สำเนาบทสนทนา 🔗
Russell: don't we also have to Clarify the fine line between content and context. Is content context? And if not,
Caspar: Ooh. Ooh.
Russell: it is one-to-one the same. Where is the actual distinguishable
Caspar: Hmm.
Russell: difference?
Caspar: Ladies and gentlemen, welcome to another episode of the BPM 360 Podcast. My name is Caspar your host, and with me is the other host, and he's back home from the looks of his curtains. Russell, my friend, good to have you back. How are you?
Russell: I'm fine. Back back at home and I just as I also told you, back in Germany, and we had two days sunshine and then the switch, the autumn switch was flipped. So so
Caspar: Yeah, true.
Russell: it all all of a sudden it seems like the the summer holidays and the summer just seems very distant.
Caspar: That's true. actually was using a broom today to get all the leaves together and put them into big bins.
Russell: Yeah. Yeah, yeah, they're the symptoms, absolutely. But the Germans saying Heißer Herbst, we've got a really hot autumn, well up in front, but not
Caspar: yeah, yeah, yeah.
Russell: so much for the temperatures. It's more about all these events that we're digging into, yeah, and where we'll be
Caspar: true.
Russell: present, all about process management event and
Caspar: yes, looking forward to that one.
Russell: So I will also be on the RS dot com partner days and the
Caspar: Okay, cool.
Russell: event and also at the SAP Zignavio and Lina X summit in in October. So there's
Caspar: Okay, cool.
Russell: a lot of events on and and I think that's really interesting because I think it's always the roadshows, the announcements. It's gonna give us a lot to talk about. not just shorts,
Caspar: absolutely.
Russell: but it's certainly going to be a lot where especially when you see the road shows of the big vendors and the seller sphere is also coming up, isn't it?
Caspar: Yep, yep.
Russell: that this will be I I think there's gonna be some very cool announcements to see where the trends are going. And and I think a lot of the topics we've talked about we will be we will see assured. Maybe even
Caspar: Yeah.
Russell: some new topics popping up that we haven't talked about or anticipated. So I think
Caspar: Okay.
Russell: even from the aspect of our podcast and the topics, I I think the autumn's gonna be an awesome. It's gonna be it's gonna be something a lot to talk about.
Caspar: absolutely. I think the Americans call it the Indian summer where the leaves turn from green to yellow, from red to yellow, but it's still
Russell: Yeah.
Caspar: a very nice way of the part of the year. So yeah, I'm very much looking forward to all those events while I'm just going to be with you at the All About Process Management one. And then of course we got Sellosphere in, let's say, think December 3rd, it's this year. So that's going to be a lot of fun. I think the international BPM conference is also somewhere in October, but that's in Canada this year. Unfortunately, I can't join. So that's going to be something we need to do from the sideline maybe. But other than that, let's dig into it. mean, last week, Russell, we did episode one of a mini series where we started talking about context. And context obviously is very important in terms of AI because AI needs a lot of context because without context, AI is just noise. It's just a lot of confidently lying about things in many cases, which
Russell: Mm-hmm.
Caspar: is fine as long as you know that. So there's about six dimensions that we ran very quickly through in the first episode. And we're going to talk about the first one, which is why, the why of a context model. So what type of information is in there and why is that important? So what I did... is I ran a little analysis together with my good friend Claude and asked it to take a look at, let's say, the 10 of the most major ERP and BPM vendors in the market right now, and just asked him, okay, check out, if they talk about context, what do they mean with this? What consists, what makes their context work? Which parts are in there? And that was an interesting an interesting exercise, Russell, in the sense, and then I'm going to really look forward to your opinion on that, is that if you look at those major vendors, there were like Salesforce, Microsoft, SAP, Solonis, they were all in their UI path, you name them, IBM, Vega systems, and it's very often limited to the type of data that they actually have in their system. They don't seem to go outside of their own systems.
Russell: okay, that's what you're okay, okay. Yeah, because if you're saying data, but you know, it's it's it's limited to the data. I would say, yeah, well, what else could the substance be of of this? so but
Caspar: What else do they have? Of course, but...
