Inventory out, sell-through up: AI in the supply chain
Posted: 12 October 2017 | By Charlie Moloney
A question we’ve been asked recently: ‘can artificial intelligence (AI) help me to manage my supply chain?’
We were sure that AI technologies could be of use, but we didn’t know exactly how. Which is why we asked Greg Brady, CEO of One Network and a columnist at the Supply Chain Management Review, to explain it for us.
What should AI be achieving in your supply chain?
“Supply chains have many parties; they’re usually quite long, when you go from the guys fork to the yard that grows the product, there’s many players in between”, Brady told us.
“There’s many processes that have to happen to satisfy what the consumer wants. So as part of an AI framework, an autonomous agent is defined as being an algorithm that can replace the need for the user to be involved at all.”
“One of the biggest problems (or opportunities for improvement) is that there’s been a lot of visibility tools that say, “Hey, you’ve got a lot of problems”, but they can’t effect fixing the problems and all it does is drive everybody crazy seeing problems they already knew about.”
“So that’s the first goal: how do you build an algorithm that can sense the problems, and then fix them on its own?”
As a consumer, you know that you don’t wanna drive all the way over and find out you don’t get the stuff
“The second thing is that these algorithms have to be forward looking in nature. Because to effect improvement in the supply chain, you can’t be looking at the past, you must be looking at the future.”
“You’re making decisions on what you’re going to do. As those problems occur, you want enough visibility so the AI system can actually fix problems wherever they occur.”
A specific use case – Fine Dining
“So let’s take an example: a Fine Dining company.”
“This Fine Dining company runs a limited time offer (LTO) along with commercials and promotion, but it’s very hard to predict how much more traffic is going to occur in the restaurant as a result of that.”
“They prepare their supply chain with some amount of products in the store, some in the DC, and then they tell their logistics guys, “Ok we’re gonna run this LTO, everybody get ready”.”
“In this case, AI powered automated agents need to be able to see in real time. After initially guessing what the forecasted traffic was gonna be, it needs to read the consumption of the point of sale data every hour as it’s pulled out of the POS system to interpret whether that forecast is correct.”
“If you think about a forecast: the forecast that you start with under your LTO is, “How much beef am I gonna sell after this commercial?” Well, it’s a whole lot different of a problem when you say, “How many people are gonna go in each and every store, and of that how many of them will buy ribeyes vs. fillets”.
After initially guessing what the forecasted traffic was gonna be, it needs to read the consumption of the point of sale data every hour as it’s pulled out of the POS system to interpret whether that forecast is correct
“It’s one thing to get a good guess of what beef you will sell, but it’s a whole different problem when you try and figure out every single plate that every customer will buy in your store.”
“What autonomous agents do is they say, “Ok, I think I’ve got a good forecast” and they set the stage for what the LTO demand is gonna be.”
“And then all day long there’s 15 million of these guys, running around in this casual dining company looking at every store for every item and what was actually bought, then interpreting Koggle impact – that means, “Why is this store not selling what I thought it was? Why is it selling too high, or too low?””
“And the agent adjusts automatically. Then it talks to its brethren agents and says, “Quickly order me more product if I’ve got a projected out of stock” – note: not an actual out of stock.”
“It will quickly create an order in good time before the store receives a delivery, thereby fixing the problem before it ever actually happens.”
Taking the burden off human store managers
“Supply chains have been managed by humans historically, and that’s even the case today. That person would have to look at 15 million combinations of store and plate definitions to even remotely try to figure out how to satisfy the supply chain’s problems.”
“Well, that can’t happen. Consequently, they put in a lot of inventory, they have a lot of waste, and people show up to the limited time offer and they don’t have the product. As a consumer, you know that you don’t wanna drive all the way over and find out you don’t get the stuff.”
“These autonomous agents, looking at the real-time data across the different trading partners, doing the work rapidly, sensing a problem before it occurs, making intelligent decisions in order to resolve the problems before it happens, takes all the inventory out and takes the sell-through up.”