In this UiPath Fusion interview, Deluxe leaders Kari Mesick (business lead, Treasury Services), and Satish Balasubramanian (divisional CTO, Architecture & Data), join theCUBE’s Dave Vellante and Rebecca Knight to unpack how Deluxe is evolving from manual spot checks to agentic auditing. They share why last year’s award-winning bot laid the groundwork for this year’s agent-driven approach, moving from sampling to full-population analysis across ERP invoice items and contracts. The discussion covers timeline and governance realities – request in January, approval by March, and an initial build in roughly four to six weeks using low-code tools – plus the data-privacy and compliance reviews required before putting agents into production.
The conversation dives into real-world mechanics: bots extract key terms from lengthy contracts; agents compare those outputs with invoices, surface discrepancies and recommend next actions with a human in the loop. Mesick and Balasubramanian explain how this shift augments teams rather than replaces them – freeing people from tedious, after-hours work and building trust through phased adoption, internal communications and training. They also outline initial POCs focused on contract/pricing accuracy and revenue leakage, the role of partners such as qBotica and how an agentic approach can help reimagine legacy workflows as Deluxe transforms from a check printer into a payments and data company.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
UiPath Fusion 2025. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For UiPath Fusion 2025
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for UiPath Fusion 2025.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
UiPath Fusion 2025. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to UiPath Fusion 2025
Please sign in with LinkedIn to continue to UiPath Fusion 2025. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Kari Mesick & Satish Balasubramanian, Deluxe
In this insightful episode of theCUBE at UiPath Fusion 2025, we join Kari Mesick, business lead treasury services, and Satish Balasubramanian, divisional Chief Technology Officer architecture and data, both from Deluxe. The discussion focuses on their award-winning implementation of artificial intelligence agents at Deluxe, as part of their larger transformation from a traditional check printing company to a data-driven enterprise.
Mesick and Balasubramanian, along with theCUBE analysts, explore Deluxe's broader innovation strategy, emphasizing the company's evolution over its 110-year history. The guests examine AI’s role in reshaping Deluxe's business processes and their approach of incorporating AI cautiously and systematically. The discussion highlights how AI agents address challenges such as agentic auditing and improving operational efficiencies, showcasing the fine balance between technology and human oversight.
According to Mesick, integrating AI has allowed Deluxe to achieve unprecedented accuracy and efficiency in auditing and contract management, enabling employees to focus on more strategic tasks. Balasubramanian emphasizes the importance of earning trust at every step, ensuring AI implementations align with broad organizational goals and compliance regulations. This transformation journey is seen not just as a technological upgrade but as a fundamental shift in how Deluxe envisions its business paradigm in the future.
In this UiPath Fusion interview, Deluxe leaders Kari Mesick (business lead, Treasury Services), and Satish Balasubramanian (divisional CTO, Architecture & Data), join theCUBE’s Dave Vellante and Rebecca Knight to unpack how Deluxe is evolving from manual spot checks to agentic auditing. They share why last year’s award-winning bot laid the groundwork for this year’s agent-driven approach, moving from sampling to full-population analysis across ERP invoice items and contracts. The discussion covers timeline and governance realities – request in January, approv...Read more
exploreKeep Exploring
What is the context of the conversation taking place during the live coverage of UiPath Fusion?add
What company was Dave involved with, and what are its primary services?add
What are the capabilities of agentics in conducting comprehensive checks and analyses of ERP invoice items across multiple clients?add
What are the goals and considerations involved in transforming a 110-year-old manufacturing company into a payments company, particularly regarding the use of AI and technology to modernize legacy processes?add
What are the three categories of AI that Deluxe focuses on and how does the company engage its employees in the transformation process?add
>> Good morning, everyone, and welcome back to day two of theCUBE's live coverage of UiPath Fusion. I'm your host, Rebecca Knight, along with Dave Vellante, my co-host and analyst. I would like to welcome two award-winning guests to our show. We have Kari Mesick, business lead treasury services at Deluxe. Welcome, Kari.
Kari Mesick
>> Thank you.
Rebecca Knight
>> And Satishkumar Balasubramanian, divisional CTO architecture and data at Deluxe.>> Yeah.
Rebecca Knight
>> Thank you so much for coming on.
Dave Vellante
>> Congratulations.
Kari Mesick
>> Thank you.
Rebecca Knight
>> Congratulations. Dave was one of the judges of-
Dave Vellante
>> Yes, I was....
