A business case for using AI in your supply chain

3 steps to help jump-start your journey

By | 3 minute read | January 29, 2020

One of the things we often hear at events and in meeting with supply chain leaders is that most believe Artificial Intelligence can drive value for their supply chain, but they need to make the business case for it.

After providing guidance to many supply chain clients across industries, we’ve found that these three steps can help begin your journey – regardless of your starting point.

1. Align your goals and use cases

Every supply chain project should start with a business value proposition. The same is true for AI-related projects. Define a specific problem that really matters to the business and the outcome you want to achieve, and make sure it’s a good fit for AI. Since this is your initial foray into AI, pick a relatively easy use case. This allows the team to become comfortable with the fact that an AI-based solution isn’t all about coding. It’s about designing a model that involves a series of decisions mapped to user workflow with results based on data. Here is a sample use case we have seen clients tackle:

Predicting return propensity and optimizing call center operations. Most returns are from “genuine returners”, but “serial returners” are a fact of life. Insights about the returner helps call center agents quickly determine the next best action to process the return and reduce call handling time. A serial returner’s behavior is not indicative of dissatisfaction. So, offering a discount or spending time determining if this customer wants a different color or size only prolongs the call and increases the cost of goods sold. To reduce call handling time, AI-enabled solutions use a natural language interface. Transactions are faster and simpler to execute because agents work intuitively by asking questions, receiving answers and executing transactions in the language they use every day. With customer context and natural language search technologies, we’ve seen clients reduce call handling time by up to 10%.

2. Explore your data landscape and assess the gaps

Regardless of the use case, the AI models that work behind the scenes to drive outcomes require data. So, the next step is to identify the data sources, data types and currency of data you have and make sure it is sufficient. For example, to understand return propensity, you need data on factors like the time of purchase, the customer’s transaction history and the item itself.

If you don’t have all the data you need, don’t give up – there could be workarounds. Approximations may be adequate, or you may be able to substitute industry benchmarks. Also, research exists that identifies and prioritizes the appropriate attributes for specific use cases. You may not need data for every single attribute. If you have data to support the key attributes for your use case, you can move forward to the next step. Tap into subject matter experts to help you work through this quickly and confidently.

3. Consider a proof of concept

With data in hand, you’re now ready to create AI models that reflect the decision-making process to solve the business problem defined in your use case. You’ll also need to make a choice of the right AI tool for the business need and user expectations. And that means validating the choice by executing a proof of concept. The proof of concept will help ensure you have the right data and the right tool for the decisions a business user would want to make with the help of augmented intelligence.

Decisions may include, “If this is a serial returner, proceed directly to processing the return.” As you start running the models in your prototype environment using your data, you’ll start seeing results and can do some fine tuning. Typically, after 30 days of running the prototype you can validate the business value, deliver a report to the business and begin enhancing the solution based on user experience feedback.

The key to getting started with AI in your supply chain is to define what matters most to the business or the users you’re trying to help, and then prove the value. Typically, within 90 days from start to finish you can deliver meaningful results and have a clear path forward.

Interested in what we can do for you? Take a few minutes to explore our IBM Supply Chain Fast Start program to learn more about how we help our clients create a real AI prototype using their own data in just 4-6 weeks. If you’d like to speak with us directly, schedule a consultation – we’d love to meet with you.