Is ChatGPT going to transform supply chain management?

Many chatbots have come and gone over the years. Alicebot, Jabberwacky, and Kuki are just a select few that you may remember.

They have all won the Loebner Prize, a contest for the most human-like artificial intelligence application. None of them has had the same impact as ChatGPT. More than a million people used ChatGPT within the first five days after its release. This is because of its public accessibility and instant, in-depth responses.

Leaders in the tech sector, such as Bill Gates and Elon Musk have said that “ChatGPT will change our world”. They have also stated that ''ChatGPT is shockingly good.'' Is it good enough yet, to pave the way for a new generation of intelligent applications for industrial automation?

What is ChatGPT?

ChatGPT is an artificial intelligence application that can mimic human responses to text-based inputs. It builds upon a family of Generative Pre-trained Transformer (GPT) language models developed by OpenAI.

These language models are ‘pre-trained’ on a large volume of structured text. This means that they are not focused on a particular topic. Thus, they can convincingly cover a diverse range of subjects.

What are the potential supply chain applications of ChatGPT?

We are just beginning to explore the potential applications of artificial intelligence such as ChatGPT and its successors. Most of its uses revolve around communication, as this is where its linguistic abilities shine the most. However, it also has some interesting capabilities that are not what you might expect from a chatbot.

1. Support and communication

ChatGPT in the supply chain could answer product, order, and account-related queries for end customers. This can be accomplished via email or a messaging widget.

For instance, it can provide customers price estimates for the goods or services they're interested in purchasing. As well as resolve a delayed shipping complaint, or provide technical support based on a product manual. It could even make product recommendations based on a customer’s past buying history.

2. Predictive analytics and demand forecasting

A particular strength of ChatGPT’s large language models is their ability to analyse huge quantities of data. It could apply this capability to analyse historical order data and use it to predict future trends and patterns of demand. This functionality could further support production and inventory management.

3. Process automation

The ChatGPT platform has the potential to automate a variety of supply chain management processes. It could generate documents, such as invoices, contracts, and tender requests. It could also automate the communications aspects of fulfilling production requests, managing inventory, and tracking deliveries.

4. Training and Education

ChatGPT could train and educate employees involved in supply chain management. It could teach an employer’s policies and procedures to a new hire, answering questions along the way. It could also assist employees with the creation of educational material for hands-on product training and generate scripts for webinars.

5. Code generation

A little-known application of ChatGPT’s text generation capability is its ability to write in other human languages. It can also write in many computer languages - that’s right, ChatGPT can code. Request that it compose a quick programme, a code snippet, or a function for you. It will perform adequately for an entry-level position.

How could ChatGPT be better suited to supply chain management?

ChatGPT has a great deal of potential, but it is not flawless and has some design limitations. Because of this, it could be challenging to employ in a range of supply chain applications:

1. Knowledge of current events

The data collection that was used to train ChatGPT is large, yet it has its limits. It does not cover events that occurred relatively recently or that are happening right now. Events like the weather, the current position of goods, or the impact of recent geopolitical difficulties cannot be taken into account.

This limitation does not apply to other emerging chatbots, such as Google Bard. Bard can find information using a Google search, and use the result to inform its responses.

2. Real-time control

The architecture of ChatGPT does not incorporate real-time control and closed-loop decision-making in any way. It has a slow response time and is unable to exert precise control over a physical system. It is unable to monitor any supply chain metrics, take any actions, or trigger any external events.

3. Response accuracy

Even though its developers have put in a great deal of effort, ChatGPT is not always accurate. Sometimes it will respond with answers that are convincing but incorrect. Whilst other times, it will respond with answers that are biassed.

It is possible to use deception and manipulation to get ChatGPT to give the incorrect answer. It is necessary for it to be dependable in order for it to play a significant part in supply chain management.

There is more to artificial intelligence than large language models and natural language processing. Other developments in the field are also benefiting supply chains around the world.