Rebecca Knight
>> this AI25 award.
Kari Mesick
>> Wonderful.
Rebecca Knight
>> And this is a back-to-back wins for you?
Kari Mesick
>> Yes. Yes.
Rebecca Knight
>> Before we get into what you won the award for, why don't you tell our viewers a little bit about what Deluxe does?
Kari Mesick
>> Yeah. Do you want me to start?>> Go ahead.
Kari Mesick
>> Yeah. We are 110-year-old company, payments and data company. Really built our success on innovation over those years. Anything you want to add?
Dave Vellante
>> So, we are a traditional check-printing company. I think everyone in US probably knows Deluxe as a check printing, a manufacturing company. Last 10 years, we have pivoted ourselves. We are now a more trusted payments and data company. So, we have like 4,000 financial institutions that we serve. There are millions of small businesses that we serve and then there are direct customers that we serve. So, we help businesses pay, get paid and grow. That's what we primarily do.
Rebecca Knight
>> Excellent. So, Kari, last year you won for a bot-
Kari Mesick
>> We did....
Rebecca Knight
>> but this year you won for an agent.
Kari Mesick
>> Yeah.
Rebecca Knight
>> And so were you at UiPath Forward last year, your eyes lighting up, thinking, "Okay, we can use agents"?
Kari Mesick
>> No.
Rebecca Knight
>> Okay.
Kari Mesick
>> Last year I was so excited about the bot. We had had this great success. We built it fast and we had immediate success with it. We immediately benefited from it. So I was super excited about our bot. Everything was about the bot last year. And we went to this conference last year and they talked about agents and I was like, "Oh, never, never, never, never." And then by the time January rolls around, I'm like, "Oh, I think I need this." And I submit a request to Satish's team. And by March, you approved it and then we start working on an agent.
Rebecca Knight
>> So, I think agents were all over the place from last year. So, even within Deluxe, while we were trying to explore a lot on AI, how AI works, generative AI helps business, because we were big on how AI is going to help us achieve value. We're not going to do AI just for the sake of AI as a new shining technology, so we are very particular about that. So when Kari came and said, "Oh, there is this particular problem that I'm trying to solve. I'm not able to go at scale. I'm very limited on bandwidth and people," we said, "Probably that's a good use case for agents." And then, that's how it all started.
Dave Vellante
>> And as I recall from your submission, you used to have billing and audit people would've to manually spot check-
Kari Mesick
>> Spot check. Right....
Dave Vellante
>> and sample adjustments to contracts, for example. And you shifted to agentic auditing. So, I mean you're going to miss a lot when you're spot checking.
Kari Mesick
>> Yes, you do.
Dave Vellante
>> So, can you talk about that transition, what that was like, how you got the confidence to get there, how long it took? I want to know everything.
Kari Mesick
>> Right, right. So, going from a couple of teams to doing spot checks, as you ought to, to say, "Oh wait, I want to go for the whole population and I want to go deep. I want to look at every ERP invoice item number for every single client. I want to go compare it all to what my projections had been before some work," and that would've been impossible. So, only with agentics would this have even been kind of a conceivable thing. And so, I even felt like when I wrote up the request to Satish, I'm like, "He is not going to say yes." Okay, so I should also say that's what I initially said is bot work. But then I also wanted to know if there's discrepancies, research him, tell me what happened there, what does that mean and what should I do? And of course, in future, go do it. But Satish, eventually, yeah.
Kari Mesick
>> Yeah, I think that's how we approached it. So, we were doing a lot of RPAs for the last couple of years. We were doing process automations, but when this particular thing came in, so we had already done the document extraction, looking at contracts, extracting the key concepts from contracts was already implemented. So, when we wanted to compare contracts with what actually exists in invoices. And then, mainly, I think the later point that Kari mentioned, so we want to take action on what we see, probably go and update something or flag that something is going wrong, go and correct it and things like that. Then we thought, "Probably that's the perfect use case for an agent."
Again, we took it in more of a phased approach. So, we started as a simple bot, converted that bot into an agent, which is going to compare and come back and tell Kari and team on what it sees, whether it's right, wrong, the missing gaps and all those. Then, we want to go from there to actually taking some autonomous decisions of going and updating or correcting the fixes.
Dave Vellante
>> And you found that the agents were able to interpret complex pricing scenarios, is that right?