Russell: but now you're saying it's so it's it's very much looking and maybe also that is like the the nature of a tool to say I'm this is my world, this is my data, I have the all the operations and the capabilities to transform it do whatever what I do with information, collect it, have a persistence on it, have logics analysis, reports, and and and the likes. so w wouldn't it just be the intrinsic nature of a tool just to look in the realms of of the tool itself. Now there may be some tools that are explicitly not so introverted as As as other tools, and that's something we need
Caspar: Okay, yeah, of course.
Russell: to look at. But there's another aspect as well I would like to throw in your direction, my friend. And that is
Caspar: Okay, here we go.
Russell: so so we started off this discussion about context. Yeah, and and that's all
Caspar: Mm-hmm. Yes.
Russell: good. And we have the why, and we already discussed about certain elements on each level. But don't we also have to Clarify the fine line between content and context. Is content context? And if not,
Caspar: Ooh. Ooh.
Russell: it is one-to-one the same. Where is the actual distinguishable
Caspar: Hmm.
Russell: difference? And
Load full transcript
especially talking about tools, data is, you could say that's content.
Caspar: Yeah, yeah.
Russell: Now, where in the in the realms of process management specifically? And also when we're talking about AI, where
Caspar: Mm-hmm.
Russell: does where is is is is content data, does just let's say raw data, maybe? Maybe content is raw
Caspar: Okay. Yeah, yeah.
Russell: data. And then you have a certain context that is then yes, question mark, big question mark there.
Caspar: Yeah, yeah. And that's interesting. And this is an interesting rabbit that you pull out of the head. My intrinsic first reaction was, well, there's only one letter difference between context and content. But the semantic difference is obviously very big. My gut instinct reaction to this, Russell, is that context is made up out of different types of content. And more importantly,
Russell: Mm-hmm.
Caspar: I think, the dependencies between the different types of content. So ultimately, mean, there is transactional data, which you could call content, but I don't think that's enough. You need to add content that describes the, let's say, the parameters within which you need to stay as a company. Because you cannot do anything that you want. there are guidelines to companies that they need to comply with, laws and regulations, and you need to put them into the context as well, and they take the form, I guess, of content. So for me, yeah.
Russell: Very nice. And and that that's that's that's really good. And I think that is exactly the the the fine line where and and you had a we were talking about it last last week last episode and we were
Caspar: Mm-hmm.
Russell: talking about the the noise, data just
Caspar: Yeah?
Russell: being noise, and only by providing let's let's just not say content but providing a structure that helps anybody, a human being or AI. To understand which of this noise or the content is relevant for answering a certain question. And that is then then done, that is done by doing like certain, like maybe like relations. So this additional like connection between, let's say, the raw data that actually then says that helps the human beings or the AI to say, that's relevant, that isn't relevant. That is the norm,
Caspar: Mm-hmm.
Russell: that is. Not the norm.
Caspar: Yeah.
Russell: That is a benchmark that I is is good. Is it just a it's it's a a level that says that's good in that constellation of other data. For example, order
Caspar: Mm-hmm.
Russell: to cash, standard sales from stock, OTIF, that would be a normal benchmark. complicated sales with make to order, blah blah blah, OTIF, different level. So the the the there's
Caspar: Mm-hmm.
Russell: data for the KPIs and there's a result for the KPIs, there's a definition of the KPIs, and then there is this context which is required to be able to interpret and differentiate between, I would say, very simple green and red, good and bad, black
Caspar: Mm-hmm.
Russell: and white, and and and so forth. And and that's where I think when we're looking at the the analysis of tools, as you said, spot
Caspar: Mm-hmm.
Russell: on. Said, okay, so they're introverted. And isn't context very often something that you have to pull in from outside? Let's say the policies, the say the
Caspar: Yep, yep, yep.
Russell: SharePoint. And and then you have the data, but then you have to be able to relate, either pull in, when we talk about that about copying
Caspar: Mm-hmm.
Russell: in or or just to building up relations. Say how well do the tools help provide context? And content they all have.
Caspar: Yeah. Yeah.
Russell: bubble, whatever their their their domain is, they will have content. But then
Caspar: Mm-hmm.
