AI’s Impact on Procurement with Kevin Frechette and Randall Moore

Hello, and welcome to the latest edition of the Techron AI DIO series. I'm your host, Mike Bazar today with Kevin Fette, CEO of Fair Market. And then we also have Randall Moore, who's chief procurement Officer for Boston University.

And two of these folks have been working together on a sourcing solution that, um, leverages AI to help people with procurement. Kevin Randall, welcome the show. Thank you.

Thanks for having us on. All Right, Kevin, um, uh, you're not a household word yet. I don't know how that didn't happen, but it's a work in progress, as they say.

But take a minute to describe what it is that Fair Market does and how you came to this solution. And, um, can we apply AI to sourcing and what does that look like? Yeah.

Uh, fair Market. We're an autonomous sourcing platform. Uh, we work with mid to large organizations to help them automate the sourcing process, so get more competitive bids for what they're buying.

Uh, when we came into the space back in 2017, where Boston based company, uh, it just blew our mind how manual and legacy the tech was. This is being served up to procurement organizations and also to end users. So back in the day when we first met Randall 2018, uh, it wasn't ai, it was basic, just like, let's deliver a better experience, let's have some integration.

And then it went to basic RPA some ml. Uh, about five years ago, they pushed us from an AI perspective, and the last year and a half, it's been all about gen ai, obviously for the right use cases and being specific with it. So it's been a cool journey.

Um, we are, uh, Inc's fast growing company within procurement. We've gotten garner recognitions kind of across the board. Uh, but it always kind of comes back to our roots of customers and partners like Randall and BU that it challenging us at every step, which it is like a hell be tension.

So excited to, uh, to be on this and talk about our story with them. Randall, how did you come to work with Fair Market? What, uh, attracted you to the platform?

Because, well, there's a lot of choices when it comes to procurement these days. Well, I, I came to bu in, in 2017 after a, a fairly long stint in procurement strategy consulting. And one, one of my key charges was to improve the, the sourcing and procurement process here at the university.

We were fairly, um, in ingrained in the way that we did things. And so I was looking for, for new ways to, uh, better cover the addressable spend, you know, get, get the spend under control, uh, for the, the categories that we're charged with. And, and in 2017 I saw a, um, a flyer for a, uh, regular monthly meeting of the Institute of Supply Management, uh, greater Boston Chapter.

And, and Kevin, uh, was on the docket that night. It Was fresh Markets, though, Randall, they got the company name wrong on that first one. Go, go ahead.

But yeah, that was, it Was Kevin from Fresh Markets? Well, it, it could have been a lot worse than that, Kevin. Um, but anyway, I, they, they presented their, their platform for managing tail spend, which is the, the, the part of, of our spend that is an annoyance, but it's very important because it, it's not covered specifically by, uh, contracts that our sourcing team has put into place.

And it gave us a way to very quickly identify capable suppliers and, and return bids for SKU u specific purchases that we needed to make. I saw the demo that night and I, I said, I've gotta have these guys in for a conversation. And, and before you knew it, we were under contract, uh, within, uh, a few weeks after that.

Kevin, how will AI be applied to procurement? And I'm asking the question 'cause I feel like there's a little imbalance in the system these days. Sellers are applying AI left, right, and center to optimize deal flow and maximize profitability.

But, um, there's another side to that equation, right? And, uh, very interesting you bring that up, uh, 'cause that doesn't come up a lot, is the fact that there's all this hype and all this AI and actual venture funding that goes into helping people sell better, helping to get more leads, more opportunities. The buy side has been kind of forgotten about when it comes to how do we not only make their lives easier, but make them more effective when they're buying on the supply side.

So, um, one thing that which try really hard to do is to not like peanut butter spread gen AI or AI for procurement, because that's when you kind of get marketing fluff. It's what specific areas within procurement can you apply AI or Gen AI to get to better business outcomes? It's not a one to 10 x, you get 5% better every month.

So for example, um, we're now helping, uh, end users or shoppers at different organizations interact with LLMs to better capture what their requirements are for what they're looking to buy. 'cause traditionally that's been kind of like a, a difficult spot where it's like not great descriptions, not great data. We then used AI to say what suppliers should actually be recommended.

We found out that AI especially better than Gen ai for that we then use Gen AI to help to create the event. So actually set it up in an optimal way. And then now we're also using AI from an autonomous negotiations perspective.

So to your point, instead of just like this AI army coming from the, the, the sell side, it's how do we start to empower buyers with that to make sure that they are getting the best value? And that's not, that's not always the best price. It could be that we wanna work more with diverse and sustainable spires.

We wanna do things faster. We want better visibility to de-risk what we're buying. So I think the main thing that we keep coming back to when we think ai, gen AI for procurement is it doesn't really matter what the features and functions are.