Kari Mesick
>> Yeah, it's interpreting properly. So far, we're still in the early stages of testing the POC. So, yeah, it's been amazingly accurate. I'm more caught now in the discrepancies it found and the reasoning why and what it's suggesting as next actions. I'm kind of at that point right now.
Dave Vellante
>> And the POCs are a contract and pricing and revenue leakage, correct? Those are the first two that you've started?
Kari Mesick
>> Yeah.
Dave Vellante
>> And how long did it take you to get up and running where you could actually run water through the pipes?
Rebecca Knight
>> See, the time it actually took to implement was very less because I think with UiPath, you have the option of the low-code, no-code, which helps you to make a prototype pretty soon. Then, I think it actually took time for us to ensure that we are not doing any PII data leakages. Meaning, does the contracts have some PII data or PCA data which we shouldn't expose? Or can we give it to LLMs? Because we had to make sure that our legal team, our compliance team are aligned with what we are doing. So getting those approvals took some time, the typical governance processes, but the actual implementation I would say took four to six weeks. We were able to make the agents up and running in four to six weeks.
Kari Mesick
>> Right. And before that work happens, I write business requirements and I have working sessions. We work with qBotica as an accelerator, as a partner, and so working sessions with them to clarify my business requirements and then they magically convert to all of this that Satish does.
Dave Vellante
>> So, the data comes out of the operational system and then the agent adjusts that and then interprets the contract and the logic and the rates and configure that. And then, a human validates that and that's the sort of process that you're using now and is that teaching the agents and adjusting?
Kari Mesick
>> Right, it'll get smarter and smarter over time. Yeah, absolutely. And things will change in our contracts over time. Things will change in maybe our billing system. So, it will be a constant set of improvement and monitoring. There will never be a time we wouldn't have humans in the loop. And of course, this isn't taking anyone's job. This was impossible.
Dave Vellante
>> Right, it couldn't be done.>> So, I think that's what we keep repeating. So, the agents are not going to replace people, it's actually going to augment people because Kari's team was not able to go at scale looking at all contracts, looking at all our customers for the price increase. That's the benefit that the agents are giving right now.
Rebecca Knight
>> So, it's augmenting them. Now, what are they able to do now in their jobs to be more creative or to think more deeply about strategy? Can you talk a little bit about how it's shifted what they do and how they spend their time?
Kari Mesick
>> So, for the bot, what's nice about that because with the agent, we're still in POC, but with the bot, what it did was I used to have to pull 10 people for two to three weeks in. And on top of their day job, read contracts and put the information in a shared spreadsheet. And then, think of how subjective things are, right? And so, it was inconsistent and people were working nights and weekends, missing family time. So, instead, with the bot, we deploy that and if it's wrong, it's very consistently wrong. It's blaringly obviously wrong. So, you can fix things more easily. So, the bot, that's how it's really helped. With the agent, I would think in the future it's going to create a situation where someone will get a message from the agent saying, "Hey, we're seeing this discrepancy. We recommend this action. Should I proceed?"
Dave Vellante
>> And you said earlier that the agents will take action eventually. So, what do you have now? You have the bots do, the agents are thinking the bots are the doers, but the bots are limited in what they can do and then the agents over time will be able to compliment and do more, is that how you see this playing out?
Kari Mesick
>> So, right now what happens is the agents actually look at both the contract data as well as our invoices come back and show, what's the difference? Is everything matching? Is there any difference where it sees that there has to be an action that has to be taken. So, that is kind of a report that goes to Kari and team who can manually review it and they need to manually go and take an action. So, right now we are more reactive because the pricing changed, we have already generated the invoice for the clients. And then ,we are going and comparing if these two are one and the same, but where we really want to go is be more proactive. So, take the information, update the billing systems, and then also proactively start validating many of these things.
Dave Vellante
>> Okay. So let me ask a follow on. So what does the bot do that's deterministic? What's the bot's role? I think it's clear what the agent's role is, what's the bot's role?
Kari Mesick
>> Bot is extracting the information from the contracts. Which earlier, used to be a manual step, where we look at the contracts, try to look at there are 10 or 12 things that you need extract out of these 40 pages of contracts. So, we were trying to do it manually before. Now, the bots go and extract the information from the bot, keep it ready, and then the agents take that and start comparing.
Dave Vellante
>> If you didn't have the bot built and all that infrastructure and plumbing in place and you had to do it over again, would you do it with the bot and RPA or would you do it with agents? Have you thought about that?