Russell: how well did they actually provide context? And that's something where I think we can do an even
Caspar: Yeah.
Russell: deeper dive at you know at the end of the let's say the the the topics where we're starting off with the why, is that we can then always you know go back and say, okay, in certain aspect of tools, is that just content? Or are they actually really providing a sophisticated way of providing context? Now that that's something we need to we need to look at.
Caspar: Yeah, yeah, and that's a very interesting thing because I I still remember back when I worked in the chemical manufacturing industry, of course they were all using an ERP system and you need that not just for the cash but also for instance for your production planning and you run
Russell: Mm-hmm.
Caspar: your MRP runs. And there is this little thing called a material safety data sheet, an MSDS, you still might remember the name. It's a blast from the past. And in a material safety data sheet, you basically write down for each material that you use in your production process, what are the things that you can do with it and what are the things that you cannot do with it and what are the regulations and norms.
Russell: Mm-hmm.
Caspar: And those data sheets, they didn't used to be included in your ERP. And then later on, they made provisions for that so you can put them in. But I would say that the content of such a material safety data sheet, is very relevant context at the moment that you want to have an AI agent run a production planning. And that is, think,
Russell: Absolutely. Absolutely.
Caspar: that's the outside information, the outside content. I love the fact that you make that distinction between content and context. You can throw more of those curve balls if you want, but they're really good. Because there's so much information that sits outside of the transactional systems. that you need to take into account if you run a, certainly if you run a major organization. I'm not talking about the little mom and dad pop-up shops left and right, but even they already need to comply with regulations. But if you are multinational, it's massive.
Russell: And and just think of this one aspect. in in an ERP system, a lot of the context is actually, I would say, hacked into the
configuration and the customizing.
Caspar: Yep. Yep.
Russell: So, and and then if you have a system that can actually read out the configuration, it can actually then also interpret or understand better the actual the data that is being produced. but sometimes the explanation explanatory factor of the configuration isn't quite clear. So here I think maybe to to help, and this is something that, you know, we'll look at tools, it's interesting to say, so what do we before? I think 100% of the well, 90% of the people, if you'd ask them, you know, what is important, say just look at the data. Or the you know, you look at the data and that's content and that's what you can use for AI. But now, you know, with this discussion, I'm thinking, hmm, but maybe that's just not enough. You need to look
Caspar: Mm-hmm.
Russell: beyond the data and the content. And think about what are the elements that you can, as a system, you do you need to read the source code, the app app, user exit, or the
Caspar: god,
Russell: the configuration, in order to get a certain context to understand the data better. Yeah. And and and
Caspar: Yeah. Yeah, absolutely.
Russell: this is then thinking completely multidimensional, much more complex than just this simple yeah, I have a database and there's a table, there's a schema, and the data's there. Nah. It's not that simple. And we know that from the normal
Caspar: No, yeah.
Russell: debugging and working with systems anyway. But I think now in this discussion it becomes highly relevant because you're you're asking certain more much more complex questions based on a system. And when I say system, I mean from the humans, the user
Caspar: Mm-hmm.
Russell: interface, the code, the configuration, databases, and the data and the interfaces.
Caspar: Yeah.
Russell: You look, you you actually sometimes need to look at the whole system in order to be able to answer certain questions. And that's where I think we are digging into okay, how how are the current systems and the systematic thinking, have they been set up in the right manner to be future proof in in
Caspar: Yeah.
Russell: in the in an in the different use cases that you have with AI? I mean I'm sure if you just have a data analysis analytics AI that's looking at the curate of the data, it's fine, but you have to look at it from different use cases. And then I think maybe even software as it is has to be rethought in its accessibility, yeah.
Caspar: Yeah, so interestingly is what you're describing. And I think this is also a very important distinction is that if you look at the data that a lot of vendors and ERP systems claim that that's the context model, that just shows you what happened. It doesn't show you why it happened.
Russell: Mm-hmm.
Caspar: And the why, I think that's your business logic, your decision logic. That's where not only do you need to get that external information in. I think
Russell: Mm-hmm.
