both interested in different parts of the supply chain and you are in the technology space there going into ERP, and I’m a lot into the operations area, logistics operations, warehousing operations, automation, productivity improvement and so on. So I guess this world of artificial intelligence is something that has attracted our attention. We were just talking beforehand that maybe in some aspects the jury is out on this suite of technologies and maybe some of them will be more useful than others and so on. And I think it’s probably fair to say, and you can add to this, I think we’re probably interested in this both maybe for applications in our own businesses and then maybe also to provide practical and pragmatic guidance to our clients along the way. Would you agree with that?

David Ogilvie:

Yes, I do. Obviously the first place I’m looking to be honest is a little bit selfish is to see how AI can help me in my own practise. But secondly obviously is how this can be most effectively used in client sites and to help them get that efficiency that we talk about. The big technology companies. And I’ve got to declare a little bit of a bias in some respect maybe or it’s a view. And that is, I come to this with a sceptical point of view. There’s been a lot of talk over the years and I think blockchain’s a really good example. Blockchain was touted as something that was going to, a technology that was going to revolutionise this company’s supply chain. And the reality is it has failed to make a big impact. It has failed to get a foothold, if you like, into practical solutions. And it’s still a lot of promise, and I classify it as something, it’s a solution looking for a problem to solve. And I don’t think it’s managed to do that.

Now, while I don’t think AI is in that bucket, I still come to it with a suspicious viewpoint at this stage.

Patrick Daly:

Yeah, it’s a fair point. I maybe say I can’t remember how many years ago, maybe eight years ago when there was a lot of buzz around blockchain and I organised an event on it here in Dublin, a breakfast forum, and it got huge attendance. So normally in these breakfast forums I might have 10 or 12 people, it was like 50 people turned up to this thing. So people were really interested in it. And there were a few examples around, there was an Irish dairy company that had exported a shipment of cheese and butter to the Seychelles or Mauritius or one of those places. And they had done all the documentation, the transactions on the blockchain, and it was a case study, but really nothing ever happened again after it. So they’ve obviously found other more conventional ways to handle their export documentation.

So yeah, it’s a salutary tale, so we’ll see where this goes. But I guess AI, it’s a suite of different technologies and applications with a wide range of uses, I guess. And today maybe we’ll try to, maybe not so much talk about future speculation or where AI will ultimately get to nor indulgence in some of these doomsday scenarios of it taking over the world and taking over the world from humans and leaving us as drones or with nothing to do, to maybe just explore what’s available now, how it’s been used, what can be done with it in a practical sense to improve your work and your productivity and how you might go about that. So what have you been doing with it? What you seen other people doing with it and what are you exploring at this point?

David Ogilvie:

Well, I must admit, I thought it would be a smart thing to do seeing we’re going to be talking about AI, is to actually use AI for our chat today and to see what it gave us. And that leads me into how I’ve been using it. And that is to give me a broader perspective, I guess, and try and find information from a research perspective that I haven’t been able to draw from my own knowledge or from other, the old googling and so forth. So it’s been a research tool for me to start with. I think one of the things that I can see it can help in my business will be some automation. It should be able to help me automate some things. So as an example where I’m able to record a meeting with clients, for example, where I might be drawing a requirements, having a requirements conversation with a client. If we record that, I can put that into the transcript of that into AI and it can generate meetings summaries and those sorts of things.

So for those organisations that have got Microsoft Teams, for example, when you deploy the Copilot component of teams, that’s the type of functionality it’s going to deliver for them. You can record your internal meetings and it’ll do the note-taking for you that you used to have to have board reports and board meetings and those sorts of things. You used to have to have a stenographer in there to take the meeting notes and to share them around, et cetera, et cetera. It is automating that type of activity. So that’s where I see a fair amount of strength with it. I guess my next thought is, is there a distinction between AI and machine learning? And again, I just researched that while we were talking and it’s come back and said that AI is the broad field of computer science, whereas machine learning is a subset of AI.

