Buying into our AI-powered procurement future

There is a global excitement around artificial intelligence (AI) and large language models (like OpenAI’s ChatGPT). But when your work is predicated on making high-dollar contracting decisions quickly—all while identifying and excising all bias in the process—being methodical is critical.

For procurement, technology disruption on the scale we are seeing with AI creates an innate tension—governments don’t like to rush. Addressing an emergent technology like generative AI isn’t easy, but many agencies are starting to do so.

In addition to announcing their “Center for Public Sector AI,” eRepublic has been keeping tabs on governments releasing policies and guidelines related to AI. Examples include the Biden administration’s executive order and AI.gov at the federal level and policies from many states including Kansas, Utah, Connecticut, and Rhode Island. San Jose, Calif.’s latest guidelines require employees to log every use of generative AI to ensure transparency.

Over time, we expect guidelines and recommendations to coalesce into a broad understanding of dos and don’ts. But here in the early days, AI and its public sector applications are essentially the Wild West. But there are a few leaders in our space pushing the envelope, testing methodologies and hypotheses, and setting out a path for responsible AI use in public procurement.

Leading the pack—Procurement experts exploring AI
I recently connected with two such trailblazers, Jared Ambrosier, chief procurement officer for the state of Michigan; and Paul Krivacka, executive director of legal and compliance for Tennessee’s central procurement office, to unravel the impact, risks and benefits associated with AI.

For Michigan, Ambrosier painted a picture of AI’s nascent promise and detailed some early or in-development use cases: “Within Michigan we focus on using AI to streamline processes, automate when possible, give better information to leadership to make decisions. In a lot of cases, I don’t know that staff even knows its AI at work. And I don’t know that it matters. They just know that it used to take three hours and now takes a click of a button.”

Ambrosier shared examples including the use of Sentiment Analysis powered by natural language processing to help decipher citizen feedback. This feedback is then used to train the agency, improve supplier performance and boost customer service in the state.

Other areas involved using predictive modeling to analyze past contracts and better forecast project timelines. By plugging requirements into their model, Michigan procurement can now provide more accurate timelines to agencies and suppliers—allowing them to level-set with those suppliers in advance of their work. Continuously feeding new contracts into their model only improves its predictive power.

In Tennessee, Krivacka has put together process improvement teams to investigate ways that AI technology can be implemented.

They conducted an experiment to see how effectively ChatGPT could write a scope of work and found that it was at least comparable to what a non-lawyer procurement professional could put together, and it was done in minutes. Krivacka sees limitations. There still need to be checks by human subject matter experts to ensure the output is viable and that the AI isn’t basing requirements on past procurement policy. He envisions the potential for a trial of this to see how the industry responds.

Krivacka’s team is also using AI in the certification and tracking of insurance policies, extracting expiration dates, insurance types and more to assess the accuracy of the details. He notes that these technologies act as a silent force, seamlessly automating tasks, reducing errors and providing consistent results.

Navigating risks: The fine balance of implementation
Of course, new technologies have inherent risks. Some of these risks are known and understood. Ambrosier highlighted concerns such as AI “hallucination” and biases within AI models. He notes the importance of government agencies beefing up AI expertise to better mitigate risk.

Krivacka emphasized the need for human oversight, particularly in the public sector, where adherence to laws is paramount. He notes that while the promise of efficiency gains is tantalizing, a cautious approach and addressing challenges head-on are crucial. At the same time, he recognized that the AI journey was just beginning for the state of Tennessee.

“You know, even a thousand-mile journey begins with the first step or two,” Krivacka said. “So that’s kind of where we are with some of our AI projects.”

The bright future: AI as the cornerstone of procurement
Both Ambrosier and Krivacka see a future where AI is integrated into everyday procurement activities. Building requisitions, drafting bids and analyzing contracts could all be streamlined through AI. However, they underscore the importance of understanding the limitations and complexities of generative AI. They emphasize that decisions on best value bids remain human-driven, reflecting the inherent subjectivity in procurement.

Both leaders expressed the need for collaboration. Governments must partner with vendors offering mature AI products for widespread implementation. Additionally, the sharing of ideas and applications of AI between governments, even with varying data inputs, fosters a culture of innovation and mutual growth.

One thing is for sure—expect that sessions on AI will fill rooms to capacity at every public procurement conference in 2024 (and beyond).

Integrating AI into the procurement tech stack
Procurement tech vendors are also incorporating AI. Here at Procurated, the leading platform for government supplier ratings and reviews, we are fully embracing this paradigm shift.

We are leveraging AI to parse through our database of more than 60,000 reviews. We know busy procurement professionals don’t have the time to read hundreds of reviews, so we added an AI-powered review summary, highlighting key themes present in reviews so buyers can get a sense for supplier performance in a matter of seconds.

In the future, we’ll be looking for even more opportunities to save procurement professionals time by leveraging AI. For example, fast-tracking the supplier discovery process by integrating AI into our supplier search. A procurement professional looking to buy a niche item would be assisted by an AI that can find relevant information and even make recommendations.

There are lots of applications within the world of vendor performance management as well. AI will help a contract manager identify at-risk suppliers on contract and prompt them to schedule a business review. In that business review, AI will suggest issues for resolution using themes found in Procurated reviews.

It’s said that innovation in private sector procurement often precedes adoption in government procurement. If this is as true in AI as it has been for other technologies, some exciting days are ahead for public sector buyers. Chicago-based G2 recently rolled out Monty, an AI assistant that provides software recommendations based directly on user requirements. I spoke with G2 co-founder Tim Handorf.

G2 was launched to help software buyers pick the right software based on the candid reviews of their peers. Like many websites, they soon launched a chat function, staffed by a G2 employee. But they saw that the questions asked were less about a specific software and were more about the business problems the buyer hoped to solve with that software. The questions required more information than any human could keep in their head. But then, OpenAI released ChatGPT.

“This really changed everything,” Handorf said. “We realized that Monty could interact with a human in a way they wanted to be interacted with to find the right software.” The chatbot began acting like an actual software consultant, trying to understand your needs, the dynamics in your organization, what your top priorities are for the software. Monty then creates a shortlist of the right software providers, along with the pros and cons of each. Monty also serves as a due diligence tool, helping buyers verify that the software can actually do what the company says it can. It does this by instantly scanning 2 million G2 reviews for answers to the questions the buyer wants to know.

And while we may not be ready to have AI bots replace evaluation committees today, it is not hard to see a day, perhaps soon, where sourcing teams use tools like Monty to identify the suppliers they want to solicit for a project.

Government procurement’s AI-driven future
The shared vision of procurement leaders like Ambrosier and Krivacka underscores a future where AI technologies, used methodically and with all due diligence, serve as the cornerstone of a more agile, responsive and citizen-centric government procurement landscape. 

Alongside procurement technology companies, the government procurement community is ushering in a new AI-powered era. One where governments are faster and more efficient at delivering the goods and services that help our constituent agencies and citizens every day.

If AI tools make the work of a procurement professional more efficient, they also make the work of a procurement columnist more efficient. This entire article was drafted by OpenAI, taking the transcripts of interviews I conducted in the last few weeks with Ambrosier, Krivacka and Handorf.

David Yarkin, the former chief procurement officer of Pennsylvania, is the founder and CEO of Procurated, the leading platform for supplier performance management powered by more than 60,000 supplier reviews directly from verified procurement professionals.

This article originally appeared in the December 2023 issue of Government Procurement.