IFS Digitalization CollABorative: Think Tank Session – Demystifying AI
Date of Meeting: 21 September 2023 10:00 AM US Eastern Time
Insights from Sarah:
Slide: 5 Things CIOs Need to Know about AI (Gartner)
- I found this article recently and thought that this would be a good starting point for us to take the conversation from. So it was five things, CIO’s need to know about AI. Now I know we're not all CIOs, right, but looking at this from the perspective of this digitalization collaborative, I thought it might be a good place to start. So, I think one of the things is, a lot of companies are already using AI in some way or some ways and it hasn't been as buzzworthy as it is today. And then of course, we have folks that are now really looking into, OK, what does all of this actually mean to our business and where do we need to invest more or evolve what we're doing, etcetera. And so this article is talking about five things that CIO should know.
Slide: 5 Things CIOs Need to Know about AI (Gartner)
No. 1: Prior to ChatGPT, most current deployments of AI focused on efficiency improvement scenarios at the business unit level, rather than enterprise wide.
- So, this would kind of intersect into some of the conversations we have had in this group previously about taking a more enterprise-wide approach to our technology decisions, eliminating some of the silos both in strategy, in function, and in technology. So, these are some of the things that historically companies have been focused on, which I think makes a lot of sense. One of the things that came up at the conference I was at most recently is, we want to be careful that just because something's a buzzword doesn't mean that we just jump right on it and do anything without thinking about how it actually factors into our objectives and our business. So there might be applications of AI that aren't, as sexy or exciting as the things that are making headlines but still have a really big potential.
Slide: 5 Things CIOs Need to Know about AI (Gartner)
No. 2: Best use case for AI require access to a sufficient amount of reliable data that is relevant, logical and high quality.
- So based on the conversations I've been having over the last six months, this is where a lot of people seem to be today. They know that they want to do more with AI, but they also know that they don't have the data itself, or the data infrastructure or the ability to make the data logical and usable in the way that they need to, to be able to do ultimately what they want to do with AI. So, this talks about what some of those considerations are. And this seems to be a spot that a lot of folks are in. So it'll be interesting to see if you guys agree.
Slide: 5 Things CIOs Need to Know about AI (Gartner)
No. 3: The growing usage of generative AI opens organizations up to legal issues over copyrighted or protected content and confidential information security breaches.
- So I listened in to a session at one of these conferences where the presenter was talking specifically about ChatGPT, and essentially talking about the pros and cons. Some of the capabilities of it that are certainly appealing, but also some of these risks and some of the onus the business has to make sure that if it's being used, it's being used appropriately. What those scenarios are, etcetera.
Slide: 5 Things CIOs Need to Know about AI (Gartner)
No. 4: You do not need to build your AI. You can acquire AI and use embedded AI in existing applications.
- I think, by and large, I've been in this space for 15-16 years. I think the days of build it yourself are largely behind us. For most organizations is not their core competency, doesn't make a lot of sense, which is why teams are looking at how to embed intelligence into technologies that exist so that you can add that functionality into things that you're already leveraging in your business.
Slide: 5 Things CIOs Need to Know about AI (Gartner)
No. 5: Responsible AI adoption and journey planning means engaging with stakeholders to understand current efforts and defining what AI means for your corporate values.
- So I think this is kind of what I was saying is, not only eliminating the silos, but making sure that, as we continue to go down this path of AI, are we doing it because we feel we have to because it's, it's what everyone's talking about or we doing it because it's cool and shiny or are we doing it because we've identified real business applications that will add value to internal operations to customer experience etcetera. So doing that in mind, and I linked this article here if anyone wants to go back and take a look at it.