Caspar: one of the best examples I've had is I was working with the
Norwegian oil and energy company, a big one. And we were talking about the fact that they need to comply with an enormous truckload of norms and regulations from an environmental perspective, from financial
Russell: Mm-hmm.
Caspar: perspective, while you can think them all up. And they were using a third party that was keeping track of all the changes and the updates that were done in the relevant regulation and then translate
Russell: Mm-hmm.
Caspar: that into what does that mean for that oil and energy company. Because even if you just step away from the core processes that we normally always talk about, certainly when it comes to AI, changes in regulations for those big companies, what they mean is a gigantic amount of change management because the relevant people need to be updated on, hey, wait a minute, there's gonna be an update in this particular clause of this particular article of this particular chapter of this particular regulation. And
Russell: Mm-hmm.
Caspar: AI can not only use that regulation as content to put into context, but it can also then help to say, listen, yeah, that's very good. This particular change in that regulation has an impact on this particular process. And that means that you need to take this into consideration for your next decision. And the point I'm trying to make here is that if you really want an AI to be able to find out what the potential impact is of a change in one part of your process landscape, you need to have modeled your landscape. Maybe not to the smallest detail, but certainly to the point where you can have dependencies and connections between activities in individual processes and how they relate to each other. Because that
Russell: Mm.
Caspar: I think is the, that's the fabric on top of which you build up your context.
Russell: Yeah, now and let's just spin this yard a bit further. So we have this system you're talking about, yeah, is is being able to understand or pin let's say put the finger on the relevance of a change of a of a of a compliance rule regulation to the actual to the actual Let's say impacted processes, let's say system processes.
Caspar: Yeah, okay.
Russell: and that that's the case, that's the capability this tool has. Now, is he capable of doing that because somewhere in this system, context has been implemented, or is the act of putting the finger on the let's say the relevant thing? Is that the context?
Caspar: Hmm, that's an interesting question. That's a really interesting question. Because you don't know what the exact information is that an AI agent's based his decision on. Because somehow it is inherently a black box, right? You give it context, and within that context, it basically does what I think is the best thing to do.
Russell: Yeah, but then AI is given, you're given it. Okay, so yeah, yeah. So in order for AI to do a job, you provide it with a context. But the the tool you're talking about, yeah,
Caspar: You do the same with humans if they need to do a job.
Russell: yeah. And you would then but I'm I'm just thinking so so somewhere in this system, context has been implemented. That that's my take on it. And there's a k there's
Caspar: Okay.
Russell: a function on there that says change here, show me the impacted process. So it was quite a straightforward functionality, which could also be consumed by an AI in a query, and that could then be used to do some additional analysis or a process, let's say design that compensates for it, or whatever mitigation action definition or something like that. But the the interesting thing is that such a system Provides context, but ultimately in the system of it, somewhere ha and this is something we need to look at. So how, yeah, is actually
Caspar: Mm-hmm.
Russell: this context now being implemented? And I think it'll be it could be a very simple methodology where you have processes with risk impacts, and the risk impacts
Caspar: Mm-hmm.
Russell: are assigned to or mapped against certain compliant compliance frameworks. And so you already have that an let's say a logical, deductible, transparent mapping of processes of their risks in the categories of risks that they fall against, then a compliance framework. And
Caspar: Mm-hmm.
Russell: and then there's a lot of details that you need to be able to interpret based on the content, but the context,
Caspar: Mm-hmm.
Russell: compliance to process, has been done over a certain methodology, and that's That's so so the c so they've
Caspar: Yeah.
Russell: they've thought about a very specific context of this the process is this way and these are the risks and and and the compliance framework is is something that has to be complied with. And and here to
Caspar: Yeah, yeah, of course.
Russell: be able to and and the so the the context has been analyzed and been implemented based on a certain methodology. Now sounds like a a way of a trip. But then just to get the terminology right, is then but this context is a structure that can be consumed by humans and AIs. It can be queried by an AI or
Caspar: Mm-hmm.
Russell: can have a function that says check out that process because the framework has changed. So there's certain tools and functions around it that help. And that's actually what we're acting exactly what we're talking about. The compliance regulations in itself is mainly content,
Caspar: Yeah. Yeah.
