8 Elements Required for Autonomous Supply Chain Planning

By
Niels Van Hove
19m
8 Elements Required for Autonomous Supply Chain Planning

Niels Van Hove explains how to achieve "lights out" or autonomous supply chain planning.

What is required for Autonomous Supply Chain Planning?

The following is a transcript of the video "Autonomous or 'Lights Out' Supply Chain Planning; what is really required?"

Bethanie Maples:

All right. Hello, everybody. Today, Niels and I will discuss Autonomous Supply Chain Planning and what's really required to make that happen. Much of this is outlined in an article Niels published in Foresight: The International Journal for Applied Forecasting, where Niels is an advisory board member. All right, Niels, by way of introduction, can you give us a summary of the situation at hand?

Niels van Hove:

Yes, I can. Thanks for that introduction, Bethanie. So well, if you operate in supply chain and supply chain planning, you can't get past the terminology of lights out planning, autonomous planning, light tax planning, they all rate high in the hype cycle and for good reason. And it's very exciting, actually. A lot of progress has been made in technology and we see autonomous cars driving around. So why wouldn't we have autonomous planning? So in this article I really detailed what's required. So eight elements, which are required for a real autonomous planning and also highlighted in my view, current systems, what I call wave one ERP systems or wave two APS systems, Advanced Planning Systems, are not sufficient to support autonomous planning and a new type of technology, which I call wave three, a technology is required to support autonomous planning.

Bethanie Maples:

All right, this makes sense. But can you give you some more context? How does the supply chain planning technology evolve to get here?

Niels van Hove:

Well, if we look way back, it started really with enterprise resource planning systems, ERP systems, they sort of came up on the scene really in the '80s and that was all about capturing transactions. So if I have a sales order or a purchase order, or I move inventory over production order, all the transactions get recorded, which is great actually. And also financially, in my financial statement, I could see all those transactions, but those ERP systems, they had some planning, like a MRP, Material Requirements Planning, and a little bit of production scheduling, but they were never really a planning system. So I call it wave one, really, ERP. They're very mature systems now, but then around the year 2000, you get what I call wave two Advanced Planning Systems. And that's really where technology started to focus on planning.

I got the demand planning, forecasting, supply planning, inventory planning, all of a sudden there was this focus on integrated planning. However, those technologies were never really made for autonomous planning. And they're coming a little bit at the end of their life cycle. They're mature now. They're becoming commodities and we need really a new type of technology to take the next step, to take the next leap into autonomous planning. And that's what I call wave three planning systems or systems of intelligence. And yeah, that's really what led to this wave three ERP, APS, and now a system of intelligence to support autonomous planning.

Bethanie Maples:

All right. Let's back up and tell me a little bit more about the shortcomings of these APS systems, so that I can understand what the leap is to these more intelligent systems.

Niels van Hove:

Yeah so, a lot of them are built on older technologies or less platforms that are 10 or 10 plus years old, and they're focusing still on planning. They got better, they got more powerful they look better for the end user as a user interface, but they're hardly ever really end-to-end, they're hardly ever across the whole value chain. And we've seen the upcoming of data and uses of data. And those planning systems often are not really capable of handling a really mass amount of data and analyzing these at scale and at speeds. And yeah, one very important thing is they're not really made for autonomous planning for making decisions. They can plan and they can maybe automate processes, but they can't automate decisions. And if we really want to go to autonomous planning, we really have to make autonomous decisions as well.

Bethanie Maples:

Okay, got it. So let me just riff off what you said, what I'm hearing is that these systems, while powerful in terms of computing capabilities, weren't really trying to take on any of the function or like didn't have a high level integration with like human cognition.

Niels van Hove:

Exactly, exactly. And that's sort of, you see now that they start at that on to those ways to planning systems so that it might get a little bit bloated, but they were never really built for that and that's where the big difference is. And the reality is that 90% plus of enterprises still use Excel to tie plans and information together. A plan is still used 50% of the time to gather data and manipulate data. That's far from being autonomous. And those stats, they are real and they're real in the enterprise. So we need that wave three in the system of intelligence for further digitization support, the ever-increasing availability of data, and being smart about that and to have the intelligence then to make and implement a business decision. This will relieve planners from a lot of cumbersome tasks and increase automation and make automation indeed possible.