Insights from Bob:
Slide: Artificial Intelligence (AI) Sub field Venn Diagram
- I think one of the big changes we've seen over the past few years is companies started to think about moving away from the idea of just doing some AI in silos, as you say, but it wasn't so much that it was just siloed, it is the fact that it was really a proof of concept and it was small teams trying things out and in some cases with success, coming up with measurable KPI improvements on projects, but actually being able to get those projects into production was really beyond most companies and there's several reasons for that. One was having the ability to take those models and host them, deploy them, manage them. They didn't have the sort of operational capability around AI models, machine learning models to be able to do that. And the other thing is I think you mentioned is that I wasn't really tied into the strategy. There was a big focus on leading with the tooling, leading with the technology and trying to find use cases to apply that technology to. Over the last couple of years, I think there's been a shift to more top-down thinking where you're starting with the strategy and finding use cases that are relevant for your business and finding errors that you can add value. And then thinking about what technology can actually help you with that. And I think that's a big shift. I've certainly seen even during my time here, and that's helping companies get a lot closer to actually get up and running with AI.
The data point is another interesting one, certainly with the asset side of things, companies have done a very good job of collecting huge amounts of data over the past few years and the difficulty again is getting going there really and you mentioned the fact that that AI resources are not necessarily a concern for companies, but actually marrying those concepts of data and AI resource and business and domain knowledge and being able to use that data has proven very difficult. And especially in something like the asset space where you might need lots and lots of different models of different assets and different situations, it's very hard to maintain, very hard to find the competency to build all those things. And again, over the last couple of years there's been more of a shift to thinking of a data first approach. Given companies have got all this data, how can we help them get up and running quickly without the need for data scientists and all this complex knowledge. So in fields like anomaly detection and also in building, sort of a lot of the more standard machine learning type models that are learning from data, adapting to different situations, those things have become more commoditized, more automated. That ties in with what you said about not needing to build your own AI, but it's meant that users who previously wouldn't have been able to interact with the technology directly and now getting the opportunity to do so. And that's helping to bridge some of those gaps between the tech and the domain knowledge and mean the companies are able to move much faster.
And just one final point really in what you were saying around ChatGPT and the large language models, I think why that's caused such hype this year and has been such a buzz, is that it's much easier to start playing with that and seeing an outcome very quickly. I think typical machine learning, it requires a level of understanding of the data, the types of outputs you're going to get. What does that mean for your business, etcetera? When you're dealing with a language model, you can ask it questions and start getting responses back very quickly, that can instantly be pertinent to your business and start helping you out with your day-to-day tasks, summarizing things, being able to generate content, etcetera. So that that immediacy I think is what's caused the current level of hype and we're not going to go back from that although, all the potential issues you have with ChatGPT and the other language models, around the data privacy, around the inaccuracy, are all things we're at the start of the journey off and going to need to be thought about and conquered. And you see that with the type of legislation that's coming through from the EU particularly, they're focusing less on the technology itself more on the consequences of that technology for different use cases and it's signing levels of risk according to whether there is a high human impact factor with the decisions they're being made and what happens if one of those decisions goes wrong. If it's like maintaining an asset or something like a plane, the consequences would be enormous. So that risk based thinking is the way that a lot of these frameworks are building up and that will hopefully help us to think about the consequences of these language models not always returning the right answers as we expect them.
That was a few comments based on what you said the slide here is really just to show you that you know there are a number of fields within AI. You probably heard a lot of these terms, some of which are used interchangeably. AI is really the all-encompassing field that includes machine learning, it includes areas such as evolutionary computing, which is taking ideas from nature, from biology, but that one at the bottom is maybe one you've heard deep learning. That's how a lot of these very complex models work, but essentially, they're all just learning patterns from data. They're adjusting themselves to be able to recognize those patterns and be able to make the same decisions. So, these are all broadly interchangeable in terms of their techniques, what one is just a subset of the other, but there's nothing new here, particularly, it's more the fact that the language models have really appealed to people because of their immediacy. There's the ability to process a lot more data now because of conversational power and companies that are now in a position to have that data and start using it.
R: Thinking about it from my perspective and I am not a technical person but something like ChatGPT has taken these concepts which have existed and folks have been working on but just put it in a very usable, accessible way for the lay person. So it makes sense how that could really spike interest because it creates light bulbs for people that have maybe been at a distance because it seemed too complex. It's like, oh, wait, OK, this is great, we can do things with this, what are we doing with this? Let's do more with this, right?
