Videos
Rheem’s Joe Palomba: AI Won’t Scale on a Broken Process
Insights
- Rheem’s governance board turned down an appealing AI use case once it became clear the use case was built on a flawed process and an ungoverned data model, because AI layered on a broken process does not scale.
- Ideas move from an AI innovation board to a governance board, which approves only use cases whose cost model holds up and whose value can be measured on an ongoing basis.
- Adoption is coming from both directions: a master class showed 35 executives how AI can drive growth, and Copilot licenses issued to each value stream have created employee demand for more access.
At Confluence Americas 2026 in Washington, D.C., Joe Palomba, Senior Vice President of Information Technology at Rheem, spoke with Jeff Kavanaugh, Head of the Infosys Knowledge Institute, about how the HVAC manufacturer is moving AI from pilots toward scale. Palomba says Rheem is still in the pilot stage, and that scale depends on reimagined business processes, governed data, and a clear financial governance model for the cost of running AI. He describes using AI for growth, such as tracking competitor pricing and customer sentiment by region for Rheem’s largest retail customer so both companies can compete more effectively. On autonomy, he says early agents will keep a human in the loop, and the first fully autonomous agents will work only with data that is neither financial nor private. Within two years, he expects teams of agents to collaborate across value streams in end-to-end processes, which will require HR to decide who governs agents that have become part of the extended workforce. The conversation also covers Rheem’s security-first approach to accountability, where every value stream and application owner shares responsibility for AI risk.
Jeff Kavanaugh:
I'm Jeff Kavanaugh with the Infosys Knowledge Institute. I'm joined today by Joe Palomba, Senior Vice President of Information Technology at Rheem, a leading HVAC manufacturer. Thanks for joining, Joe.
Joe Palomba:
Thank you.
Jeff Kavanaugh:
We're going to discuss today how AI is changing business and delivering value. What is something that your team's doing with AI today that may have seemed impossible even 18 months ago?
Joe Palomba:
I think 18 months ago, we were talking about LLMs and the introduction of LLMs in the industry, right? So if I look at what we're doing now versus 18 months ago, it's really interesting. We've started to roll out Copilot. And I would tell you, every single value stream in the business is using generative AI technology to learn things, to become more efficient and productive in ways that I could never have imagined. On the technology side, it's amazing to see software developers really enhance and become so much more productive with vibe coding and all of the kind of software AI capabilities that are out there. I think finally, us leaning into agentic capabilities, the ability to fuse agents into processes that might be broken or not as efficient. These concepts were unimaginable 18 months ago.
Jeff Kavanaugh:
Where is AI, that you're seeing, delivering value, and especially in an area that your senior executives and board prioritizes?
Joe Palomba:
We're looking at AI from multiple different facets, but one of them is growth, right? How do we use AI technology to be able to figure out organic growth opportunities within the company? So imagine, imagine the scenario where one of our customers is a large retailer, and what's very important for them is to understand how their competitors are pricing their products within the industry. Imagine if we, on behalf of our largest retail customer, could see into regions, markets, and see competitive pricing adjustments that were being made, and also being able to understand the customer sentiment of that pricing decision. Imagine us being able to go back to our customer and show them that data so that us and our biggest retail customer can collaborate more effectively so that we can both win in the market. These are things that we're surprising our largest customers.
Jeff Kavanaugh:
How far along are you moving this AI progress from pilots to scale, and what's been the biggest challenge?
Joe Palomba:
We're still in the piloting kind of stage. To move AI at scale, things have to be set in motion. And what I mean by that is, if our processes haven't been really looked at, if our business processes haven't been reimagined, and we try to layer AI on top of a broken process, that doesn't scale. But I would tell you that we've really started to reimagine process. We really have a sound data governance platform where we're starting to lay AI on top of. I think we're on the cusp of really scaling. Other obstacles, I would tell you, financial governance. You hear the word tokenization. Really understanding the financial governance model around scaling AI is a big deal. Just overall governance of the technology as well is, are things that we're still working on.
Jeff Kavanaugh:
There's the people side of things as well, and how do you help your employees know when AI is appropriate, it's the right tool, and how to use it well?
