Skip to main content

AI in Procurement: Opportunities and Risks

Artificial Intelligence (AI) has become a significant disruptor in the business world in recent years. While AI has been incorporated into source-to-pay (S2P) technology for the past decade or more, what is new is the accessibility of general AI tools to the masses. However, businesses are grappling to understand how AI can and should be used in their respective organizations. Procurement is no different, and this article will examine how AI can be utilized in procurement and the necessary requirements for successful implementation.


What is Needed for Effective Use of AI in Procurement

The success of using AI in procurement depends on several requirements that need to be met for the technology to function effectively. Most importantly your organization should focus on training, data, and input.

Essentially, AI tools require systems training to be able to accurately identify patterns and predict outcomes. This training is typically done by exposing the AI tool to large amounts of data (aka information), allowing it to learn from past experiences and patterns. The quality of the training data and the methods used to train the AI tool are critical to the accuracy of its output.

AI tools require large amounts of data to learn from. For S2P organizations, this data must be both relevant and representative of the procurement process being analyzed. It should also be up to date, as using outdated data can lead to incorrect predictions and decisions. This means that businesses need to ensure that they have access to the right data and that the data is properly organized and cleansed.

AI tools require inputs and prompts in order to function properly. These inputs and prompts can vary from across different tools, but they are essential to ensuring that the AI is focusing on the right information and providing accurate output. Prompting will become a new skillset within procurement, as effective prompting is the difference between acceptable output and great output. Procurement professionals will need to learn how to structure queries and prompts in a way that allows the AI tool to provide the best possible output.

Overall, meeting these requirements is essential to the success of AI tools within procurement. If these requirements are not met, the AI tool may produce inaccurate results, leading to poor procurement decisions and outcomes. Therefore, businesses must take the time to ensure that their AI tools are properly trained, have access to the right data, and are being prompted in the right way. This will allow them to maximize the value of AI in procurement and gain a competitive advantage in the market.


Internal vs External AI: The Options Available to Procurement

There are two ways to utilize AI in procurement, internal and external. Understanding the differences between these two types of tools is essential to knowing when to use them and how they can benefit procurement processes.

External AI tools, such as ChatGPT and Bard, are designed to be used independently, or integrated into existing business processes through APIs. These tools are trained on the entirety of the internet up to a certain point in time (currently 2021). This means that they have access to a vast amount of data and can quickly identify patterns and make predictions based on that data.

On the other hand, internal AI tools are built into existing source-to-pay technologies that are currently licensed by the business. These tools are trained on the specific data of the company, meaning that they can provide highly customized output.

Internal AI tools are typically trained on cleansed data from the S2P user, and they can be built into the solution user interface or behind the scenes within business processes.

One key difference between using an external AI tool like ChatGPT and internal AI functionality built into a source-to-pay tool is the level of customization. External AI tools are trained on general data from the internet, which means that they may not be as effective at providing output that is specific and valuable to the organization, but they have a wide scope in terms of usability and applications. In contrast, internal AI tools are trained on specific data from the company, which means that they can provide highly customized output that is tailored to the company’s needs and requirements, but the scope of what they can do is limited.

Use Cases for External AI in Procurement

External AI tools can be used to automate many repetitive tasks that procurement faces on a daily basis. Some examples include:

  • Risk assessment: External AI in procurement can be used to assess supplier risk by analyzing data from various sources, such as financial statements, credit reports, and news articles. This can help procurement teams make more informed decisions when selecting suppliers and mitigate risks associated with supplier relationships.
  • Supplier identification and discovery: AI can be used to identify and discover new suppliers based on various criteria, such as spend category, geography, and diversity status. This can help procurement teams expand their supplier base and increase competition, which can lead to better pricing and quality.
  • Recommendation engines: AI can be used to provide recommendations for RFP questions, such as which suppliers to invite, which evaluation criteria to use, and which pricing models to consider. This can help procurement teams make better decisions and save time during the RFP process.


Risks and Limitations of External AI in Procurement

AI is a rapidly evolving technology, and there are many areas to consider when adopting the technology. Some areas to carefully consider before adopting external AI solutions in your source-to-pay process include:

  • Data privacy and security: Using external AI in procurement to analyze your suppliers, spend, contract, or other sensitive data means that you are sharing this information outside of your organization. Consider your data privacy policies, and the information that you are working with. If you would not be comfortable making this data publicly available, then do not use it in an external AI tool.
  • Bias: AI systems can be biased if they are trained on data that is itself biased. For example, an AI system used in sourcing may be biased towards certain suppliers based on historical data, even if those suppliers are not necessarily the best choice for a particular project. Remember, external AI tools do not use real-time data. Recent changes at a supplier may affect their overall risk score, but that would not be identified by AI.
  • Lack of transparency: AI systems can be difficult to understand and interpret, making it hard to determine how they arrived at a particular decision. This lack of transparency can make it difficult to identify errors or biases in the system.