And I want to make that distinction because I had a client undertake a machine learning project with one of the big software suppliers. I won’t name specifically which one, but the purpose there was they were going to be using the machine learning capability on their platform to help them with their demand forecasting to try and give them a more accurate forecast. And if this was a pilot in conjunction with this software supplier that went on for a number of months, it was about eight or nine months. Cutting a long story short, they ended up stopping that pilot and went to use some statistical, went back and used statistical forecasting application because they found it more reliable. So I just think yes, there’s a lot of potential sitting here, but I think we are going to, there’s going to be a lot of trial and error and mistakes made and subsequent costs too, I would say, going to be made until we find where that sweet spot is.

I’m no expert in medical science, but I do understand it has revolutionised the detection of cancers and cancer diagnoses and those sorts of things. So you can look at scans, which is obviously just ones and zeros at the end of the day, you can look at scans and make much more accurate diagnoses of those sorts of issues than the doctor’s visual review of the scan. So that clearly in my mind, again, is another example of where there’s going to be huge improvements made.

Patrick Daly:

It’s interesting some of that terminology and to draw those distinctions. So artificial intelligence is a wide field, which is basically using, creating systems that are capable of performing tasks that would typically require human intelligence, whether that’s understanding natural language or recognising patterns and data, making decisions, learning from experience and so on. Within that, then you have machine learning as you said, which is the development of algorithms and statistical models so that computers can perform specific tasks without explicit instructions. So they learn and they improve their own performance base and the exposure to data. And then the interesting thing, so what you were talking about in terms of the examples that you set up was generative AI, which uses machine learning. So GenAI, which is AI that produces new content. So you might say to it, inventive marketing slogan for me about this, that or the other, and it generates something new, so it generates new content.

And then an example of that, so you’ve got GenAI, generative AI, which is doing that say with text like ChatGPT that we mentioned. But then there are other elements of GenAI which are not text-based, they might be video-based or they might be image-based. So there’s a whole raft of stuff there, particularly in the creative fields. So those are some of the nesting technologies that we’ve got there.

David Ogilvie:

Yes, I don’t know whether you’ve tried to use it for images, Patrick, but I have. And one of the results I’ve managed to generate so far are, rudimentary is probably our best way of saying it, that it contains, the images that I’ve managed to generate, contain a lot of spelling errors. So if I’ll ask it to create an image around some parameter, I can’t think of one off the top of my head, but the wording has been spelt incorrectly. Funnily enough, some of it looks like it’s Russian, and so the accuracy of these things at the moment is still missing.

Patrick Daly:

Yeah, yeah. A few examples from my own work. So in terms of research, I was using it for research and what I found it made it very efficient for me to find references and citations. So as I was asking it questions about certain things, I was asking it to give me where that information was coming from. So it was actually giving me the citations, and therefore I could go and look at those citations and read up about them and also mention them in the reports that I was doing. In terms of summarising, as you know, we have a podcast series and I would pretty much once a week interview somebody from somewhere in the world about some topic and we end up with a transcript of 25, 30 minutes of conversation back and forth. So I’ve used it to summarise those transcripts and to, if you create a summary with quotations in for what the guest has been saying, and then I can use that to make a newsletter and then send that out as a summary.

David Ogilvie:

And I think our profession is going to get a lot of benefit out of this once we’ve mastered the way that it can help us. And I think organisations can. That’s really just another use of that Copilot thing that I was talking about before in Teams where you summarise a meeting.

Patrick Daly:

Yeah. And I had a young man on my podcast, hasn’t gone out yet, but it’s recorded. We recorded it earlier last week, actually it will go out next week. And he’s a young guy in his early thirties, I think he’s in California, originally from Nepal, and he’s working on projects at the moment where for healthcare providers in the US normally when you have a doctor and a patient and a consultation, often you’ll have a scribe who’s recording what’s being said, and then the doctor has to take those notes and do his administration stuff like that with it. So the project is they’re using AI to substitute the scribe, and then they’re also using it to take the bureaucratic load off the doctor’s back so that the clinician can keep working as a clinician rather than being dragged into administrative tasks. So that’s quite interesting.

And I was approached last year at this stage by a company in Berlin who are developing AI to help with the specification and analysis of requirements for logistic systems like sortation systems or warehouse automation systems for order picking. So for example, the meetings and the specifications for those are quite complex, so there’s lots of technical meetings between customer and the technicians, and then there’s a lot of data that has to be obtained and it has to be analysed. And what they’re looking to do is to provide an AI agent that can come to those meetings and that can also, after those meetings, take the data and analyse the data and help both the vendors and their customers to get to the optimum solutions quicker. So that’s an interesting one.