Bethanie Maples:

Tell me a little bit more about the architecture of these third wave systems. I mean, you started a hinting at it, but could you break it down for me?

Niels van Hove:

Yeah so, I defined in the article, eight elements. And let's take the first two, which is the digital twin and a common data layer. So the digital twin is really a copy from, I have to be able to copy exactly what a planner does in a digital way, and that has to be across every part of the planning function. So that could be forecasting, supply planning, production planning, inventory planning, transport planning, order allocation, you name it, every part of that end-to-end value chain. You need to be able to digitize that and let's call that the digital twin.

But for a digital twin to be effective and even to be possible across all those functions and across all those different elements, you need to have a common data layer. You can't have the siloed ERP or old APS approach that still beat siloed information in those different functions, because we have to make global decisions. We have to make integrated decisions if we want to plan autonomously. So you need one common data layer and you need to have the ability in the wave three system of intelligence to tap in all those different data elements, internal, but also external to support that digital twin. And so we need a digital twin and we need one common data layer really to support it without those two autonomous planning wouldn't be possible.

Bethanie Maples:

Okay. I think I understand what you mean by end-to-end data layer in order to have a digital twin. What else would you say is required besides this end-to-end value chain data and a digital twin, for this third wave of this next step in the industry?

Niels van Hove:

Let's take the next three. And that's really advanced analytics, automated problem solving, planners need to solve problems, and flexible goal settings. So a lot of the sort of wave two assisted, they have advanced analytics, or they start to add advanced analytics in terms of descriptive of what has happened, predictive, what will happen, but what's new now, as well, and what's required is prescriptive. What's the best thing I should do? What should I do? And an autonomous planning system should advise on what you have to do and execute that then itself as well, of course, but it starts with that advanced analytics to a support decision, automated those planning and decision making. The second one is problem solving. And this is about automated simulations for example. What type of scenarios in my value chain, which can be a thousands or millions that explodes very quickly are possible and the probabilities around that?

And then decide on the handful of the best ones to take forward. I saw that automated problem solving, the tradeoffs. Again, that needs to be part of a digital twin, that process, but the problem solving it's smart... a tradeoff of opportunities and risks that need to be part of it as well.

And then finally, you had a flexible goal setting. There's still, however smart the machine is, you're still in the autonomous planning age. You still require a goal. So the machine has to work to what's a goal, and it doesn't always have to be logic. I always give the example as logically a business tries to make a profit and optimize the EBIT, but you can go for market share and make a loss, for example. If you don't program that into a machine, the machine will know. So you require flexible goal setting and that human interaction in that element still, whether human provides goals, because what I always said as well, like an autonomous car, without a goal, you never arrive at your destination and that's the same with autonomous planning. You still require some flexible goal setting.

Bethanie Maples:

That makes sense. Okay. So what I think I'm hearing is that this new breed of system has much more connected data. And then the intrinsic design is much more flexible. Would you say that's right?

Niels van Hove:

Yes. Well, it needs to be connected again, one data layer. There still will be required. Some human interaction, of course, you have still have the guidance and goals, but maybe also ethics or other elements in terms of decision making, of course.

Bethanie Maples:

Yeah. Tell me a little bit more about that. How do you really embed human cognition? What are those last few points?

Niels van Hove:

Well again, if you go to full automation, I think there's always an element. And we've discussed it previously. There's always an element of augmentation required and the further you go out, the business arise and in planning and decisions become bigger. Some human interaction is probably required and decision above $10 million, will you let the machine make it? Decision of a thousand dollars you might. But where is that boundary? So that's in terms of value or impact, but then you've got the human element, ethical or moral questions or investment decisions. Of course there is some human elements is still required. So look, I pull out these eight elements to say what's required for autonomous planning, but deep in my heart, I always think that some human interaction is definitely required, yes.