Russell: but building up Correlations. So I think a lot about, so if you think about a graph, it's a lot about very, very sharp kind of like
Caspar: Mm-hmm.
Russell: relations that say this is a structure. And if you're looking for answers in this realm, take into up take into consideration the relevant compliance frameworks. If you want to do a process design,
Caspar: Yeah.
Russell: make sure that you comply with these frameworks. Make sure that you try to standardize the guys. So this this is actually so ultimately if you break it down, it's quite could be quite simple, but then we still these and that's when it gets coming back to what the original question we want to dig into today is then it is so interesting to say, but there are different forms of context. Now now we're getting down to the level of understanding content versus context. But I think that's where when we are looking at say the why, you say, that's easy. It's the business model. And you said, yeah. Okay, the business model
Caspar: That's not enough.
Russell: is somewhere somewhere's written down there's a 20 pages of strategy or business model.
Caspar: Yeah, yeah,
Russell: But we now have to say, okay, that's content. That's the business model content. How do you get that into a context into context? And where so like like the compliance framework to the process, how
Caspar: Yeah? Yeah, yeah.
Russell: is the connect there? And how can that be done? And then I think the why we can look at different. aspects of why and why it it it is important for contextual layer. And of course it is to be able to deduct good and bad and differentiate between things you do want to do or don't want to do. The why I think is the most important context you need to provide. Or at least one of the
Caspar: Yeah. True.
Russell: I most probably it's the most important one. But then we let's let's just discuss then also
Caspar: Yeah, but
Russell: what examples we have and how we then would say How it what what structure does that need to be prepared in? Is it just a balanced scorecard as a model? Is that sufficient? Would that be more than
Caspar: No.
Russell: content? Would it also be context? they're not bad the balanced scorecards. So let's let's dig into the why.
Caspar: So here's, so no, no, I agree. I agree, but I mean, this takes me back very quickly to the one picture I always show clients when I start talking about business process management to them. And that is a picture that basically says you have a strategy, you translate it into an operating model, an operating model has different components, one of them always is your way of working. So your process landscape. enriched with things like your risk controls, your roles, your applications, the inputs, the outputs. And if you want to really go level deeper, then you end up with the work instructions. That whole structure from your strategy with your strategic initiatives, the business model that you have, the different components of your operating model, which application landscape do we need? Which process landscape do we need? How do we motivate and reward people? All that And then the process models on how basically the graph that you talked about on how these things are related and connected. That is, I think, the skeleton of your context model. And once you have that, you can then start to say, okay, I now know how we agreed internally on how we are going to work together because we know that if we want to be able to send out a proper invoice to a customer, We need to have a proper sales order with a good delivery and a shipment. And if that doesn't
work, then the billing will not succeed either. So we know how that works. But then there is all of the, I would say, extended contextual information that sits outside of your organization. Because you're not running your organization in isolation in the world, right? So you're doing that in interaction with local governments, with... international bodies, with regulatory bodies, with auditing companies. And all of that information is contextual as well. The fact that there are financial accounting rules in Basel 15, I'm not sure how far Basel is right now, but, or Sarbanes-Oxley, that is content that should be part of your context. Now the real question then becomes, and that will be my question to you, Rus, in this case, would you need to simply have a reference to that type of content or would you actually formalize that content into your context model?
Russell: Mm.
Caspar: Because that's a shitload of information. Excuse my French.
Russell: Yeah, yeah. And I I think ultimately you're always gonna have to trade off the the depth of importing, let's say, building up the cot that is part of your content versus the referencing based on the the actual use case and the the the speed and the then do you have time to actually then do a a research and go into the details of that, or is it s sufficient enough just to have the reference? But it also shows a lot about how important it is that companies say like Salesforce in a customer c environment, the whole systematic is around trying to get a three and sixty-degree view of a customer. And in order to really deduct the right decisions on when and how to influence a customer, you're not just going to have pure data, but the way the cases are strung together. In a like a big network that is ever changing, every customer has a different behavior and a different set of activities and cases and relations and contracts and all this. That that is actually I think a good example to think about the actual the in the context of this customer, what is the next best action we can do to close the sales? Yeah. So so
Caspar: Yeah. Yeah.
