Nobody needs to fix their data before using AI

Good to be on too. Thanks very much for inviting me.

Yeah, so the first 20 odd years of my career were spent in a variety of operations and maintenance management roles, predominantly, as you say, within the brewing industry and predominantly within Heineken UK. And during that time, I got to be a site general manager of one of our breweries up in the northeast of England, making Newcastle Brown Ale. And I also got to be the engineering lead for our UK sites. And in that I got to run some very large and very exciting capital programs as well as managing our maintenance function and our energy management. The next 10 years of my career, I then moved into service provision. And predominantly that was within an industrial distributor where I was leading a corporate startup. around industrial Internet of Things and industrial data acquisition. So starting to build out the technology platforms that I'd seen the need for when I was on the customer side for a variety of reasons, sadly that corporate startup wasn't successful. And so I left the business that was with just over two years ago now and decided that after 30 plus years in corporate roles, You know, it was the time to set up on my own. So I set up two, six services to look at, you know, industrial asset management with an AI twist. And just had our second birthday, was joined by my business partner at the beginning of this year. So there's two of us now. And yeah, that's where, that's where I am after two years.

So as I said, we focus on predominantly on industrial asset management. So we work with manufacturing businesses to look at their capital investments. So their acquisition of new assets and bringing those into service and their asset care of existing assets. So how do they operate and maintain existing assets to ensure that they get the maximum life at the best possible life cycle cost. out of those. And, you know, we believe very much in the importance of doing the basics right. But we also believe that there is a huge role that technology can play as an enabler in that space. And indeed, today, I'm in one of our partner offices today, because we work very closely with an AI startup looking at building technology solutions for industrial asset management.

Yeah, I mean, think, you know, obviously, businesses know they have a challenge around either that capex and bringing assets into service or all that maintenance approach when we come in. And we always start with an audit of the customer. So we go and do a structured assessment of where we see the opportunities for improvement are. And, you know, I think we very much focus on some of the human related failures. I mean, we understand machine failures, both myself and my business partner are reliability engineers. So we understand the mechanics and the theory behind equipment failures. But actually, it's normally the organization that allows things to go, you know, to happen that shouldn't happen. Yeah, I was in a large pet care business, a global pet care brand on Tuesday. And we were looking at repeated failures on a machine. And we were talking about just the importance of doing the basics well. And I think they, yeah, the surprise is often how simple the fixes are, but how hard it is to embed that into the culture.

Again, it refers to... I mean, in the world of industrial maintenance, there's a process called precision maintenance, which is about making sure that you set yourself up for success. So it's about making sure that you are looking after your assets and your components in the best possible way so that you don't allow failures to occur in those components. You focus on stopping the failures that occur in the same place. If you think about cars, we all know that if you keep the oil topped up, if you keep the water pressure, if you keep the tire pressure right, if you run the car within the right revs and you don't cause excess wear to the engine, we know that extends the life of your car and reduces your fuel consumption. The same mindset applies in industrial assets. There is a way to... a way to care for your assets, a way to run those assets which maximises the life and minimises the cost to operate.

Yeah, I so I'm currently researching a PhD in the use of AI in industrial asset management. So, you know, it's a topic I'm clearly very passionate about and I can see huge opportunities for. And I think AI can create value in a number of layers of value. So first of all, you you create value in the tactical layer, the how do I execute my tasks on a day to day basis. And in insight to six, use a lot of AI tooling. We use, you know, call recording and meeting recording. use, chat GPT to help us analyze data to help us structure data. we use a bunch of web scraping tools to allow us to understand as much as possible out of our, our customers. Now, in that case, we're being more efficient versions of ourselves, but we're not really transforming things. We're just operating our business more effectively. You then can think about using AI in the operational layer. So how do we improve the performance of a function or a process? So the piece of work I was doing on Tuesday in a pet care business, we'd been looking at repeated failures on on one of their pet treat lines. And we're using AI to help us to enhance and improve the maintenance program for that entire line to look at how they cared for the asset and how they managed their spare part flow through that. And also using AI to look at how we could design out some of the issues that saw. So, know, operationally, know, huge benefit to not just one or two individuals, but also the whole of that line. And then you can also use AI strategically. So yesterday I was in a large greenfield factory, very, very major investment. And they're looking at the handover from capital acquisition of assets into the operations and maintenance of assets. And we're looking at using AI to structure the data to hand over from one team in the acquisition and CAPEX phase. into two different teams, which is the operational team and the maintenance team, and making sure that that data handover is done effectively, but also the data is presented in a way that can be consumed by those two new functions of operations and maintenance, rather than just having to consume it in the same way that the capex team presented it to them. So three levels of three levels of value, the tactical layer, creating value for yourself, the operational layer. creating value within a process stream and the strategic layer creating value across the entire life cycle of an installation.

