AI as Your Procurement Teammate: Jennifer Lyons (MA) on Human-AI Collaborations

AI is showing up in procurement workflows whether we invite it or not, so we wanted a grounded conversation about what actually works and what can backfire. I’m joined by Jennifer Lyons, Deputy General Counsel at the Operational Services Division for the Commonwealth of Massachusetts, to talk practical AI in public procurement with real examples from the field, not theory.

We dig into where AI delivers immediate value, like speeding up repeatable administrative tasks, summarizing complex RFP and RFR documents, and generating better options during policy development. We also talk about how suppliers are using AI to quickly decide whether to respond to a solicitation, and what that means for clarity and competition.

We explore how AI can help maintain continuity across supplier issues, surface patterns in supplier performance, and support risk awareness. Jennifer also explains why Massachusetts built tools inside secure environments, including a procurement chatbot used for onboarding and help desk support, plus a contract editor nicknamed 1L that accelerates first-pass redlining. Throughout, we keep coming back to the non-negotiables: transparency, fairness, bias awareness, and the reality that humans remain accountable for accuracy and final decisions.

If you’re building an AI policy, planning a pilot, or trying to choose your first “quick win,” this conversation will help you separate safe productivity gains from high-risk automation. 

Follow & subscribe to stay up-to-date on NASPO!
naspo.org | Pulse Blog | LinkedIn | Youtube | Facebook 

Transcript

Copy Transcript Copied

Welcome And Guest Introduction

Julia McIlroy 0:06

Hi everyone and welcome to NASPO's Pulse, the podcast that focuses on current topics in public procurement. I'm your host, Julia McIlroy. Today's guest is Jennifer Lyons, Deputy General Counsel in the Operational Services Division for the Commonwealth of Massachusetts. We'll discuss all things AI and practical use cases for public procurement. Hi Jen, welcome to Pulse. Hi, Julia. Thank you so much for having me. Excited to be here. I'm so glad you're here.

Where AI Delivers Real Value

Julia McIlroy 0:35

So AI, let's start with the big picture. Where are you seeing AI deliver the most practical value in public procurement today?

Jennifer Lyons 0:44

Yeah, so I mean I think we're seeing a lot of practical procurement across many different realms. Everything from obviously those uh repeatable administrative tasks that we're trying to tackle and make a little bit more consistent and easier for people on a regular basis, but also uh, you know, bigger things such as policy development, where, you know, in one case we were setting up a brand new grocery contract to try to purchase groceries directly from farmers. And while at first we were trying to define locality by within a given amount of miles from the Statehouse in Boston, and then we realized this is not how school purchasers and public purchasers are actually looking when they're doing their menu. And that's really something that we found by generating different options of you know, what are other states doing, using that deep research functionality, finding out, and then also using that to really take a quick dive into what programs are all of our buyers trying to navigate and what categories would it be easier? So we ended up instead of defining local as 30 miles away or 50 miles away, we just find, you know, we set it up so there's Massachusetts farms, there's regional New England farms, there's New York, you know. Um, and so this is how a lot of the schools were being reimbursed. And that was really something that um I think we may have gotten to on our own, but I think we got there a lot faster by being able to generate the options. So I for me that's where I'm seeing these kind of immediate first wins is the everyday repeatable tasks of just summarizing or comparing, but also really being able to generate more options and think of the assistant as kind of one of the smarter people on your team, but just like a smart person on your team, you're not gonna just run with whatever they tell you, you're gonna um use it to brainstorm. So I I find it immensely immensely helpful for that.

Julia McIlroy 2:37

That's a great way to look at it. Another member of your team that's really bright. But at the same time, you have to look at what they're doing, right? And I love that the farm to table for K-12. That's fantastic.

Drafting Better Solicitations With AI

Julia McIlroy 2:51

So, what are some ways AI can help with the creation of solicitations, such as drafting requirements, developing evaluation criteria, or identifying gaps?

Jennifer Lyons 3:00

Yeah, so uh there are so many different ways. This idea that it can amplify intelligence. Um, so you know, we can come up with better questions faster, we can do clearer comparisons better, we can have uh more consistency in the types of reviews and how re what we review for, um, and again, more creative options. And I think in the end, this saves us uh the time for where humans really need to be in the loop, right? I think um it's become much easier to summarize really complex technology projects so that even vendors, if they're going online and they want to see if this is something they even want to bid on, they don't necessarily want to start digging into a 30, 40 page RFR right away.