It's what business outcomes can you actually get to? And that's been something that BU has always pushed us on. It's like, yeah, like, heck is cool, but what are we trying to deliver back to our students, to our faculty, to our team members over here?

Mm-Hmm. Randall, how dynamic can procurement get? And I'm asking the question because, you know, in, in my experience talking to folks, it takes a while to onboard a new supplier.

And in invariably there's some sort of crisis somewhere in the world, and I can't get access to something. But it's not like I can just dynamically add somebody to the approval process for somebody in a B2B transaction. So how fast is fast these days?

And, um, how much pressure is there to be more agile? Uh, tremendous pressure. You know, we, we have to be agile and, and compress the cycle time, uh, or otherwise would be just absolutely overwhelmed.

You know, the, the university is a, a big place. We have, um, more than 34,000 students, 11,000 faculty and staff. We're, we're City of Boston's second largest employer.

You know, the, the transactions and the request for sourcing and the request for new suppliers and contracts are just coming at us left and right continuously. And we would be quickly overwhelmed if we did not innovate and find ways to creatively do, you know, more throughput with the same set of resources that we have. And, you know, we see Gen AI is helping us in, in many ways to, uh, uh, manage that workload, uh, get better throughput and make better decisions.

Yeah. And Everybody has a vision in their head about ai and it basically says, I'm gonna have some magical AI agent will represent me, uh, execute deals based on the parameters that I set forward. And this is gonna be an awesome experience.

The sellers though, have a similar vision. And I'm trying to figure out in my mind, if the sellers have an AI agent that's optimized for one thing and the buyers have an AI agent that's optimized for the exact opposite, and won't the two of these AI agents eventually converge and more or less cancel each other out? Um, like the war on AI versus ai, it's like a fun, it's, it's truthfully like the future.

So like that, that's an accurate thought. Um, the one area that we have learned, 'cause when we talk to, we have a million suppliers that sell through Fair Market as well, is, um, it's usually not a game of winning and losing. It's a game of are we getting out to the right supply base?

And it could be to ESG suppliers, to trusted suppliers. It could be specific initiatives you have in different regions. Like BP we work with, they wanna work more with black owned businesses in the uk they're part of the billion dollar round table.

So like, they have initiatives of like, all right, how do we optimize what we're buying? Can we get our arms around it? 'cause before Fair Market, a lot of companies, they have tens of thousands of transactions.

They focus on their top maybe, I dunno, 10%, 5%, but everything else kind of just like falls under the radar. But that's a huge opportunity. Tens of thousands transactions, billions of dollars that you can potentially optimize and make available to the supply base.

Now on the supply side, it's not like a win or loss. They're trying to have an equal shot and a fair like, level playing field to work with these big businesses. So I think the future state, you're absolutely right, it'll be configured on either side, a great end user experience.

It'll be, uh, almost like human by exception. So it'll be triggering when people need to get involved, but they're both just looking for like a best value or an optimal transaction. And I think the challenge that you have today is for a lot of organizations, it's so black and white of like this old way of doing it, okay, you are buying this category in this region.

I always go with these one or two suppliers. Is that the right supplier? Is that fair?

Are there other potential suppliers? Are there sustainable? Are there diverse we could be working with?

But because the tech available today, uh, prior to fair market, it's, it's just very manual. You're not gonna spend an hour across 10,000 purchases to figure out what is the optimal strategy. But if you can start to automate that and you can start to actually ingest, like inject it with AI and gen ai, you can put it out to a wider audience.

Still, to your point earlier, being controlled though from a risk perspective, not everyone with the sun, but you can start to actually execute on category strategies, which is kinda like a crazy thought for a lot of companies that have a, this book of a strategy. And then how the hell do you expect an end user to know how to follow the policies and procedures? If you make it easy, then you can actually accomplish that.

Randall, when is the future of a procurement officer in this great AI expanse that we're entering? I mean, a lot of people say there'll always be humans in the loop, but how do you envision that job evolving and what exactly is that loop gonna meet? Well, you know, the way we kind of measure success is we want to deliver best total value to, um, the, uh, budget holders across the university who are spending, uh, research dollars in, in mom and dad's tuition, money, you know, that is a, a, you know, very high order, um, requirement that we have to deliver best hold value.

And it's not necessarily best price, which supplier can meet our specification, our timeframe, uh, meet, um, you know, risks around, uh, reputation, uh, legal risks, terms and conditions. You know, it, it is all these factors that go into making the, the best buying decision. And we want to support our customers because, you know, we're not making the supplier decision.