Russell: and and and that's where I think a lot of and here. maybe the the quotes are in SAP and they're just related. Yeah. But the key information
Caspar: Yeah.
Russell: of this, you know, the which products, product group, maybe, and the the some of the like core corner points of the the contract, maybe you do want to have them in your in your system. so so I I think it's it's it's really interesting that also there's every system in the in their in their content, somehow for their use case, they are also having providing a certain element of context for as for the bubble
Caspar: Mm-hmm.
Russell: that's say that they're working in. And certain AI capabilities are just placed directly onto that. And and they they
Caspar: Yeah, yeah.
Russell: know how to use it because the people know their meta model and they and they know where how how they can say, just there's a data in Salesforce, just use the the a the Salesforce AI Capability will know exactly how to use it. Same for the BPM tool. What I think is really
Caspar: we have a-
Russell: it ex exciting is that all these systems are incomplete. So, you know, who has all the data for the CIST degree? And and and that's where where
Caspar: That's an interesting one, Russell.
Russell: it becomes interesting to see how you can,
Caspar: Yeah.
Russell: you know, expand that. Yeah.
Caspar: Two remarks. To start with the second one, and I think I wanted to make that remark earlier, because the way that you describe it basically means that virtually any system, sorry, an AI context model transcends virtually all applications that you use in your company, because none of them contains everything. They always
Russell: Mm.
Caspar: focus on certain part of it. And of course, an ERP system contains a lot. but still it doesn't contain everything. And then the question really is, do you start with, let's say, your biggest heap of data and you start to build out from there, or are you going to basically say, well, I just want to have something independent which sits on top across every application that I have and have an independent context model? That's
Russell: Oof.
Caspar: one, but before you dive into that one, and we're getting almost at
Russell: yeah.
Caspar: time. The other one that now slipped my mind, which is a pity. Yeah, no, I know, I know. The other thing that you were describing around, instance, with Salesforce and that quotation and how that is important for the decision how to best attack the customer or deal with the customer. Do you know what this comes down to in the end, all of it? Human behavior on how to properly set up and maintain your data. This is about hygiene in systems. And I'll give you the stupidest example. We use Salesforce, right? I run a go-to-market team and I'm depending on the pipeline reporting I have in Salesforce. But if my esteemed colleagues, the account executives, do not put my product in the opportunity product list, which is an administrative action they need to do, I don't see that opportunity at all. So every time I see deals popping up that... are completely have gone under my radar because they simply didn't put it in. And if they don't put it in, the AI context model also doesn't know about it. So in the end, a lot of this stuff will ultimately fall back to do people put in the right information in the right part of a system.
Russell: That's a good a aspect. I mean, but that's a the the classic, I would say. there's also there's all there's also
Caspar: Of course.
Russell: maybe an element on I you know in Salesforce sometimes methodology like they have the radar and they have some kind of like selling and you know, like selling methodology implemented. I'm always just wondering in how far that interpretation of the data, it says so the classification of a customer for In the pipeline, he's he's near to closing, anything like that.
So you have a
Caspar: Mm-hmm. Yeah, yeah, yeah.
Russell: classification, which is also done by you, you know, says how much percent do you guess is going to be? And sometimes we have mechanisms that say, This data is all there, he's ready for closing. And sometimes you do it manually, and somebody says, Hey, you know, give me your give me your commit. Are you gonna close them in two weeks? And all stuff like that.
Caspar: Yeah, yeah, yeah.
Russell: And I think that's that's where the limitation of gets blurry that the the content and the context sometimes is distorted by well let's say humans as you mentioned also with bad data or just kind of wrong interpretation
Caspar: Mm-hmm. Yeah.
Russell: of data and the setting of the data and and their judgment ultimately and sometimes false I would say false models that are applied there rules that are based on some sales dude from the sixties
Caspar: boy.