Bethanie Maples:

Okay. What we're through if you will, the final three points from your article about how to make these intelligent systems truly intelligent.

Niels van Hove:

Yeah. The final three are really automated execution, planning without execution is, well, what is it, it doesn't get you anywhere, and self-maintenance, and self-learning, and a system of intelligence this way, three planning system need to be able to execute planning decisions.

So, I decide to produce something or decide to move inventory. How do I execute that? So yeah, the system of intelligence needs to be able to ride back to ERP, underlying ERP and APS systems. And that can be a global company. It can be dozens of them.

Third, it needs to be very smart communication protocols and the ability to ride back to underlying ERP and APS instances, right? So that's really required to perform autonomous planning, execution part needs to be done as well. Communication after you execute the plan, you might need to automate communication to a group of stakeholders of what you thought the machine has just decided.

The next one is self-maintenance. So after planning run, or the machine can observe that lead times change that production quantities or minimum order quantities or inventory setting should be changed. Usually this is a planner that optimizes or improves changes those master data and planning sentence.

If you want to do autonomous planning. And the system of intelligence has to do that. So again, you have to update your master data and planning data autonomously. And so that's an important element of autonomous planning. And then finally, it's a self-learning aspect. I don't think you can have an autonomous planning system without a self-learning capability. Because if you execute a plan automatically and it's continuously the wrong execution and you don't learn from that then, okay, where do you go?

You never reach your goals, or again you don't have a good autonomous planning system. So it needs to be self-learning. You need to be able to capture all the decisions, learn from those decisions. I have feed it back partly into the machine, but maybe also even into the human too, this self-learning cycle as well, that should be part of the autonomous planning session. That's really the eight elements I described in my article.

Bethanie Maples:

And that's really interesting because so much of the research and machine learning has been around like pure automation, but really what you're speaking to is a system that flexibly inputs, human cognition and learns from maybe the gaps in human cognition in order to have like a more complete picture.

Niels van Hove:

Exactly. And I agree totally with you that the focus is on machine learning and artificial intelligence and that's the whole thing to talk about, but let's not forget about the human and the augmentation, what the human can bring. And a lot of the automated decisions don't need machine learning. It's a simple trade off. So you don't start with machine learning either.

Bethanie Maples:

Yeah. You use the technology appropriately.

Niels van Hove:

Exactly.

Bethanie Maples:

And just bulldozed everything with it. Okay. This has been really great. So you've taken us over these eight requirements for kind of autonomous and lights out planning, but I was wondering if to kind of wrap up, you could summarize what this journey might look like for supply chain professionals, maybe in a few bullets.

Niels van Hove:

Yeah. So basically we touched on that. If you go on a journey of autonomous planning, remember it's not just about the machine and machine learning and autonomous planning, you have to understand the whole picture of where is the human involved, if you want to do this right, I think. So that will be my piece of feedback and to supply chain leaders and the supply chain planners.

And secondly, understand what's required. So I provide these eight elements, know those elements, if you go to what's the journey of autonomous planning. And understand that a platform or a system really needs to be one single integrated platform that provides all those eight elements, it need to be flexible, it need to be scalable, it need to be able to implement it at speed. And it needs to be able to interact with hundreds of entities within your business. So keep that in mind, if you want to get on that journey of autonomous planning.

And finally, it's what we say, don't focus only in autonomous planning where does the human play a role here, but look, we see the technology is here now. It has arrived. It's being used all over the world and it's getting adopted more and more. And it's really exciting times for supply chain and the supply chain planning in specific, Bethanie.

Bethanie Maples:

Awesome. So, flexible, powerful, well-designed systems that incorporate human cognition leads the way to autonomous or lights out planning. This has been really great. Thank you so much, Niels.

Niels van Hove:

Thanks, Bethanie.

Get social with us!

Follow us on LinkedIn: Cognitive Automation Community
Follow us on Twitter: CognitiveAutomation

 


By
Niels Van Hove
,
Senior Engagement Principal, Aera Technology
Published:
February 25, 2021
Share:
SUBSCRIBE TO WEEKLY EMAIL
-->