Julia McIlroy 3:44

Yeah, and as you said, another team member that can amplify the work, do things faster, better, but not a replacement, another team mate, which is great. I know I was spoken speaking to an e-procurement supplier who said that they receive RFPs from everyone from higher ed, uh, state, local, K-12, etc. And so they're using AI to quickly review the RFP and to decide if they want to respond or not to that particular RFP because they they take a lot of uh work and money, right? A lot of resources. So that's uh an excellent use right there.

Contract Management And Risk Signals

Julia McIlroy 4:22

Contract management is another area with enormous potential. How could AI help agencies monitor contracts, identify important dates and obligations, analyze supplier performance, or even flag potential risks?

Jennifer Lyons 4:35

In Massachusetts, we've established uh what we call the innovation group, and this is made up of members of every unit within the agency, and we are doing exactly that. I would say I don't have a single answer one for that yet. Um again, I think there's probably numerous different solutions, anything from building your own uh you know, kind of contract library solution to maybe if you have Chat GPT or Gemini accounts within your agency, you start setting up different projects for different contracts, which then can compare against different vendor proposals, or if you can start logging in whenever you get, you know, a complaint about a certain vendor on a certain contract, you know, it's very good at being able to maintain context around one thing.

Julia McIlroy 5:22

Yeah, I think that's a great point that you really have that continuity because it can save all this information, right? And reference it again. Absolutely.

Jennifer Lyons 5:32

So we can remember, you know, if you set up a project for a certain IT contract and you have given it the list of vendors that are on contract, and somebody calls in and says, so-and-so has been non-responsive, or you know, the tech technology is broken and we can't get um a service person out, you start logging that, right? And then now that it's you know, and so and there are different ways to do that manually. Um, you know, I'm sure people are already tracking things like that with Excel and everything, but to your point, AI can just amplify that by setting up notifications, by warning you if it is in this kind of you know, giant mass of 1400 vendors that you're trying to get feedback from or pull in from your procreated event um you know, uh account to make sense of. Um, there are absolutely you know different ways that you can set that up to help.

What To Automate And Avoid

Julia McIlroy 6:21

So what repetitive or administrative tasks should procurement professionals consider automating first? And what are some that they should maybe be a little more cautious about automating?

Jennifer Lyons 6:32

Yeah, I would say, you know, again, uh I would start the the lawyer in me wants to caution everybody to start with anything public-facing. So any kind of public-facing data that needs to be constantly updated or needs to be, you know, is um is a manual task that's repetitive, that's um so I'm thinking in terms of we are on a constant like regular basis updating our contract user guides. Well, even if it's just for uh vendor contact information, I would say at least once a week there is something there that needs to get updated. The flip side of that coin is to be incredibly cautious about any sort of regulated data. So if you have health information or quarry information or tax information, social care, any of that data that really comes with a heightened risk. Um, that a lot of the models, even if you do, even if you do work in a jurisdiction that has negotiated a third party such as ChatGPT or Reclaude or Gemini, and it doesn't necessarily feed back, even in Massachusetts, we would tell you to still keep out that confidential and sensitive information, even though we know it's not getting fed back, it's still, you know, there's still risk there. Um, and we try to we try to ridicule mitigate that risk as much as possible. So that being said, I think there's also a place for um, I think you know, we've teamed up with Northeastern Burns Center for Social Change in the past to build. We have uh an AI procurement chatbot, and then we also have a contract um editor. So both of those live in Massachusetts only environments, so I'm not as concerned about uh regulated data getting in there because you need to actually go through our firewall to get there.

Julia McIlroy 8:13

That's some great advice. And I know that Massachusetts is leading the way with your chat box, etc. You are the golden standard for procurement, for public procurement.

Accuracy And Human Accountability

Julia McIlroy 8:25

So one of the biggest concerns around AI is accuracy. If an AI tool produces an incorrect answer, recommendation, or analysis, who is ultimately accountable? How do we keep humans firmly in the loop?

Jennifer Lyons 8:39

Yeah, and I I think this is again just you know, really thinking through where machines are best and where human judgment is best. Um so absolutely, we're always worried about AI accuracy. And but I do remind people that you know human judgment accuracy isn't at 100% either. So, you know, we just like we need to uh you know double check our our intern's writing, we also need to double check the outputs that we're being given from AI. And again, keeping in mind that AI is great at those tasks that they're repeatable, they're mechanical, they're comparative, they're organizational or they're first pass type of things. But humans, we want them making the legal judgment. We want them making final decisions, ethical judgment, you know, maintaining those relationships, making key strategy decisions for negotiation, you know. Um the the the technology can help us be better prepared, but it can't make those um certain decisions, you know, that for trust and accountability that um that won't replace humans.