It's, it's the, the dean of a school, uh, who has a, a limited budget, uh, that they have to, um, that they have to control. We can act as, um, consultants and help them support that decision, but it, at the end, it's their money and, and they make, they make the decision. So we need to use these tools to help them achieve the, the best answer in the shortest amount of time, uh, we're, we see using ai, uh, to help us, uh, red line and negotiate, uh, supplier contracts.

We have a massive volume of contracts that come through, uh, our office, uh, on an annual basis. We have, uh, thousands of users across the university who, who come to us with, with trouble tickets and using our P two P system or, uh, suppliers who are having issues with, with P two P or getting paid. Uh, we can use AI to, to field those questions and, and play back, uh, responses or suggested outcomes or best ways to fix their problem, uh, using ai.

And then finally on the sourcing intake side, what Kevin's been talking about, you know, put together better, quicker, faster, more intelligent RFPs, uh, for our user community. Mike, if I could jump in, 'cause once something that's very unique and kinda interesting about Randall and team at bu, um, is just like the, the mindset of like, let's start the journey on AI and gen AI knowing it's not gonna be a 10 x like Uber night. You don't become Taylor Swift on the, a tour of like, on stage for everyone.

Like you're in a garage band, then you're kind of going through it. Like that's the mentality that Randall's team has had. It's okay, we know what Mar's law is that people typically overhype things in the short term and under hype things in the long term.

So right now we're in the overhyped payer period, and Randall's team has been really good about saying like, all right, let's be real. Like what is the value we can get out of it today? Like, what can that mean back to our end users?

But then over time, if we go on this journey, we can get to that future state where we are like top tier from an EDU perspective. And the best, like, and I've shared this story before, the best uh, example of that is back in 2023 is March. I sent an email to five customers.

We just said, okay, we're all in on Gen ai. And I said, I sent email Randall. I said, we're setting up a gen AI SWAT team.

I don't know what the time of commitment's gonna be. I don't know where it's gonna go, but I think it'd be fun to learn together and are you in, and Randall rolled back literally in three minutes. He says, sounds fun, let's do it.

And like, this is a busy person, but to say like, all right, I'm gonna go on the journey with this company and I'm just gonna go and let's just learn and learn. Like that's, that's why we've been able to get to where we are today. It's because we're on the journey with people like Randall who bring back that real side, the Accenture back on the BU side.

We bring kind of that outta the box, like gen ai, like thinking, but then we make it real and let's get a use case, use case use case. So I just wanna share that, 'cause that was a, that was a pretty cool moment for our partnership. Yeah, it, it's grounded in, in reality, but also we learn from each other and, and that's what has made this journey so far.

How did you overcome some of the inertia that exists in procurement systems? If you go back in time, people have been using everything from spreadsheets to, uh, huge ERP platforms and and extensions to things like Ariba and they come with weight over time. How did you kind of get people to go?

Uh, Well, not too long ago, and I have this framed in my office, you know, we were on a carbon paper, paper-based purchase order system. You know, this is before my time here, but you know, the, the systems have, have continuously evolved. Um, and now, uh, I believe we are on a, a state-of-the-art procure to pay system that is entirely, uh, paperless.

We, uh, issue better than 100,000 purchase orders a year and, and, uh, pay, pay out hundreds of millions of dollars in, uh, invoices either on purchase order or disbursement. And it's all done without paper. Um, I think it delights our community, uh, community to be able to use a system like this.

It does require training. You can't just jump in and expect to use it, um, on day one, but once you know how to use, uh, the system, it really is a, a, you know, a complex machine, but it has tremendous, uh, capabilities and the fact that we can pay 98% of our invoices, uh, without human intervention. You know, if the invoice comes back to us electronically, it matches the purchase order.

And as long as that rule is met, that test is met, it posts an SAP and it goes out for payment. It's, um, it's really a, um, I think a highly evolved procure to pay system that we have today. And Anything involving AI these days requires a certain amount of, uh, oversight governance.

There are these slowly things called audits in procurement. Um, how do you kind of make sure that the AI is actually doing what we intended it to do? Yeah, no, we get the question almost every day.

I've got there about 60 security assessments, which I never used to be part of since we started using Gen ai. Uh, so the way to think about it would be, we've had AI for the last five years, uh, AI running essentially in-House at Fair Market, where we're using collective intelligence to make sure we're bettering the platform for all of our customers by getting an understanding of how suppliers are bidding, where they're bidding, uh, how fast they are and if they're winning or losing. So that's our supplier recommendation engine.

Once again, that's all in-house. Uh, customer's data is secure to that themselves. It's not training external models, but once you start to get into Gen ai, there's a fair amount of education that needs to get done.