David Ogilvie:

Yeah, it’d be interesting to keep an eye on that one because I think that’s one of the areas where AI can help. Now, I’m not a coder, so I haven’t experienced this myself, but my tech colleagues do tell me that AI can solve a lot of technical issues. So they’ll have a piece of code that probably has got a bug in it, for example, they’ll get AI to basically look at it and identify where the bug is and write the code to rectify the bug. So there is a school of thought out there that says that programmers or developers, code writers, that’s one of the jobs that will be somewhat under pressure. Now, whether that actually comes to fruition or not, I have my doubts. However, from that coding perspective, I still do think that the demand forecasting conversation that my client was having with that vendor about trying to… I do see that there’s potential that it can help in those complex areas like routing, logistics routing, transport routing, manufacturing routing.

So there are some very complex optimization problems to be solved in Australia, for example, our coal chain. So while everybody’s trying to close the coal mines down while they’re open, actually getting the coal out of the ground, shipping it down rail lines to ports and scheduling that coal onto that ship on that dock is a very complex arrangement, because you’ve got so many different players involved and they’re all trying to game, and there’s costs involved in having the trains tied up, et cetera, et cetera. That’s a very, very complex supply chain, one that I’ve worked in a bit, but I do think that this could potentially be the one thing that could crack that complexity because no one really has successfully done that.

Patrick Daly:

Yeah, yeah. I guess as well, we all know in manufacturing and logistics transport, and there’s so much waste, there’s so many empty legs on driving, there’s so many vehicles going past places where there’s an opportunity for them to take on a load. So there are lots of opportunities. I guess the thing is having the people who understand how this works and being able to set up the projects to get hold of the knowledge and the data that’s required, make that data available to the system in compatible formats. And then-

David Ogilvie:

That’s an interesting point, Patrick, because I don’t know whether you remember my supply chain model that I developed and that I use on projects, but it fundamentally consists of at the high level for these, and what I say is we look about the visibility of material information and money. And so, one of the things that AI could do is improve that visibility of information. The second V stands for velocity, obviously the speed at which things occur, whether that be the shipment of the velocity of information, material or money. Again, it could potentially increase velocity at which that information about money or material is brought to the system.

And then if you’ve got a computer system running this, well, that obviously is probably as fast as it can be done and certainly faster than the human. So getting back to your point about empty legs and so forth, if an order is placed on a carrier somewhere and once that information hits the database, well then that speed of information can then be brought back in. The AI can consider that, and you could be making last minute changes to a route to make things more profitable or more cost-effective.

Patrick Daly:

Yeah, that’s quite interesting. I guess there is already a skills issue in general. There’s going to be requirements here for skills and just as I was saying there, so in an organisation getting hold of the data that’s required, and often in organisations, a lot of the data to do with decision-making is tacit and it’s buried in people’s heads. And getting that stuff out explicitly is going to be one part of the conundrum. Another then is translating that into digital knowledge and making it AI compatible. And then the other thing then is using platforms because you’re, to make this powerful, it’s got to be able to roam wide and far to do these optimizations. And then every application, I guess is going to be customised in a way where you’re going to have to co-develop generative AI algorithms for your particular applications. So it sounds to me like there’s a whole world of opportunity there for people who can learn about this and who can insert themselves into-

David Ogilvie:

100%.

Patrick Daly:

… that space.

David Ogilvie:

Another obvious one for me is the maintenance area. So for a firm that is asset intensive, I’ve been running my business owner and CEO lunches of late around the country, and the topic at the moment is around the connected factory. So how do we actually connect the machinery on your shop floor back into your ERP system to make it more productive, more efficient, and all those sorts of things. And one of the things that comes out of that is the use of data lakes. So all of these machines you’ve got on your shop floor will be feeding all of this temperature information, number of items produced, all of this information that they have inherently in there. But from a maintenance perspective, we can have engine temperatures, oil temperatures, oil pressures, all of those, that data coming back into this data lake.