Slide: ML Algorithms
- I think I've covered most of it, but in a sense, certainly the level that I try and get people in our organization to think out in the different industry teams when they're coming up with use cases and trying to implement things is, you don't need to worry about the technology right? Those pieces in the middle, whether it's supervised learning, unsupervised learning, reinforcement learning, they're interesting things to know, but the crucial stuff other pieces around the outside, what’s the functionality that these things can give you and how can you apply that functionality to your to your business? So, what we're doing at IFS, the way we're trying to set it up is as you say, we're embedding the AI within our core products, but we're trying to do it in a way so that even our own developers don't need to see what's under the hood. They just understand that there's something that will allow them to do forecasting so they can take their data, apply that forecasting functionality and it's going to work for them. It's going to allow them to solve a problem, and really that's what we're trying to break it down into a set of skills, if you like, that you can apply and turn to different problems. And I think that's the way we see the general market moving with how they're applying AI.
Slide: Why now? What’s behind AI explosive growth in the past couple of years?
- I think I have already covered this, but its really just the rise of that compute power, coupled with data and the touch play, the ability for people to actually start playing around with it. And as you say, having those light bulb moments, I think that's why we're seeing the rise in 2023.
Customer Experiences (Questions / Answers / Feedback / Responses):
- Q: If we think about planning and scheduling optimization in the IFS platform, so that would be the sort of predictions and process optimization, right? It would kind of fall into that regression area of supervised learning.
A: Planning and scheduling, it probably covers a few categories. I guess it falls into the sort of supervised learning in that there's elements of data that you can use to make better decisions, like understanding how long a journey's going to take based on the time of day. Understanding how long an activity would take based on historical precedent, but there's some other techniques when it comes to scheduling around optimization and the evolutionary side I mentioned before where you're trying lots of solutions out, seeing which ones survive and thrive, seeing which ones die away a little bit, so it's slightly different techniques, but again, the idea is that at the end of the day, you're producing a schedule that's what's happening. The piece under the hood is only important in so far as you need to trust it and understand why it's coming out with the outcomes it is.
- Q: We are on our journey to IFS Cloud, and I would say I get questions weekly. How will we be able to utilize AI in IFS Cloud? And of course, a when I got those questions, I think there are as many definitions of AI as there are persons out there as you just highlighted as well. What should I answer? I mean, I know we are going to invest in IFS Cloud and I know that you are investing quite a lot in AI, so it shouldn't be necessary at least within the sort of the supply chain area maybe even the sales prediction area, CRM area, shouldn't be necessary in order for us to invest isolated into AI, what will I get out of the box from IFS in the next two years?
A: So, the general answer is that all the AI that we're putting into the platform comes through to the users in the form of industry specific use cases and functionality, right? So, the benefits you'd be getting would just be part of the standard way that you're interacting with the product. It will allow you to do processes faster, or allow you to automate it, allow you to automate certain things, but allow you to make better predictions, but all those benefits will be tied into pieces of functionality within the product that you won't necessarily know are using AI under the hood. So, we're not trying to sell an AI platform right alongside IFS cloud, it's all that AI functionality is embedded within there, and as we develop the software and as our different industry groups develop the solutions and the functionality for their particular area, they are able to take advantage of the AI capabilities in the platform to build that into the solutions that they are that they are giving you. So, 1, it's embedded 2, its always business use case driven and 3, it's really across the whole scape of the platform you'll start to see examples coming through. I mean we can talk about specific examples if you like, but the way we're trying to set it up is that can iterate much, much faster. I mentioned this idea of having the POC's and you know having to build a model and then think about how you can go live with that. The advantage of having it embedded in there is much, much quicker for the teams to pick up the tooling, use it, build it into what they're designing generally, and then roll something out as part of our standard development process. They don't need to worry about AI being a separate piece or something that that's going to be so slower to build.
R: That is part in line with what they haven't answered as well, but it would be great to actually have some specific examples in which areas, what is your focus areas? Are you all on the broad scale on the whole application? Are there specific areas you're focusing on, we don't need to do that here, but it could be great to have that input.