Dave Vellante
>> It'll definitely be with agents. I think it has been an evolution because the entire generative AI space has been an evolution journey. So, we all were going against what was available two years back, which was more of... Even our entire journey with UiPath, we started more like an RPA, where we were just trying to automate the individual step-by-step processes. Then we evolved into more of a bots, which were going and doing extracting data using the generative AI technology. Then we are now augmenting it with agents, which are going to go and do much more cognitive work. So, I think it's been an evolution journey. If you're starting from scratch, everything new right now, probably we could just do everything in an agentic AI, forgetting all the process automations and all the bots that we need.
Kari Mesick
>> But I would say from a business perspective, there's a trust component here. There's a trust component for myself, there's a trust component for my colleagues, there's a trust component for our clients. So, I really like taking the time to test everything, slow roll it a little bit more. I really like that I built trust internally as a company with a bot. And then, now we are still building trust with the agent doing videos and all sorts of things to bring the company along with us on this process, so that when we have the output that we all trust it and rely on it.
Dave Vellante
>> So, this is my premise. My premise is if you have RPA in place, you're going to use that because it is trusted and you're going to be ahead of a company that has a greenfield and is going to be starting with agents. It makes sense to start with, but it's going to take longer to build that trust.
Kari Mesick
>> Agreed.
Dave Vellante
>> So, you have a competitive advantage, I would argue, relative to a company that doesn't have any of that deterministic infrastructure in place. At the same time, it's good news for UiPath because the market's changing and everybody's saying, "Oh, RPA's going to get disrupted," which is probably true, but there's a big market for companies like UiPath because you can build with agents. It might maybe take longer to build that trust.
Kari Mesick
>> And if you're solving a business problem, the business has to trust it. And we're a little slower on that.
Rebecca Knight
>> For good reason.
Kari Mesick
>> I think so.
Dave Vellante
>> Every solution that we built, especially being on the financial services side of the world, I think we want to ensure that the trust for business partners, like Kari and team, and also for our customers, I think everything that we do needs to be under the trust umbrella. We are making sure that we are, from a regulation standpoint, we have all complied and we also ensuring that the security is top-notch. And that's why, as I said, even from LLM usage or any AI and agentic usage, we are making sure that all the data privacy, the data leakage, all those are also taken into account. So, it's a long process. But I think to your point, Dave, I think the way we approach it, like starting with an RPA, ensuring that we are building it brick-by-brick and not jumping suddenly onto a 60,000-foot of agents really helped us in this journey, earning that trust, ensuring that we are doing things right and then we are just augmenting it with all the evolution that's happening in the market.
Dave Vellante
>> So, day-to-day, you've got to be prudent, you got to protect the crown jewels. But last year you were like, "No way. I'm not doing agents."
Kari Mesick
>> That'd be crazy.
Dave Vellante
>> So, next summer when you're at a Woods meeting and you got this whiteboard and you're thinking about, "Oh, what is possible?" Not that you're going to push that day-to-day tactically, but what is possible? Because clearly, you've got new opinions about the potential. What's possible?
Kari Mesick
>> I'll take a stab.
Kari Mesick
>> I appreciate it.
Kari Mesick
>> So, we are 110-year-old company trying to transform our business from a manufacturing to a payments company. And then, as part of the business transformation, we also need technology transformation. So, we see AI and agent as a tool in our toolkit, which is going to accelerate the transformation. So, there are abundance of processes, like the pricing audit is one of the processes that we have. There are abundance of processes across our company, which is all traditional legacy, which needs to be modernized, which needs to be improved, so that's what we are actually going to go against. How can agentic help us in say workflows? How can agentic help in autonomous decisions that could be made? How it can help us transform our business model and transform our entire technology landscape?
Dave Vellante
>> Do you look at your existing processes as opportunities to automate things that we couldn't do before with RPA? Or do you look at it, what you just said, Satish, as, "We need to rethink all of our processes," because the potential is there for agents to actually, like Lego blocks, put together new workflows that are going to drive productivity. How do you think about that?>> The later, but I think going to later, we need to do the former. We have to look at individual processes right now to understand on what is there in that process? What I could change in that process using agents? But we definitely want to reimagine multiple of our process. A good example could be one of the business that we provide is log boxes. So, onboarding a customer onto a log box and setting up a log box for our customer has multiple steps that has to be followed one by one. These are traditional. These have been there for quite some time. Now, we want to improve the process, reimagine the process, so that we are able to go to the market. We are able to onboard our customer at the earliest possible timeframe. So, it's a mix of both, I would say. It's not either-or. We start with the process, improve the process, but at the same time we definitely have to reimagine because agent tech is providing as a completely different way in which we think about things. So, just going and looking at existing process to say, "I will move step one and step two, put a bot in between," is just like an RPA. Agent helps us to reimagine, I think we should take benefit of it.