Um, 'cause you have like three different setup you could do. It could be through open ai, uh, it could be through maybe like Ara on Bedrock, so almost like a managed, uh, like platform. Or you could do it in-house.

So say you're running a lot of three, one on AWS, which is more private. But if you're thinking about OpenAI in general, which most people are on, um, we have signed up for the API platform. Uh, and then on top of that, we've created a partnership with them where they've validated our use case is a low risk use case.

Uh, we found a very good job setting up controls and guardrails to make sure that like there's very little risk that you can have when interacting with the LLM. Uh, they then gave us a zero, uh, zero data retention badge, which means the two main questions we always get in every serious assessment are, like, at the core of it, is your data stored on any external models? Does it train any external models?

For us? Both the answers are no, and we can show that we have documentation, but you still have to go through it. Um, so I guess the point is for a lot of our customers, especially back last year, it was a, like an education phase with them.

They'd bring in like seven different law firms to ask a bunch of different questions, but as they started to understand, all right, this is the use case, it's low risk, we understand what the controls are. Okay, we're good with that one. And then that one, and then that one where it gets a little interesting is if you start to get to autonomous negotiations because now you're interacting with external suppliers and the AI essentially is your brand and your tone.

So you have to be a little bit more thoughtful in terms of how to train and fine tune that so it matches how you want your brand to be out in the world. Right. Randall, um, as you kind of think about the future a little bit, you know, what is your crystal ball telling you you're gonna be in a year from now?

What's your kind of roadmap of things that you wanna achieve in automating procurement processes? That's a really good question. Um, and one that I, I probably only have a partial answer to.

Um, you know, we, we know that the stakes are, are huge for the university that, you know, we, um, we have to compete. Uh, higher education is in, in many ways under, um, a bit of duress these days with the, with the current political environment, um, uh, higher ed as, as a product itself. Um, so we, we need to, you know, really sharpen, um, our sword and, and just be as, as, um, efficient as we can with, with limited, with the same or limited resources.

And, um, I, I see, you know, AI and solutions that Fair Market is, is bringing us as, as a way to, uh, remain competitive and continue to deliver, uh, best total value, um, and maintain the throughput that our customers expect from us, uh, on an annual basis. Kevin, a lot of people are asking this question, how smart can smart get, we are seeing better reasoning engines being added to these LLMs and, um, people are starting to wonder, um, you know, how much can these things recognize the semantic relationships between processes? So same question, looking at a year from now, where are we gonna be?

Uh, I almost, my answer I wanna say is I have no idea because of how fast everything is changing. Like, do you see like open eye, uh, launched, uh, oh one that is more for like complex reasoning, which are some of the challenges they had before, and then you saw like LAMA three one, so Metas platform, uh, and now looks, I think it's 405 billion parameters. So if you think about like the, the scale that we're hitting so fast that I think that, uh, EGI is still like in the future, I, I think that the fear of that, uh, you need to have controls and guard rails, we need to have like, obviously safe ai.

Uh, that being said, I still think we're in Amaras Law. If you're thinking a year out, um, it's still probably gonna be overhyped compared to what it can actually deliver. I think we're probably two or three years out from that inflection point where it does start to really kind of out, like really, really take it to the next level.

Um, and I think that it's not gonna be, it's kind of like, uh, I, I, I kinda associate a lot with the journey to the cloud where like you didn't know where the cloud was gonna go. You had your on-prem hardware and then you went to your private cloud and that was the main thing. And then you went to the hybrid cloud, you're like, oh, that's awesome.

Like, it's a, it's public and private. No one would've ever thought you would go fully to the public cloud. And then no one had ever thought AWS or Microsoft would've been the biggest players in that space.

So what, once again, it's kind of difficult to predict where it's gonna be or like how like daunting it's gonna get. That's where we kind of keep pulling it back to like, what does the next three months look like? What just got rolled out?

So for us, it's all about how do you stay flexible every month. So okay, if we start to wanna bring in more multimodal s people can have like voice, text, image. Cool.

If you wanna start Gene multimodal where we might wanna use different models for different use cases within the procurement process. Cool. So for us it's all about like we don't know where it's going.

We know we want to focus on a great end user experience and deliver results and value to businesses or to organizations. And we think the best way to do that is just to stay in like agile and keep iterating as tech keeps evolving. All right folks, you heard it here.

Procurement and ai, it's not just a job, it's an adventure. And we're not quite sure where it's all gonna end up in the end of the day, but one thing for sure is it won't be the same as it was yesterday. Gentlemen, thanks.

Being on the show. Yeah. Appreciate you having us.

Thank you for having us all. And thank you all for watching the latest episode in the Textron AI video series. You can find this episode and others on our website, and we invite you to check them all out.

Until then, we'll see you next time.