The impact of AI on public procurement: Introducing a new series

If you’ve been at any event or meeting this year, we bet that at least one session has been on the promise and challenges of AI. The topic is fascinating even if there is little discussion of concrete applications of AI that are actually working in government.  

We recently published a blog post that provides a gentle introduction to AI in procurement, illustrating the thinking and work that needs to go into even relatively straightforward processes such as comparing green criteria for public contracts against real tender notices. Very intentional design and awareness of the limitations struck us as key considerations in the modelling. 

We felt it was important to go beyond the hype and have a grounded and evidenced conversation exploring the use cases for AI in procurement and the use case for procurement of AI. And we work with an amazing community of frontline reformers so we thought we would plan a community call and a series of blogs that cover these topics as we seek to make more sense of the promises and the challenges of AI. 

As ever, we like to bring a goal-driven perspective to public procurement. It starts with an articulation of the problem that we are trying to solve and then brings all the right actors around the table to explore how we can work together systematically to change the policies, practices, information and agency of actors to open data and tools can support that change. 

1. Procuring AI technology

First up, partners have asked us how to go about procuring AI technology, for example, an AI-based solution to address key public service delivery challenges. 

We have already identified lessons and best practices in purchasing IT & software and many of the best practices apply here too. 

• – An integrated, multifunctional team.
• – A focus on goals and a clear articulation of the problem that you are trying to solve. You should only procure what you need. AI then may, or may not, be a part of that need. 
• – Solution scoping and testing, including whether AI really, really is the most appropriate approach or worth the potential additional cost compared to a traditional solution. 
• – Understanding how data is used and managed and any related risks.
• – Procurement planning and market engagement, particularly as we are looking at a rapidly evolving market. 
• – Understanding pricing and cost management, particularly as much of AI is sold as a software. 
• – And finally, performance monitoring with a focus on accountability regarding the ethical and data management specifics. 

While there is already some useful general guidance on procuring AI (such as here by the United States’ GSA or here by PUBLIC and Paradigm), in our series, we’ll dive deeper into these questions and best practices, discuss some of the details, and share practical tools from the community. For example, our partners at ChileCompra will talk about how they developed their ethical framework for procuring AI technology, including templates for tender documents, and our friends at Connected by Data will discuss how do to better public participation throughout the process.  

2. Practical examples of how AI can improve procurement processes

Here we’ll look at how AI is being used by procurement agencies and elsewhere. We want to look beneath the high-level conference presentations and shiny surface to give you the nuts and bolts, highlighting some specific tools and implementations where there is compelling evidence that AI is helping to streamline procurement processes.

A recent study presented by Isabel Rosa, the Director of Finance, Studies, and Strategy at the Portuguese Public Procurement Regulator (IMPIC) at #TEDConf2024 explores the use of AI in procurement agencies in the EU. According to the responses by 26 countries, seven have ongoing pilots and implementations from analyzing data and monitoring red flags to text analysis and preparing tendering documentation. Elsewhere, Brazil has been using a tender analysis platform called ALICE.

We’ll also explore private sector services that rely on AI for private sector services use AI for data cleaning (e.g. Tendertrace), pulling summaries from procurement documents to facilitate finding relevant opportunities (e.g. Spend Network’s Open Opportunities) and preparing bid documents (e.g. Open Opps and Bidhive). 

3. AI improving data management and analysis

Finally, public procurement means thousands and thousands of documents. Not all too long ago, most of these were paper-based, requiring boxes and boxes to hand-deliver bid proposals to city halls. While we’ve made some strides in turning them digital, PDFs still capture most of the information on the procurement process. 

We have a keen interest and quite a bit of knowledge in using procurement data. The Open Contracting Data Standard, which we administrate, has provided a clear, global schema to structure data, powering up data analytics via business intelligence tools or calculating red flags.

When we’re exploring the use of AI in public procurement, use cases may include aiding digitization throughout the procurement process, conducting data analysis and indicators, or applying Large Language Models (LLM) to extract information (and meaning) from these documents. 

We’ll follow-up to our gentle introduction on green criteria in public procurement and also discuss the limitations of generative AI and how to mitigate bias when using machine learning. 

We hope that this series will help define the right questions to ask when exploring whether AI should be part of procurement solutions, whether the goal is improving the quality of the meals provided in schools, supporting local businesses, or delivering more affordable health care.