Russell: That that wrote a book, and then Salesforce takes the book and implements it in in part in their software. And and and then and and that can be distorting for the interpretation of the AI if you say just purely do it based on this pipeline of the way the status is and the way this is. So and and maybe external kind of like human behavioral like patterns and interpretation on a more personal level could be much more valuable. So What I'm saying is I believe that these systems that have built-in AI capabilities, I I know I'm I'm repeating this every time again, somehow
Caspar: Yeah, I know.
Russell: they are they're they're they they will I think they're always they would they can they become become blind, betriebs blind for the content and the context that it's giving them. They cannot question that, which is good. Rack is all about this is your structure, stick to it. But you have to be aware and you have to question yourself is Do I not need to look outside of the box? Yeah. And and that's why I'm a strong believer of that the the future of agentic systems where agents talk to agents is something where they need a contextual layer that is to a certain extent independent and and can help cross-relate context and
Caspar: Yeah.
Russell: content between different systems and which are like guiding, let's say principles. Different agents that are actually working in an orchestrated manner or talking to each other have one more there's a overarching contextual layer over which they can then f find a common ground to discuss or work against things and they can be validated
Caspar: Yeah. Yep.
Russell: outside of the systems. But to the other question, is that something we could say, and I I would not negate. The real value of content and context in the applications. That is the that's where
Caspar: No, of course.
Russell: you are. That is there. And you have
Caspar: Yeah.
Russell: to understand it more and more to be able to use it properly in the context of the, let's say, these isolated IA capabilities. But a company, an enterprise thinking digital
twin and context for the enterprise, they is a I think as a as a leader, you have to already start thinking independently, tool agnostic. You have to think about your company and say, this tool,
Caspar: Yeah.
Russell: I di I have my Linux, let's say my have my application infrastructure and I have all these data, I have content, I have context, but I need to think on on I I need to think beyond that. I I need to really look on it as a like a contextual architecture on on that, how I can make that all fit together, because these will be the cutting edge use cases. That that will make the difference in in how you
Caspar: Yeah, yeah.
Russell: steer your business big time and how you have agents talking to each other and not being confused because they have their own separate contextual layer, and then they have a conflict. That how is how is that going to be resolved other than
Caspar: Yeah.
Russell: by some overinstance that has its own context, says, Yeah, yeah, yeah, that's fine for you, that's fine for you. But my context says,
Caspar: Yeah, yeah, yeah, yeah. Yeah, yeah, that's interesting.
Russell: You're right, you're wrong. So like a like an overlord on AI and based on a on a on an enterprise context. And and and that's where and maybe that just we we need to somehow get back to the original topic we wanted to talk about, straying away the why. So so let's take this picture of everybody can have their little tools and their little agents and their little content and their little contextual layer, the islands, yeah.
Caspar: Yeah, yeah.
Russell: But for the for the for the overlord agent.
Caspar: The Dark Lord agent, yeah, there we go.
Russell: Exactly. Exactly. The main context he needs to have is the why. Because it's like you will have some on traceability here and something there, and they can all have their their I think the if we look at the six dimensions of context that you were, you know, you we listed up and discussed last week, the Ultimate decision on hmm, we have to take a decision between left or right. Yeah, and and this agent says here, this tool says this, this content says that. The ultimate escalation and decision should be taken on why are we doing this? Now,
Caspar: So.
Russell: is there a dilemma on a on a level said, yeah, but if I just say why are we doing this just to earn money, wouldn't that be that
Caspar: Mm-hmm.
Russell: you always overrule? compliance for the sake of making money and the I will go crazy. So it cannot just be so
Caspar: Ooh. Yeah.
Russell: simple says the why. To define the why, it's you need the guardrails around it.
Caspar: Yeah, because if you don't, you run into non-compliance issues that can run you into serious fines, which is losing money. but
Russell: Yeah.
Caspar: I just came up with a new business model, Russell. It's highly illegal, but it's funny. So
Russell: Ha ha
Caspar: I'm going to pitch it anyway. If an organization really needs to find out why they do the things that they do, I would suggest
Russell: Yeah.
Caspar: that all employees who have children between the age of two and five, bring them
Russell: Yeah.