Julia McIlroy 9:44

So as we've discussed, it's another teammate, right? Right, right, exactly. And I like to think of it as another tool in our toolbox, meaning it doesn't matter if I used an encyclopedia or AI to create content.

Jennifer Lyons 10:00

Yeah.

Julia McIlroy 10:00

I'm still responsible for that content. Exactly. And for its accuracy, etc., right?

Jennifer Lyons 10:05

Exactly. We tend to have this in our uh negotiations with vendors all the time, you know, who don't want to provide us any warranties of this. And I'm saying, well, you I know you can't warrant the accuracy of the technology. What I need you to do is warrant the accuracy of the output you're giving me.

Julia McIlroy 10:22

A hundred percent.

Transparency Fairness And Bias Risks

Julia McIlroy 10:24

Public procurement is built on transparency, fairness, competition, and public trust. How do agencies make sure AI use strengthens those principles rather than creating new risks or an intended bias?

Jennifer Lyons 10:38

Right. And I think uh again, it the goal should never be to let the AI decide. Um, the goal is to let AI help, you know, kind of prepare that terrain so that you're better informed as a human to make that decision. And I think, you know, kind of keeping that uh structure in our head of just like you said, still thinking of it as a very smart member of our term of our team, but that it's a team member. It's not it's not leading the charge. I love using it to maintain different like guided intake processes for people when trying to navigate different complex templates and these types of things. The goal is to help us be more informed, better prepared, but not to let the actual technology to be the decision maker.

Starting Small With Policies And Pilots

Julia McIlroy 11:22

So for a procurement office that's just beginning its AI journey, what's the best place to start? Should agencies begin with a small pilot, focus on internal productivity, develop an AI policy first, or something else?

Jennifer Lyons 11:37

Yeah, um, I think I said before, I'm a big fan of kind of those small, quick wins. So, you know, I hopefully I think you do have to start with some sort of AI policy so that, you know, um there's some guidance for for your users. Um, but beyond that, um, you know, I think one of my favorite first uses and quick wins that really got some confidence and some energy or just um moving along with these projects was simply showing, like, hey, we can we can use AI to put in all of our very complex regulations, statutes, and ask it to rewrite them at an eighth-grade level. Um, and it sounds so simple and so silly in some ways, except for we have spent hours upon hours thinking about different guides or different ways of breaking down the regulation and how to inform people of the obligations without getting them lost in the legalese. And you know, I think within five minutes it was like took one of my most complex regulations and just rewrote it, and I brought it to uh the director of that unit, and she was just mind-blown. She was like, Oh my gosh, I have been working for months trying to prepare basically exactly this, and you know, and so that really changed her whole attitude because she wasn't thinking in terms of like, you know, what are kind of some of these quick hit wins? So, you know, what could we, what is our most painful stress point in our workflows? And if we break down that workflow into tasks, what tasks can the machine do better and what tasks do we want to reserve again for that human judgment? Um, so yeah, I do think you want to kind of go through it in stages. So, you know, starting out very uh high level with just your policy and kind of guidance for within your agencies, but then yeah, finding some of those small, quick, everyday use wins that people can just start playing with it and just start talking to it and um you know, telling it about your job and what you could do better and how it thinks it could help you, you know, because that's one of um the things thing people don't think to really do, too, is just to have a dialogue with it or ask it, you know, I'm I would love for uh I'd love to be able to pilot something in my agency, interview me about my agency so we can come up with the best pilot. And you know, so you can ask it to really like pull the right questions out of you as well.

Julia McIlroy 13:59

You know, that interview is a fantastic idea. I love it. That is that is great. That's a great use.

Procurement Chatbots For Onboarding Help

Julia McIlroy 14:06

And how has your bot been received by the public?

Jennifer Lyons 14:10

It's great, it's great. Um, I feel like uh particularly new employees, um, you know, I had a wonderful, um, wonderful message from a legal intern who was here for the summer saying, you know, this was like immensely helpful because usually when we first have our interns start, they would sit down and we hand them a 130-page best value handbook, which I tell them is a user's guide to our regulation, and here spend your first week happily reading this very, very dry, dense material. Um, and instead, you know, I gave her the link for our chat bot, Abe, and I said, ask Abe to teach you about procurement and to teach you the best value handbook. And she was like, This is fantastic. Um, so you know, there there's some um, you know, we didn't really quite see, or I hadn't seen at the time it even being that sort of tutor. Um, but because we really have a scope now where it can walk you through the entire procurement process, um, it's very helpful to even onboarding new employees. I know our help desk likes it, you know, they've been using it to kind of assist them for different calls. Um, and particularly our training department, if they're on with different um people or if they have a question about a specific contract, it's much easier to go in. So it's really it's helping uh our agency, I think, provide just better, more consistent answers, you know, kind of on a regular basis. So that's been a huge help.