Slide: IFS.ai conceptual architecture (IFS Cloud)
- A: So, I pulled this slide up as Sarah says, the concept of IFS.AI is kind of positioning that we are working on it at the moment and we're going to go be going external with it in the next month or so. But, I'm happy to give you a flavor here and I understand this slide is very busy, but I can just use it as a framing device to answer your question or at least to hopefully give you some more context. The bits in the bottom, represent what I was saying, you have a number of services that are embedded within the products that are taking advantage of different datasets, be that telemetry on how you're using the product, be that unstructured data, your manuals, technician notes, documentation, pictures or just the standard stuff that you'd have in the IFS database as an ERP product. But the actual uses of it by the different industry groups are really across ERP, EAM, FSM, and these purple areas represent the specific pieces of functionality that either are in the products or on the road maps of the different groups, or that they're thinking about how to bring in. So, in the ERP we have areas around opportunity conversion, lead conversions in CRM, we have various pieces of demand forecasting within manufacturing, they're based on sort of short term forecasts around weather, there's a lot of demand forecasting. We have our optimization products on the service side, which is doing optimization and scheduling using AI and that's something that we've rolled out. It's typically just been for field service, we now rolled that out for manufacturing. We're rolling out in the asset space as well to do things like optimizing production planning. So that's another area. And and another big area is we've recently brought in some tech that focus very much on the anomaly detection space. So, this is the idea that you are streaming DATA maybe from an asset or you've got data that's coming in in real time. And normally that data is fine, it shows a certain pattern, but when that pattern starts to deviate in some way, it's an anomaly and there's some reason for that. In the case of an asset, it might be that there's some fault or some potential fault that's pending. In the case of financial transactions, it may be there's some anomaly, something fraudulent, etcetera. So that concept of anomaly detection really can apply across a number of areas as well, and we're using it initially in the asset space around that predictive maintenance angle, but certainly in finance, in monitoring our own clouds capability and how well services are running, we're again taking that anomaly detection tech and we're building out across a number of different areas.
So that's a bit of a flavor. I think that the other things are, the light blue area at the top, this kind of represents the interaction experience where again AI is very important. I mean, it was mentioned in one of Sarah's slides, the idea of the personalized experience. So, as a user of the platform, we'll be using AI within the products to reflect the way that you're interacting with IFS based on historical data and the types of things you as a user will be doing. And the other concept to talk about is these co-pilots on the right-hand side. These represent assistance if you like, that help you with the way that you're working that can provide recommendations around processes, whether they're optimized, whether they're things you could be doing better, they can provide you ways of doing bits of development in the product, doing certain queries, being able to produce reports and lobbies much more effectively. So, there is a range of co-pilots that that help you and help the people within your organization with their different roles to be able to go about their day-to-day business. So specific pieces of functionality co-pilots represent, making it easier to use the product overall and the user experiences about making that more personalized and make it more attuned to you. I hope that gives you a bit of an idea. This exact branding and positioning are something that we are going to be pushing it certainly in the next month or so.
Q: Is there anything in this content or anything Bob said or didn't say that you would push back on and say would be good to include or give some more specifics on? If you have any constructive feedback, it would be good to hear it as well.
A: So, we are high volume producer, meaning that and we have a lot of data. We have IFS in all our 40 factories, so we have a unique position to gather data and to standardized data, and I would like to utilize that position to streamline and improve the operation and utilize AI to actually be more efficient and get the speed through the system. That is what I'm looking for and I see that there are areas here within planning, within forecasting, within prediction, so on the different areas, but I would really like to see if it is possible to get some business case where we actually can highlight and showcase utilization of AI. I appreciate that it's kind of built in under the bonnet, when you buy a car, you don't bother about what's in under bonnet, you just wanted it to work. In this case, I'm happy that it's under the bonnet, but I would actually like IFS to showcase it as well because AI is a big buzzword and it would be much easier to say that yes, we actually are using AI and IFS is supporting in that track.
