The Blueprint for AI in Large, Global Supply Chains

This article was originally published in the Supply Chain Alpha newsletter on LinkedIn. Read the original here.

A couple weeks back I had the opportunity to deliver a keynote address to senior executives and board members of an Asia-based diversified manufacturing business. With senior leaders from sales, supply chain, finance, manufacturing and sourcing in attendance, my job was to present the potential for AI in Supply Chain and what it could mean for their business.

This manufacturer was clear that AI adoption offers radical, new opportunities. Their question wasn't "IF" AI could help, instead it was "HOW" and "WHERE" can AI help practically. I thought it would be a fun challenge to tackle this question.

I had so much fun putting the presentation together, I figured I'd share a few excerpts from it with my newsletter audience as well. If you are an executive or a leader looking for ways to augment your operations with AI, I hope these points help you.

As a clarification:

AI = Predictive AI + Prescriptive AI + Generative AI

1. The Purpose of your Supply Chain

The Supply Chain exists in the context of the business, its industry and its customers

Everything must start here. After two decades of running nearly every supply chain function, I know that the supply chains only exist in the context of the business, the industry it operates in and the value it delivers for customers.

Defining the purpose of your supply chain from first principles is a healthy exercise every executive team should undertake. In fact, if you haven't done so recently, I recommend going through this exercise to bring clarity to your team on why your supply chain exists. Do not fall back to consultants or industry/academic definitions. If you need a start, here is one I find useful -

To enable the company's strategic objectives by delivering its goods and services to customers in the most optimal manner to ensure long term success

2. Supply Chains are Not Linear

Supply Chains are a connected system of systems operating at varying frequencies with varying goals

Lets stop pretending they are. Modeling supply chains is difficult. They are complex, non-linear and dynamically connected global systems. That alone deserves respect. Invest in technology and systems that are capable of operating despite the presence of high complexity, variability, uncertainty and risks. Reject rigid systems that impose artificial predictability on your supply chains - it will cost you more to unwind artificial predictability than it will to embrace uncertainty.

3. Supply Chain ROI is not about replacing people

In my experience, the best run supply chains enable superior outcomes.

Best-in-class supply chain performance can drive : Revenue growth +2-3%, Operating Margin +50-100 bps, Excess Inventory -35% and Return on Capital +2-3X. vs average-performers

The inputs that enable this are extreme visibility to customers and suppliers, integrating customer and suppliers objectives into your operating model, improved demand anticipation, optimized inventory targets and placement, smart trade-offs for network design levers across fixed and moving assets, radically fast decision-making under uncertainty/constraints and ability to price and implement optionality/redundancy across the network. It starts to get obvious how technology is the only way to leverage scale across these inputs and deliver impact.

4. Enduring Problems in Supply Chain

But where should you focus your investments? I find this Jeff Bezos quote helpful

What is not going to change in the next 10 years? - Jeff Bezos

In the same spirit, I am defining here 5 of the most enduring problems in supply chain that will yield transformative results for you today and ten years hence.

Invest in AI to address the problems that will be worth solving today and 10 years from now

5. Connecting your Supply Chain

After talking to scores of supply chain leaders over the last few months, I am convinced that a huge opportunity exists in "Connecting the Supply Chain" to enable high visibility, low latency decision making end-to-end. Too many supply chains are operating broken, disconnected networks and are complaining about overwhelming complexity, demand variability, poor inventory performance, high lost sales, unseen risks and poor asset utilization.

Map your Supply Chain, Gather critical inputs and model the design of your network

The objective is to minimize the distortion in the demand signal from customers to suppliers while using smart investment in redundancies where needed to protect from uncertainty and variability. This is impossible to do without actually bringing in external signals and inviting crucial customer and suppliers into the design exercise. And this is precisely where AI radically improves your supply chain's decision-making speed by connecting vertically and horizontally.

Vertical = Deep modeling in areas like forecasting/modeling/optimization across millions of data points

Horizontal = Enabling decisions through risk-detection/simulations/workflows/trade-offs by cutting across vertical layers

6. Demand Planning is not just about Forecasting…

Invest in demand planning and AI/ML to understand your customers, what they value, how they interact with your product and services etc. Rigorously define demand clusters around these customer-product segments and watch it transform the operational culture of your teams. Your algorithms will also perform better as a result - for instance Customer ordering policy models, demand elasticity in response to price/promotions, inventory optimization and freight models and even critical components sourcing plans.

With this rich intelligence, demand forecasting and S&OP starts to look a lot more capable of making difficult decisions faster and with quantified options/trade-offs

An AI-Enabled Demand Planning and S&OP Process

7. Network Design should reflect the dynamics of each business

Most enterprise supply chains operate multiple brands, banners, product lines, channels etc. In fact, each can be large enough to be its own business. And yet, many teams design their supply chain networks with a one size fits all approach. Spend the time to be intentional about the end-to-end design to maximize the outcomes you seek in each business. A great heuristic for this is:

Work backwards from the customer requirements
Organize Critical inputs to feed network design (facility details, costs - opex, capex, lead times, products, capacities, suppliers etc)
Use AI powered simulations to compare strategies and designs
Select the strategy that maximizes output goals under uncertainty and constraints

8. Win with AI

As you evaluate the opportunity to win with AI, here are a few closing reminders -

"AI works great middle to middle, Humans work end-to-end" - Balaji Srinivasan. Upskill and charter your teams to inject value in defining clear inputs and goals for AI models and empower them with the decision-making authority to adapt AI-driven recommendations to formulate strategies and execution plans in the real-world.
Pick strategic problems that fit well with AI model capabilities. Don't use the wrong tool for the right job.
Don't spend time obsessing over collecting all the data upfront. Focus on getting narrow but deep data to get started and then add over time. Just start.
Invest in a culture that rewards algorithmic decision-making