Generative AI in procurement: A fast-moving landscape

With the release of ChatGPT in late 2022, the world's view on the speed at which AI will disrupt industries changed overnight. It captured everyone's attention because (a) it's a significant change in capability, (b) it can solve a huge range of problems, and (c) the conversational interaction model unveils the AI process at work. These factors make generative AI (gen AI) significantly different from the AI that has been more subtly changing our lives in applications such as speech recognition, navigation, and social media.

This release wasn't just a major shift for the public but also for industry experts. McKinsey & Company's prediction of when AI would attain human-level capabilities accelerated by over 10 years after ChatGPT's launch. This sudden pivot in expectations is challenging all industries and business functions to reassess the opportunities and risks from AI at an unprecedented rate. Some functions have a tremendous opportunity to streamline their work and unlock new sources of creativity. In contrast, companies whose business models are based on prior versions of AI could see their entire business disrupted.

In addition to the typical challenges associated with a new technology, generative AI also poses questions in terms of ethics, accuracy, and transparency. The flood of gen-AI technologies such as Bard, Llama, and a host of purpose-specific tools means the combination of rapid change and opaque decision-making presents unique challenges for companies.

We can look at the impact of AI on procurement from two perspectives:

1. The impact on how procurement operates. Generative AI will have a significant impact on how procurement functions since much of their activities include creating and analyzing documents such as request for proposals (RFPs) and contracts. A recent McKinsey & Company study estimated that just under 20% of the procurement function could be eliminated by generative AI – the third most impacted business function behind software engineering and customer operations
2. The impact on the supply base. Procurement teams will need to understand and manage the opportunities and risks as suppliers embed AI into their offerings. For instance, if a supplier of application development services uses generative AI to write code, it could be more cost-effective and better quality. But it could also result in erroneous code if not managed properly. Procurement will be on the front line in managing opportunities and risks of these developments

In terms of the impact on procurement execution, there are a few types of use cases that will rapidly become standard. These use cases can be grouped based on some common underlying needs:

• Insights. Effective procurement is built on insights. These could be from internally generated data, such as supplier performance, or from external data, such as benchmarks, market insights, or supplier discovery. Generative AI can surface these insights and present them in easy-to-digest form in response to simple user queries
• Document creation. Managing the procurement function requires generating all types of documents. The procurement process itself can be thought of as a series of documents being created, culminating in the ultimate document – the contract. Along the way, procurement teams generate numerous documents: requests for information (RFIs), RFPs, supplier communications, nondisclosure agreements (NDAs), statements of work (SOWs), master service agreements (MSAs), and various types of contracts. Generative AI specializes in this type of document creation and can already generate credible RFPs and SOWs. Contract creation will be one of the more exciting applications of generative AI for procurement teams. However, since each word in a contract can have important implications, relying on generative AI to do this will require additional advances in the ability to merge client-specific clauses with more proactively controlled AI content
• Summarizing documents. Another time-consuming activity for procurement is analyzing supplier responses during a competitive bidding process. Analyzing prices is just one part of this, and the qualitative analysis can be more problematic. For complex purchases, responses can be dozens of pages long, describing the supplier's qualifications, services, specifications, and how they drive value for your company. Generative AI can digest and summarize this text to create a succinct executive summary that allows the sourcing team to compare bids and identify areas where they should do a closer analysis of the actual responses
• Interactions. The procurement process necessitates many interactions between the procurement team, internal stakeholders, and suppliers. Teams can use generative AI to create human-like conversations with these stakeholders, the most obvious example being with the procurement or accounts payable help desk. Say a user has a question about a procurement policy. Instead of having a chatbot point them to the policy in question or providing a canned response, generative AI can respond directly to their question using the policy as input for a tailored response

These use cases highlight how procurement is ripe for transformation from generative AI. But the speed at which they're adopted will be largely dependent on a team's ability to tune their AI models using proprietary company data.

As well as assessing how generative AI will impact their own function, procurement professionals also need to think about how it will affect their supply base and the appropriate steps and controls to manage the technology. This challenge isn't new, but the rapid evolution of ChatGPT amplified the problem. There are several factors procurement teams should consider:

• Segmentation of the supply base. The opportunities and risks from generative AI will vary by supplier and the product or service being purchased. For example, software development shows great potential but also some risk, but if a commodity chemical supplier uses generative AI to optimize its internal processes, the risk and opportunity are lower
• Understand your company's gen AI policy. Generative AI is highlighting some challenges that have always been present with AI. But since some systems have been trained on "all" the data available on the internet, additional challenges have emerged:
  • Bias/accuracy. Generative AI will replicate what it's found across the data sources, including any inaccuracies and biases
  • Transparency. Fundamentally, machine-learning AI systems such as ChatGPT are sophisticated black boxes. It isn't possible to explain why a result was generated. This presents challenges when trying to establish accountability or liability for an event
  • Confidentiality. Many generative AI models also learn from the prompts entered by users. This new information becomes part of the corpus of data that can be used to generate results for other users and companies. Using a private cloud environment will reduce this risk, but it's a necessary consideration

These factors are just a sample of the risks and challenges that procurement leaders will need to address, so it's vital that you understand your suppliers' approaches toward managing AI responsibly.

Given the speed of evolution and the significant potential for this technology, procurement leaders should take the following steps to maximize results and minimize the risk:

• Identify an AI lead on your team. Since the market is moving so quickly, it's essential to nominate someone to keep track of developments and the implications for procurement. You won't be able to rely on a corporate IT team for this insight
• Establish an AI code of conduct for suppliers. Just as many organizations now use a supplier code of conduct, procurement teams will need to incorporate AI-related dimensions into supplier governance. This can be incorporated into RFPs as requirements or as questions to assess suppliers during the sourcing process
• Influence technology firms to be responsible with AI. Much has been written about possible doomsday scenarios for generative AI. While these concerns may be overinflated, until the technology industry or governments establish protocols, tech companies will respond to the market. Procurement teams have a significant opportunity to influence these firms to use generative AI responsibly
• Investigate the opportunities for your function. Generative AI applications for procurement will continue to evolve. While there is no need to be among the earliest adopters, this technology will mature quickly, so planning for the implications and being prepared to act when the time is right will pay ongoing dividends

To successfully get the most out of this technology, we recommend that all business teams:

• Identify where AI can create the most business value
• Embed AI into workflows and rethink processes
• Ensure strong data management and governance
• Partner with internal IT or external providers to scale the technology and operating models
• Nurture new skills and roles for employees

Generative AI will have a significant impact on procurement, and the capabilities in the market will continue to evolve quickly. Procurement teams that keep pace with these developments will quickly evolve their function and be confident that suppliers are driving value from this new technology in a responsible manner.