Well, somebody’s got to look at that. And at the moment you said that’s all in somebody’s head. There might be somebody wandering around having a look at the temperature gauges. He’s either thinking about the fight he had with his wife at home and not taking any notice of it or all of those things that happen on a daily basis. It’s flippant, but it’s real. And so all of this data going into the data lake, so this is going to change the technology landscape of organisations like manufacturers who wouldn’t really be thinking normally about the use or how do I use a data lake?

Patrick Daly:

Yeah, yeah, that’s interesting.

David Ogilvie:

You apply that AI over that sea of data that you’ve got, and getting back to that medical piece that I mentioned about a better diagnosis, well, if it’s really good at that, it’s diagnosing which of those machines has got a cancer, has got an issue that needs to be resolved.

Patrick Daly:

Yeah, yeah. I’ve seen this in terms of anticipate anticipatory maintenance or just getting, say the settings on machine tools just right to minimise waste and to minimise wear and tear on the machine, prolong the life, improve quality of the parts that are being produced. So yeah, there are huge, huge opportunities in-

David Ogilvie:

Well, it’s interesting you say that because one of the questions I did ask ChatGPT for this morning, as you know, was, and it came back and said, AI can help with enhanced quality control. Now, at a first glance when I first read that, I was sceptical about that, but to your point, if there are some precision improvements that it can be made in the machinery’s tolerances and so forth that could improve the quality.

Patrick Daly:

Exactly, exactly. So in terms of learning more, do you have any plans? Are you aware of any material or courses that are worth looking at in this space?

David Ogilvie:

Not that I’m aware of, Patrick, no. It’s more just trial and error. And at the moment for me, it’s more error than trial, so to speak.

Patrick Daly:

Yeah. Yeah.

David Ogilvie:

What I have heard from a couple of my other colleagues is that the power of it comes from the way you frame the question. So I think there is some learning for us there in how we ask the questions.

Patrick Daly:

That’s true, that’s true. I have become aware that there are some very good free courses and materials on Google. There are two or three starter modules there provided by Google that explain the basics of the technology. And people who are in the know have said to me that the content is quite good. And I’ve also had a recommendation. There’s a module made available by NVIDIA, which is free as well on generative AI, and it’s a two-hour introductory model. It’s free, but it’s good stuff. So I think a lot of people are going to be looking into it. I think those who are really interested in it can maybe get a head start and just seeing what can pragmatically and practically be done to improve productivity, improve quality, improve efficiency. I think there’s a lot here to be looking at.

David Ogilvie:

Well, there’s no question the software companies, and Microsoft specifically, because they put the big investment into ChatGPT. And so they’re leveraging that and getting reward for that, I guess now through their Copilot piece in Microsoft Teams and so forth. But even there’s a lot of webinars from the Microsoft partners out there at the moment that are saying that they’re embedding AI in their ERP systems, and they potentially are, technically, they’re probably embedded. The examples they’ve used to date, for me at least, have been really rudimentary. And I often wonder why they even think about them being of benefit. I’m going, well, okay, well why would you do that when they demonstrate these things? But they are starting to make steps and they are starting to find improvement. So I think it won’t be long before those use cases if you like become more productive.

Patrick Daly:

Yeah. Yeah. Okay. Well, that’s interesting. I guess our verdict is, it’s interesting. There are some good examples where we’ve experienced use, we’ve seen other people experience use, but maybe in the wider sphere, the jury is still a little bit out, and we’re in wait and see mode as to what further developments there are. But as I said to people, if you’re interested in learning more, have a look at that material that’s been made available by Google or by NVIDIA. And then there are also some good courses, some paid courses that you could do at MIT and some of these institutions. So some good online modules and even qualifications that you can look at. So I guess we’d have to leave it there, Dave, the clock is against us as always.

David Ogilvie:

Always is [inaudible 00:28:38].

Patrick Daly:

Thanks again for being here. Been a pleasure to chat to you again as always.

David Ogilvie:

Thanks for having me, Patrick. It’s good to talk to you.

Patrick Daly:

You’re very welcome. And thanks also to our listeners for tuning in again today and be aware that if you enjoyed this episode of Interlinks, you can find the full series, which is getting on now for almost a 150 episodes of Interlinks on Spotify, Apple Podcasts, Acast, and other major podcast platforms. So until next time, keep well and stay safe.