Businesses constantly seek innovative solutions for their supply chains to streamline operations, reduce costs, and enhance customer experience. Generative Artificial Intelligence (Gen AI) offers promises to fulfill those exact wishes. But can it do just that? I’m going to answer that question for you as well as offer some practical insights into how you should be implementing Generative AI in your supply chain.
What is Generative AI?
Before we dive into its impact, let’s briefly discuss what exactly is Generative AI. Now, there’s a lot of definitions out there, but here’s how I define it. It’s a system that can create new patterns based on the interpretations of learned patterns. It also comes down to three main components.
Foundational Model
This is where the training/learning takes place, where you’re teaching the AI how to look at things and look at input.
Large Language Model (LLM)
This model is trained on vast amounts of text, can interpret what you’re asking of it, and can put a response in words that you can understand.
Natural Language Processing Model/Chatbot
This is the most visible portion of current Generative AI implementations, like ChatGPT. This model component is generally a conversational chatbot leveraging the LLM to create content. It was first trained on a foundational model and then fine-tuned with human feedback.
Generative AI is first trained on a foundational model and then fine-tuned with human feedback and additional data.
Now, none of these are really new technologies. They’re just available more widely today because of the lowered cost of processing power, rise in cloud technology availability, and ability to push it out in more cost-effective manners. 10- 15 years ago, it was far too expensive. As prices have come down and processors have become more powerful (think Moore’s Law), it’s allowed Generative AI to become more easily accessible.
When you talk to Generative AI it feels like it understands you. Spoiler alert, it doesn’t. Its responses are based on data it has consumed and a resultant powerful prediction mechanism. Don’t think that Gen AI “understands”; it doesn’t. This has significant impact when it comes to focusing a Generative AI into a particular industry, like logistics. Generative AI can’t apply specific nuances of an individual industry unless it has been trained on them, and even then, there are other obstacles to Gen AI being useful outside of targeted use cases. This is a key point in how we think of getting Generative AI to effectively impact supply chains.
Public vs Private Generative AI
When you use Generative AI, there’s a big thing you should be concerned about – public vs. private Generative AI.
Public AI
We talked about ChatGPT. That is one example of a public version of Generative AI. Anything you input in there is being reused for training purposes. Any of the prompts you put in, it’s using that to train itself and feed back into that foundational model for others to use. That becomes more or less public domain once you put it in there. You must think about it that way. It’s like posting something on the internet – it’s there forever.
Public AI is great to use for idea generation, but only put in anonymized data, that’s if you put in any data at all.
I also cannot stress this enough – fact check what it generates. Generative AI will make stuff up – the industry term is hallucinations. It’s the worst people pleaser you can imagine. It will give you what you ask for, whether that’s true or not. There are advances in the works to help “fact check” the tool, but you must remember whatever Generative AI does today is opaque. We don’t know where it’s pulling this information from. You’ve got to go in and check its responses.
Private AI
There are opportunities to use private Generative AI. Amazon Bedrock and Microsoft’s OpenAI implementation have private options to use such that the prompts and any data you put in are not used for training the core foundational model. But you must make sure you’re on the right (private) version of the tool.
The second thing is that if you feed it data, you must make sure that data is right because Generative AI has no clue. It’s assuming what you’re telling it is true. Though it does have the ability to fact check your data, it will only do so if you ask – and ask correctly. You must have clean data and validate the results that come out of it.
Again, private AI is great for idea generation but not so great yet as a fully automated toolset that can be used to drive business forward. It’s getting there, but not there yet. Private AI options will get there faster for certain topics, industries, and businesses than public AI.
Generative AI Latest Features
There are some new features that have come out for Generative AI recently. One of those is the ability to relate images and text. You can put an image up, and it can write a story about that image. You can input text, and it can create an image, like Dal-E.
There’s also code creation and injection. It can generate code, and when I say code, I mean programming. OpenAI has functionality where you can do code injection. Github has its copilot technology where it can write code alongside you based on a plain English prompt.
Another feature is enables tools to call digital actions. So, I can go from a Generative AI tool and call out to create an action in another application. This is where we shift over into physical actions that we find in Logistics 4.0. For example, a robot in a warehouse that results in a physical action, or it could just be something as small as, if this event takes place, then make this phone call or change this data.
The Impact of Generative AI in Supply Chain
What’s the logistics impact of Gen AI? Biggest one is data analysis.
Data Analysis
Generative AI is great at taking data in and putting it into a format that can be easily used to extract meaning. Think about this from a summarization perspective. You’ve got somebody on a call with a customer. Based on their phone number, Generative AI identifies them and comes up with a prompt that tells you the last 30 days of that customer’s history (products, contacts, issues, etc.) so you can have a much more informed conversation with them.
Trending is another great one to look at across data sets. As you put the data sets in and tell Generative AI to relate these two things, you can start to see trends against each other. It’s interesting what you can do with those pieces.
Then there’s assistive information delivery. At points in time where actions need to be taken, information can be delivered as a prompt to the person performing the action. This can happen in real-time with some tools, or it can happen in the background.
These are all pieces that you can put into place today.