Dave Vellante
>> But there's a learning curve there that's going to take years to play out?>> Exactly.
Rebecca Knight
>> So, this is an enormous transformation going from a check-printing company, to a data and financial powerhouse. And as you said, you're not doing AI for AI's sake. You are really trying to be particular about it and be thoughtful and intentional about where you're using AI. But I also want to ask about how you are bringing employees along. You talked about the trust element, but there is also a lot of real concern and reality that AI is taking over a lot of human work. And it's lovely that those people have more time with their family and they're not working nights and weekends, but can you talk a little bit about the day-to-day and about bringing humans along on this journey too?
Kari Mesick
>> Yes. So, as 110-year-old company, that really has been a check printer, a printer. So, we have printing equipment. We have manufacturing facilities and that have always relied on process improvement and innovative strategies to make ourselves more efficient, more accurate, super important in all this. So innovation isn't a stretch for us at all, that is absolutely woven into our culture. It's just taking this step into this other type of innovation. So, building this trust, I've had our communications team involved with the agentic process from the start. So they have been shadowing our working sessions, and so they've got all the background on it. We did a lunch-and-learn a month ago, two months ago, where we really talked about what was our intent? What's our early learnings? Where are we going with this? No one's losing their job, human in the loop, all these kinds of things. And then, next week we're doing internal video, again, to a larger internal audience like, "This is what we're doing. This is what we're seeing. Again, no one's going to lose their job. Someone might have an augmentation to their job. Hopefully, you won't have to do tedious tasks. And also, if there's something in your job that is tedious or repetitive, let Satish's team know because it's not so tough to build a bot anymore, to just take that little bit off your plate and you can focus on mindful, creative, customer-centric kind of work.">> As a business leader, I've always said technology transformation is easy. You have toolkits with you, use the technology, we could do transformation. Transforming people is the hardest part. So I think in AI, at least in the last one and a half years that we have been in the AI journey... So, in Deluxe, we look at AI in three categories. AI for technology, which is all AI for employees. AI for business, which is how I bring AI into operations side of the world. And then, AI for customers. So we have specific focus on each of these three areas. So, bringing employee along is one of the key pillar that we have. So, there are evangelization that is happening. We have new newsletters that go out to the entire team. We have a central SharePoint site where we are sharing everything that is happening in AI within the company and what's happening in AI outside the company that would change their life. So, we have a strong vision. There is a big leader push from CEO and CIO to make sure that the entire company is AI-enabled. So, people are coming along nicely because they're using ChatGPTs in their personal life, so they definitely want to see how they could use AI in their day-to-day life at work. We have enabled Microsoft Copilot, so people are using Copilot effectively. And we are also rolled out a few of the internal products similar to ChatGPT. So, we see that people are interested, people are coming along, and we are also ensuring that in the three pillars, we have dedicated focus on how I bring in employees? How can I change the SDLC process? How can I train the product team to use AI to generate their product requirements? Similarly, in the operations, how can I use AI to change the business processes? How can I change sales, servicing, marketing? How can I augment them with AI? And definitely, for products, that's a long pole, but products is where we are seeing how AI could help us reimagine some of our products and make sure that we are able to give a better experience and better product for customers. So, there is definitely the top-down push. Also, the interest of people is keeping us making sure that we are able to roll out AI across the company.
Rebecca Knight
>> Great vision. Thank you so much, Satish and Kari.
Dave Vellante
>> Great story.
Rebecca Knight
>> This was a great conversation.
Dave Vellante
>> Thanks a lot.>> Thank you.
Kari Mesick
>> Thank you.
Rebecca Knight
>> I'm Rebecca Knight for Dave Vellante. Stay tuned for more of theCUBE's live coverage of UiPath Fusion. You're watching theCUBE, the leader in enterprise tech news and analysis.
Rebecca Knight
Dave Vellante