Joe Palomba:
We've been starting with business case first. We've been starting with value first. We've developed an AI innovation board, an AI governance board, and ideas are kind of cultivated through the innovation board, and then brought forward to the governance board to make sure that cost model makes sense, the value makes sense, and the value can be measured in perpetuity.
Jeff Kavanaugh:
As AI changes the work, let's talk about adoption for a moment. How do you bring people along towards adoption?
Joe Palomba:
It's been top-down and it's been bottom-up. I'll start with top-down. Just recently, we ran an AI master class for 35 executives at Rheem. And the reason for the master class was to really get our executives to understand the art of the possible, to understand how we could innovate, understand how we could not only ascertain efficiency and productivity improvements at the employee level, but really figure out ways how we can grow. I think the executives really left the class kind of wowed. When you go bottom up, we've been kind of selling and articulating the value of tools like Copilot, and we've been, let's say, doling out X number of licenses to all of the value streams. And all of the value streams have been leaning in, learning, doing, imagining, and that has taken foothold. So that is starting to seep further and further into the organization, and now I'm getting pulled. We want more licenses. We want access to this cool technology because people are becoming more efficient and productive.
Jeff Kavanaugh:
Tell me about a situation where humans have been overruling a recommendation and why.
Joe Palomba:
So back to the innovation and governance board. We talk about process, we talk about data, we talk about financial governance. The innovation board brought forth an AI use case. And it seemed cool on the outset, but as you really leaned into what this AI use case was doing, it was really leaning into a bad process. It was really leaning into a data model that wasn't governed. So, catching those things, I think is important.
Jeff Kavanaugh:
When AI does get it wrong, who's accountable?
Joe Palomba:
The world of AI has introduced so much discussion about cyber, right? And security and financial control. I would tell you that there's no one value stream that's accountable. I would tell you we're all accountable, right? And what I mean by that is, our cyber organization would be, let's say, accountable, but what we're working on at Rheem is kind of a security-first mantra, where value stream owners, application owners, we are all security-first mindset.
Jeff Kavanaugh:
Got it. Well, at some point, you're going to continue to progress to where AI is more autonomous. What would it take for you to trust AI to work independently?
Joe Palomba:
We are starting to introduce agents now. But we're being very careful. I would tell you right now the agents that we're starting to think about deploying will be human plus or human-in-the-loop. For it to be kind of headless, I would tell you that out of the gate, it would be probably non-financial data, data that's, let's call it not private data as well.
Jeff Kavanaugh:
Two years from now, how do you envision AI changing the organization, and what do you think its limits will still be?
Joe Palomba:
The whole world of agentic capability will be so much more advanced in a couple of years. What I see happening at Rheem is agents will start to have been deployed within value streams. There'll be agents within agents and teams of agents within value streams. Then, what I foresee happening is those teams of agents within a value stream collaborating with other agents to the left and right of them to start working in an integrated, autonomous, end-to-end business process. I see that happening, and when that begins to happen, that's where you see the HR organization really beginning to lean into who governs the agents, and who polices the agents because the agents really become part of our extended team.
Jeff Kavanaugh:
Switching gears a little bit and stepping back. There's been a good, long relationship with Infosys and Rheem.
Joe Palomba:
Yes.
Jeff Kavanaugh:
Can you tell me a little bit about the work with Infosys and what the partnership made possible?
Joe Palomba:
So Infosys has been with us, I think for about four years, and they've been a terrific partner in, I think, a lot of different ways. So first, they have been providing our ERP Oracle support, right? So they're our first line of defense across most all value streams in the ERP space. So that's point one. Point two, they've really been leaning in with us in, let's call it the engineering and manufacturing thread. So we like to say the digital thread that Infosys is helping us innovate around. They've been relentless in hanging in there with us trying to modernize our R&D capabilities, modernize the way we think about PLM processes and PLM systems, and that's now starting to more from R&D into the manufacturing space. In particular, I'm hoping that Infosys is with us for MES selection and MES implementation moving forward.
Jeff Kavanaugh:
Good, and hopefully, helping solve some of your hardest, most valuable problems.
Joe Palomba:
Yes, 100%.
Jeff Kavanaugh:
Well, thank you so much for your time. I'm Jeff Kavanaugh from Infosys Knowledge Institute. Until next time, keep learning and keep sharing.