Predictive Analysis
A primary use case within logistics is predictive data analysis.
Routing is one but that’s going to take some work to be effective. That’s because it needs specific knowledge sets around the model and those knowledge sets are not available to most of generic models. There isn’t a specific logistics model yet that understands all the nuances to the different modes and can bring in the data sets that are necessary to truly understand routing. There are tools you can use for routing right now, but they’re not Generative AI tools.
Exception identification is another one. Again, this is a large data problem and there are a lot of factors that come into it. Any supply chain is essentially a complex system. Being able to identify those exceptions earlier is going to be something Generative AI will help with when it comes to logistics and supply chains.
Hyper-personalization is a great one. That’s getting information into a format that speaks to a particular person, entity, or customer. You can hyper-personalize a piece of communication to them based on the information you know to make sure it’s something you can use to improve service delivery.
These benefits of Generative AI are really where we’re going to be able to improve service delivery across supply chains as it evolves.
How to Use Generative AI in Supply Chains Today
You’re likely wondering, how can I use it today? I’m happy to report that there are several different ways you can being using it.
Improving Customer Experience
One of those ways is improving customer experience. Using a private AI, you can feed it customer data. Then, when you’re talking with customers you can bring up insightful information. It can bring up things like a notice of a customer’s birthday and save you time from having to go and look up specific pieces of information.
If you’re feeding the data into a system, you can use these private models to intake that data and provide relevant information back to someone that’s interacting with a customer. This is where I like to remind everyone that Generative AI isn’t a replacement for people in customer service but is great for putting information into a digestible format to support people in their roles.
Summarization
One of the big things is summarization. Summarization is an awesome feature of Generative AI. It’s ability for it to take data and create a summary is a tool that you can use today. Getting summaries of articles, getting summaries of customer history, getting summaries of a lot of different types of data is something the current application can do and do it well. You still have to fact check it, though!
Personalization
The other big piece is the creation of personalization. Getting that data in and creating a personalized email with some information about who you’re sending to is something that Generative AI can do and at scale. There are tools you can use to do that and it’s something that you can put into place today and try out.
Think About Your Data
I would encourage everyone to get started and try things. It’s not going to be perfect out of the gate. I can tell you that right now.
One of the key factors that you must think about as we move into this Generative AI world, is data. Good, clean data is at the core of successful AI use. You must think about how clean your data is, what data you’re going to send to an AI, are you public or private, all those pieces.
Accuracy and consistency are the core. You must establish your data governance. It’s great if you have clean data, but if you start feeding bad data in again, all you’re going to do is skew your results coming out. Now is the time to start evaluating your data sets. Understand where you have good data, how your data is inputted, where you can start, and where you need some work. Then, get your data in line and start to think about how you can enrich that or what can you do with it.
Before You Get Started…
There are other pieces to consider before you fully dive into Generative AI.
There’s an ethical component here. There’s change that’s coming. People’s jobs are going to be affected. You must think about how this impacts your team. What trainings do you need to provide? How do you get people brought along for this ride instead of leaving them behind? What’s the impact to your customers? If you’re relying on this as an assistive technology, how are you making sure that you’re still servicing your customer effectively?
One great thing about automation is that it makes things happen in the background. The bad thing about automation is things happen in the background. Unless you understand what’s happening, can react to that, and leverage it to effectively service your customers, you’ll end up doing them a disservice. That’s true for any automation and Generative AI is no exception.
Then there’s change management. How do you prepare your people? How do you get to the point where you have prompt engineers?
Each AI works differently. If you give the same prompt to different AI applications, you’ll get different answers. So, how do you understand how to engineer those prompts? How do you identify this skill? These are critical questions you’ll need to answer over the next few years as we move into this Generative AI world.
FAQs
Are There Logistics Focused Tools in the Market Today Leveraging Gen AI?
There are. I would say that the use of Gen AI in logistics tools is still a little new. Tools like Parade, which is a very specific toolset around carrier relationships is leveraging some Gen AI capabilities. Most Gen AI is relatively generic. Tools like Observe.ai or Freshworks are available today with some Gen AI capabilities.
How Do You Measure Success When Leveraging Generative AI Initially?
We always look at success when we look at automation as reclaimed time. What we mean by that is, are we giving someone a technology that assists to a point where we’re able to redirect them to more strategic activity. The other one is really asking the people who are impacted by it. You need to get feedback so you can understand if it’s being helpful or harmful.
When Should I Start Using Generative AI?
Now! Encourage its use, though with appropriate guardrails and guidelines. It’s going to be a disruptor. Gen AI is not a disruptor quite yet, but it’s getting there, fast. The widespread availability and use will accelerate its impact on many industries and roles.
What’s an Example of How Trinity Logistics is using Generative AI today?
We’re using it in a couple of different ways. One tool we use is Observe.ai, which is a natural language processing tool. We use it to look at our interactions with customers, look at sentiments, look at certain events, look at what type of interaction was this, and that uses that Gen AI to provide us with that information. Then we have a Development Team, so we’ve started working with some of the code copilot, so assistive coding technology.