The Next Wave E-Procurement Sourcing Negotiations

Julia McIlroy 15:36

So looking ahead three to five years, what do you think will be the most significant AI use case in public procurement? And what will the procurement professional of the future be doing differently because of it?

Jennifer Lyons 15:49

Um so yeah, looking ahead the next three to five years, I think that there's several opportunities for significant use in AI. The the three big areas I see hopefully that'll be um significantly impacted in procurement for through AI in the future, I think, um, are really break down to um kind of three major segments. So one is that e-procurement segment, right, that's kind of common to every uh state and jurisdiction doing public procurement, or becoming more and more common. We didn't all have electronic procurement systems for sure forever. Um, but it's becoming more and more the standard. Um and I think being able to make, you know, make that system as uh, let's say user-friendly as possible, as integrated as possible with your payment system, with you know, vendor onboarding, um, to the extent that we can synthesize and simplify a lot of that entire posting of the solicitation process, um, I think will be a huge win because I think we spend a lot of time really, you know, just stressing about the details of that and then constantly we reworking the workflows. Similarly, the that sourcing piece, that major, you know, doing the market research, putting together the specifications for the RFR, um, you know, making those key decision points, doing the evaluations, all of that is incredibly voluminous and time-consuming and you know, nuanced detail. Um, so I think you know, that will be a huge item to be able to, again, break down that workflow into different tasks, figure out where the legal judgment, where the final judgment, where the you know, all of the ethical judgment, all of these kind of things that need to be reserved for the humans still stay in that process. Um, but helping, letting again some of the stuff that AI does do better, maybe making some of the comparison charts for us, um, or making some of the evaluation charts for us, um, just could help us become better informed humans when we get to those evaluation decisions. So I think one, e-procurement, two, the you know, the sourcing process, helping that sourcing team. And then three, and one that we've been tackling for a while in Massachusetts now, and that no matter how much you tackle it, it is just the you know, over-bearing, never-ending system, and that by that I mean the contract negotiations. Um, so you know, trying to go through 30, 50, sometimes 100 pages of clarifications, particularly um, you know, in some of the technology fields where vendors also have their own master, you know, service terms or service level agreements or you know, uh end user license terms that they want incorporated into part of the contract, but obviously cannot supersede any of the Commonwealth or the public terms. Um, so those can be really, really repetitive, redundant. They, you know, we were constantly, we used to have kind of a long um list of interns coming in and out, and we would try to train them to do first-round reviews with us. And um, it was great because they could flag, you know, major issues and that sort of thing right off the top, and so it would help the attorney save some time right off the top. Um, but we then were able to build a contract editor, which we lovingly named 1L after the first-year law student, um, and that is able to save just so much time right off that initial first review. But it's something that you know I would like to see, you know, continuously working on and see a 2.0. I would love to have this particular technology being able to, you know, it can help me create the red line now, it can help me analyze the red line, help me get that back to the vendor 10 times faster than previously. But then once that's happened and the vendor sends me their red lines back, I don't have anything that can distinguish between what I've said and what that's said, and then help so something that could help an attorney or category manager navigate the throughout the entire negotiation process. We have uh a team of four attorneys and we have about 80 vendors right now that we're trying to onboard with negotiations that are getting prioritized. So we're building uh basically a library of prompt answers. So we have, you know, the whole team has an idea of what um what we've agreed to as acceptable clarification language and how far we can kind of push that versus where we're really up against, like, no, that's uh that's a conflict for us. Um, so it's really helped analyze that line, and I think as we're building out that library of prompts and answers and understanding that more and more, um, I'm hoping that there will be room for kind of a 1L 2.0 that's gonna be even more informed and be able to help the attorney or the cat or contract manager through the entire negotiation process, not just the initial triage. So that is the goal.

Julia McIlroy 20:44

I'm hopeful for the 1L 2.0 because anything that can save you time and resources on redlining is a win.

Jennifer Lyons 20:53

100%.

Final Takeaways And Sign Off

Julia McIlroy 20:56

I don't want to spend the rest of my life redlining. Jen, thanks so much for being a guest of Pulse. Until next time, keep asking questions, improving processes, and moving procurement forward. And remember, we work in the sunshine. Bye for now.