Yeah, I mean, look, you know, everybody's everybody's using an LLM at home in some way, shape or form. Everybody's using an LLM at work or the majority of people are using an LLM at work in some form or another. Some of those work based LLMs are formal and inside the control of the organization. Some of them are informal and possibly outside the control of organizations. So people are using LLMs, but Yeah, and LM is, know, chat GPT or Claude or whatever. These are incredibly powerful tools. And as a result of which they're also incredibly complex and people are lulled into a, know, because of the simplicity of the chat bot interface, you know, people are lulled into this, into the thinking this is a simple tool to use, in order to really get the value out of it, you know, you need to learn how to work with these tools to best effort. You know, I, as I say, I'm, I'm in my AI partner today. We all use a lot of these tools and we, we constantly share tips on how we're extracting maximum value. And there's a bit of a competition as to who's getting the most, who's found the next new feature and embedded it into their workflow. And we make a bit of a sport out of it, but there's a serious point behind it because we're really stretching ourselves to learn how to use these tools. And as a result of which we're getting maximum value out of it. You know, just because you get the use of Excel doesn't mean you can do a pivot table. Just because you get chat GPT doesn't mean you can extract all the possible value that comes out of it. I think once you move beyond that, that tactical layer in asset management into the operational and strategic layer, know, businesses are very much at the start of that journey. I'm working with a number of organizations who are looking to put their toes in the water and looking to see the value. industrial asset management is a complex space and therefore it is more complex to apply AI in that space than maybe some of the other functions that exist inside of business.

I think it very much depends where in that technology, where in that value stack you're playing in the tactical, the operational, the strategic layer. But I think my biggest recommendation to anybody was really stretch yourself to understand what the tool can deliver. One of the things I do, I'm a ChatGPT user. One of the things I do about once a month is I get ChatGPT to analyze my usage of ChatGPT. and tell me how I can be better, how I can create more value from that. And that surfaces new functionality that is potentially available, but also teaches me the skills about how to get value from it. And so I will regularly audit, get it to run a self audit, get it to run an audit of my use of it, get it to give me a little scoring matrix about how my, and I can start to see, am I really extracting maximum value? So, invest the time in really pushing and the tools. Once you move into the operational strategic layer, you need to be solving material problems. So you don't just sprinkle AI fairy dust on a factory and think you're going to get any benefit from it. You need to be looking at your material problems that exist and really thinking about what role can AI play in this. And of course, there's different types of AI. Everybody thinks now that an LLM is AI and they forget about a lot of the other AI tools like machine learning, like anomaly detection, like forecasting and prediction. Tools that have been around for decades and are also incredibly powerful. So you need to make sure that AI is not one homogeneous set of technologies. is a range of technologies. And you can't just throw one technology in an LM at a problem. You've got to think about the different problems, different technologies.

you Yeah, yeah. Yeah. Yeah, just yeah, ask the prompt and, know, get it to tell you how to write better prompts, better get it to tell you how to get better connectivity, get it to tell you how to organize your, you know, your other data sources so that the LLM can make best use out of those data sources.

Yeah, you've clearly got to understand, you know, understand your different types of data that you're looking at. I mean, as a maintenance engineer, I talk a lot about condition monitoring and measuring leading indicators of failure. You know, these are fundamentally time series data sets. So if you're looking at a time series data set, you need to understand, you know, are you measuring the right indicator? So are you measuring the right thing? Are you measuring it at the right granularity and are you measuring it at the right frequency? And are you doing any post analysis on that data to prepare it properly? You know, these are relatively easy things to understand if you have the domain understanding of the problem you're trying to solve. But then you've got to think about how do I manage gaps in data? What do I do if the data stops? What do do if I get a backlog of data? So that's all relevant for time series data. If you're looking at transactional data and a lot of what you do in a business is transactional data in your ERP, again, you've got a different data quality challenge there. So I do a lot with engineering notes in people's maintenance management platforms. And you just need to accept the fact that the data quality is as good as the engineer typed in at two o'clock in the morning at the end of a breakdown. So the quality of that tends to be fairly poor. So then you need to think about how can I educate people of the value that they're going to see for this data? in order to encourage them to put more in. How do I lower the burden for them to input good data? Can I move onto the shop floor? Can I make it voice rather than text? Can I get some, can I load photos rather than wording? How do you make the data acquisition burden lower? If you're looking at things like this handover from capex to opex that I mentioned earlier, that's about understanding all the possible data sets that exist in this particular factory. And how do I meaningfully merge those data sets? So I go from disjointed data to joined up data. But of course you can also use AI to address data quality. So if you know you've got late quality data, you can use AI to start, know, web scraping, to go searching, to go and look at, you know, triangulating that data with other possible data sets. And enrich that enrich that data source. So there's lots and lots of tools to use to, to lead into that. And you don't need to, you know, for me, it's a fallacy to say, we've got to fix our data before we can use AI. You've got to, you know, solve different problems with different tools and see what, what data I've got that's good enough already. What data have I got that can be enhanced through AI and what data do I need to put that effort into enriching before I can throw it at an AI tool.

Yeah, I mean, I'm towards the end of my career, whereas you're towards the start of your career. So, you know, different motivational factors. you know, we want to, you myself, my business partner, we want to, we want to solve real engineering problems. We want to, you know, have a lot of fun and meet good people and have a good time. We want to, you know. stay at the cutting edge of technology and really explore how AI can create value and support people into getting that. We're probably less motivated now than we were about going, you know, I need to scale this, I need to bring more people in. You know, we, yeah, we're less fussed about scaling it into and selling it and we're more fussed about having fun on the way and continuing to learn and broaden ourselves on the journey.

So, I mean, we've got a website to six.co.uk. But if one is the best place is on LinkedIn, that's where most of our activity happens. Just reach out to me on LinkedIn and let's connect. And I post some stuff about AI on there, I post some stuff about maintenance on there. yeah, join the conversation and see where it goes.

Okay, well thanks very much for inviting me.