Use Cases for Internal AI in Procurement

Internal AI tools are trained on your specific data and include a higher level of data privacy. This makes them ideal for automating more sensitive processes.

  • Predictive Analytics: AI can be used to perform predictive analytics to identify patterns and trends in purchasing behavior, allowing procurement teams to better anticipate demand and optimize inventory levels.
  • Supplier Performance Management: AI can be used to monitor supplier performance by analyzing data such as delivery times, quality, and stakeholder feedback. This can help identify potential issues and allow procurement teams to take corrective action.
  • Supplier Selection: AI can help in the supplier selection process by analyzing various factors such as performance history, pricing, and delivery times. By using machine learning algorithms, AI can predict the likelihood of supplier success, allowing procurement teams to make more informed decisions.
  • Contract Management: AI can help in the contract management process by automatically reviewing contracts and identifying key terms and conditions. It can also track contract performance and alert procurement teams when there are potential issues.
  • Invoice Processing: AI can help automate the invoice processing process by extracting key data from invoices and matching them with purchase orders. This can reduce the time and effort required to process invoices and reduce the risk of errors. Ai can also be used to automate non-PO invoice coding based on a user’s historical coding combined with details from the invoice.


Risks and Limitations of Internal AI in Procurement

While internal AI solutions offer more data security and are tailored to your business processes, there are still several areas that procurement organizations need to be aware of as they begin to adopt the Ai functionality built into their source-to-pay applications, including:

  • Data Quantity: In general, the more data an AI model has access to, the better it can learn and make predictions. While there is no fixed amount of data that is required for an AI model to be effective, it is important to understand that AI will need to be trained on your specific data, and evaluate if this will be effective for your organization.
  • Data Accuracy: The quality of the data is just as important as the quantity. Poor quality data can lead to inaccurate or biased models, even if the model has access to a large amount of data.


How to Adopt AI in Your Source-to-Pay Process

Adopting AI in a source-to-pay process can be a daunting task, but if done properly, it can yield significant benefits in terms of cost savings, efficiency, and accuracy. Here are some steps that businesses can take to successfully adopt AI in their source-to-pay process:


  1. Identify what AI capabilities are in your current tech stack: The first step in adopting AI is to assess your current technology stack and identify the AI capabilities that are already available. This could include features such as natural language processing, machine learning, and predictive analytics. It is important to understand the capabilities and limitations of the AI tools that you already have, as well as the potential areas where they could be improved.
  1. Make sure the AI is trained with correct data: Once you have identified the AI capabilities in your technology stack, it is important to make sure that the AI is trained with the correct data. This means that the AI algorithms must be fed relevant and up-to-date data that reflects your business processes and workflows. Without proper training, the AI cannot provide accurate or relevant results.
  1. Identify sample use cases for external AI: External AI, such as ChatGPT, can be used to perform tasks such as identifying information that would normally take hours of internet searches and data consolidation, finding recommendations for RFP questions, and identifying key attributes for a spend category. It is important to identify sample use cases that can be tested to determine whether external AI can provide value to your business.
  1. Run sample scenarios, both through AI and manually: Once you have identified the sample use cases, run them through the AI tools and compare the results to those obtained by manual methods. This will help you to determine whether the AI is providing accurate and useful results. It is also important to identify any gaps in the AI output and determine whether these are due to AI limitations or query structure.
  1. Start small and scale up: It is important to start small when adopting AI and scale up as you gain experience and confidence in the technology. Begin with a pilot project or proof of concept to test the AI in a controlled environment. This will help you to identify any issues and fine-tune the AI algorithms before deploying them on a larger scale.
  1. Train your staff: Adopting AI in a source-to-pay process will require a shift in how your staff works. They will need to learn new skills, such as how to effectively prompt the AI tools and how to interpret the results. It is important to provide training and support to help your staff adapt to the new technology.
  1. Continuously evaluate and improve: Adopting AI is not a one-time event. It is a continuous process that requires ongoing evaluation and improvement. Regularly assess the performance of the AI tools and identify areas where they can be improved. This could involve refining the training data, improving the algorithms, or adding new features.


Adopting AI in a source-to-pay process can be a complex undertaking, but if done properly, it can provide significant benefits to businesses. By following these steps, businesses can successfully adopt AI and use it to improve their procurement processes, reduce costs, and increase efficiency. While AI has become a hot topic recently, providers such as JAGGAER, Basware, and Zycus have been adding AI functionality to their source-to-pay processes for years. Velocity is releasing a video series with thought leaders from these organizations to discuss the state of AI in procurement today, and what the future holds.

Velocity has also been exploring various use cases for external AI tools in areas like Soucing, supplier management, and risk identification. If you have questions on these topics, feel free to reach out, we are always excited to talk!