- F: AI is a funny conversation these days, right? It's both used and abused. Pretty aggressive right now, in that buzzwords tend to do that, but I'm really on board with having some practical samples. As a technology leader, I face this all day long, right? What technology do we need? Is IFS really the right technology for us still? Because NICE is now very good at advertising AI. So, these conversations are happening very frequently and it's going to make a differentiator in the market too. If I look at the difference between boomi and power automate, andmake.com make.com is doing a fantastic job in explaining how they are using AI. They're showing use cases and what they do with the AI ChatGPT integrations and how that leverages as an automation platform, how that helps people. From an IFS point of view and just to throw it out there, you show how any incoming email to a Microsoft or any email server is being parsed through an AI engine and instead of it requiring the customer to submit things in a certain format to you for proper parsing, you leverage that AI to say, hey, this person is asking for an update on the ticket, it can actually through the conversational AI, the NLP (Natural Language Processing) engine, really get the right information back to them immediately, or create a note in the right place in the IFS system for the right follow up from your team. That, as a practical example, if you show that off, you're going to reassure your position from a client facing delivery platform.
- F: Self-service right, being a B2B organization which metals very much into B to B to B and B to B to C, sounds funny, but it's truly how we do business. It's self-service capabilities is really the key. Everybody is now looking for more self-service and as we are self-servicing ourselves through AI, in the professional world, by looking at jobs or job descriptions, clean this email up for me, we are using AI, at least I am, and a lot of my team members and people in my organization are to make our job easier, faster and get better results, get better communications out to the business, get more meaningful keynotes etcetera, etcetera. That self service capability is something that I would like to offer to my customers as well when they submit a ticket when we have a chat bot session, when they send us a note, when they call us, when they log in into their customer portal and our identifying issues truly making the journey of service easier to our clients. And that can take many shapes. Like I said, in the NLP option earlier, but also if we have a chat engine that really takes a look at the person, the issues, what's installed at the store and not just this, one size fits all kind of thing which is resolved by implementation of AI because it really takes all of the data and parameters available within the variety of systems and data sources that we have and kind of makes it tailored solution and experience to each of the clients, which is nice.
- F: We're manufacturing company and we're definitely interested in some of those scheduling capacity planning things that are on the horizon. I know he talked about MSO, that's something we're definitely very interested in, and we have strategic initiatives currently ongoing and all of our companies trying to get better at scheduling and planning out our future, right. And we really struggle with that today. We use some demand planner. We use some of the tools that are available, but at the end of the day are our processes unique enough for those things don't always fit, so I think where AI can help us get smarter in the future, find trends, finds ways for us to maximize our schedule and make our plans run as efficient as possible, is crucial for us. And when we look at especially competition and what goes on in the future with them, we have to be best in class or we're going to be behind and that's one of the reasons we picked IFS and we're hopeful that that can help us put us in a good place in the future as far as competitive competition and efficiencies and cost savings and all the things that that AI can do for us.
R: So, basically your data that you input is super important, right? So without the data being input into AI, it's not going to work. It's not just about what IFS can do for you, but also kind of what data you have to make that information come out correctly at the end of the day.
R: That's a really good point, and I think it's something that's important to understand whether you're thinking about your partnership with IFS or your partnership with any technology, is it is a partnership. This is what happens when something becomes buzzworthy. There can be this thought that you can just sort of wave your magic wand and put it to work, but in reality, this is a technology that really does depend on the data to do what it's supposed to do, and that's where sometimes people have really extensive objectives that ultimately are attainable, but they're maybe work that needs to be done on the customer's end as well to get the data where it needs to be to make that happen. So, I think IFS is doing everything it can to simplify and streamline that process, but at the end of the day, there isn't any technology that can do literally all of the work for you, there's still some requirements. So that’s a good reminder.
If you are an IFS Customer and would like to watch the recording, please email jessica.foon@ifs.com
A copy of the slides can be found in the attachments section below.
Next Meeting: 18 October 2023 10:00 AM US Eastern Time
IFS Digitalization CollABorative: Tech Talk Session with Andrew Lichey, VP, Platform Product Management at IFS
If you are an IFS Customer and you would like to join the CollABoratives, please click here to fill out the form.