Caspar: to work because the only thing they will do is ask why all day long and you need to answer them. I mean, that's it. So it's the five whys. I mean, it's a very normal lean tool that is being used. If you start asking why and you ask it a couple of times, you're going to get to the core of it. And I think that
Russell: Mm.
Caspar: that repetition of let's say five times why would give you the context. that you need on the white part. We need to start wrapping this up, unfortunately. But we have still a couple of like five or six or seven episodes on this topic to go. So we can be creative on the topics that we do next time. Maybe do a little bit of why again.
Russell: Yeah, and and and build the bridge from the why to the second topic so we can look at the differences as well. And and maybe look in
Caspar: Yes. Yep, absolutely.
Russell: some I think the next episode we should check out which systems we know that really have why content and context. And and I wanna come back to the the ballot scorecard. as as a as a because
Caspar: Okay, yeah sure. that's good.
Russell: Because I'll I'll tell you why, because I I don't know, I had that experiment as chat GDP came out in a new version, I think it's two, three years ago. And I took some some models, and one of them was a balance scorecard, organizational models, some process models, and I don't know, sort of system architecture. And and I and I threw it all into chat GDP and then I started asking questions that were correlating the the models. So I said, you know, how can the person in logistics sus support the sustainability objectives of this company, which were in the b business scalance balance scorecard as well. And he
Caspar: Mm-hmm.
Russell: really managed to derive how she could change her behavior, what she needs to do to contribute to the sustain sustainability objectives of the company. And and I've really liked and I was thinking I this I completely disjunct from our discussion here. That's what I did that, just to see how good ChatG2B is in interpreting PDFs actually. But the the the point was that now thinking about it, isn't that model type balanced scorecards something that specifically is focusing on creating context? That's that's something
Caspar: Yeah,
Russell: we need to begin to.
Caspar: that's interesting. Maybe it its time far ahead for that. So let's dive into that next week. And then we're gonna revive the balance scorecard. Cool. Yes, yes.
Russell: Okay. And separate models.
Caspar: And what else we can think of? It was...
Russell: No, I mean seriously, if if if we come if we with this episode man manage to revive some why contextual models and and can and and say look, this is something really valuable, you should start digging into it. Because I think they were
Caspar: Yeah.
Russell: very vogue decades ago.
Caspar: Yeah, of course.
Russell: but but maybe models like that, because then they were, you know What would you do with them? They were there, they were documented, then you could have some statuses and follow up with them and there was it
Caspar: Yeah, yeah.
Russell: was a nice methodology. But if we now identify on each of these levels recommendations on, hey, you've got your business model and PowerPoint and blah blah blah, but for AI, you should think about modeling together. I said the model is dead, log as it's the model. Yeah. And and then
Caspar: Yeah, Yeah, yeah. Cool.
Russell: you can you can start putting that together. That's a good outside for the outlook for the for the next episode.
Caspar: Absolutely true. I did thoroughly enjoy this one, Russell.
Russell: Me too.
Caspar: So thank you for listeners of keeping up with us for the last 40ish minutes where we dove deeper into the why of context models and business decisions. And we decided to bring back the Balanced Scorecard. So that's going to be, do we need to go with like a crowdfunding, crowdsourcing initiative to bring back the Balanced Scorecard, Russell? Let's discuss that next week.
Russell: Well no, it's I'm I'm gonna I'm gonna we're gonna close this down. I'm gonna go to Claude Code and I'm gonna ask him if he can program me some some a an a super app, a simple app for for creating balanced scorecard models. And you wouldn't need
Caspar: cool.
Russell: to do them all manually, would you? You would just say, Hey, think about, you know, you would in a dialogue you could create of these models. And it's not just balanced scorecard, there's a lot of different models out there, but they I think now in the time and age of AIs Creating them and consuming them and the relevance of them. Everything's changed. Yeah. So we're not even talking about
Caspar: Yeah, yeah, Yeah, all right, cool.
Russell: processes, we're talking about the why, the you know, the business strategy.
Caspar: Very good. Let's continue that conversation in the next episode, Russell. Thank you very much for today. To our listeners, I say thank you for sticking with us and we catch you in the next episode. Thank you very much and have a great day. Bye bye.
Russell: Thanks for listening. Bye bye.