Oracle NetSuite on How AI-Native ERP Is Redefining Business Operations

Oracle NetSuite is rebuilding its ERP platform around a new form of collaboration. Not just a user and a computer, but a user, an application, and an AI that can talk to both. That shift sits at the center of NetSuite Next, the company's next-generation release, and it reframes what an AI-native ERP system actually does.

At The Six Five Summit: AI Unleashed 2026, Keith Kirkpatrick and Nick Patience welcomed Brian Chess, Senior Vice President of AI, Product, and Technology at Oracle NetSuite, for an Enterprise AI Software & Agents Spotlight interview focused on how AI-Native ERP is redefining business operations.

Chess laid out a working test for what counts as AI-native: is the AI essential to how the system functions, or is it a feature a user can ignore and still get through the day? For ERP specifically, he tied that test to three questions: does the AI understand the data model, can it flex the application's functionality, and does it meet the user where they are instead of asking them to translate their request into system language first? That principle shaped the engineering work behind NetSuite Next, where AI had to communicate directly with the application layer, not just the user, so it could act on existing functionality rather than inventing new software on the fly.

On trust and autonomy, Chess drew a direct comparison to how organizations build confidence in new employees. He pointed to NetSuite's anomaly detection capability as a concrete example: the system can flag a transaction that doesn't match the pattern of everything around it and show the user what the surrounding data looks like, but it stops short of correcting the record itself. That human-in-the-loop boundary, he said, is where trust gets established today, one task at a time, the same way a new hire earns responsibility gradually rather than all at once.

Looking at multi-agent systems, Chess described NetSuite's current approach through the lens of AI skills, packaged units of knowledge, instructions, and tools that agents can draw from, and framed the open question as whether specialization or generalization produces better outcomes at scale. He also pointed to early signals from NetSuite's AI Connector Service, which lets users connect tools like Claude and ChatGPT directly into the platform, describing user reactions that go beyond productivity gains into a sense of empowerment to attempt work they wouldn't have tried before.

Key Insights:

🔹 AI-native has a specific test, not a feature checklist. Chess defined it as whether the AI is essential to the system's function or a layer a user can bypass entirely. For ERP, that means the AI has to understand the data model and act within the application rather than sit beside it.

🔹 NetSuite Next was built around a three-way collaboration model. The engineering shift wasn't adding a chatbot. It was giving the AI a direct line to the application layer so it can execute using existing functionality instead of generating new solutions outside the system.

🔹 Trust with agents is built the same way it's built with employees. NetSuite's anomaly detection flags irregular transactions and shows supporting context, but leaves the correction to a human. Chess frames that boundary as the current stage in an incremental trust relationship, not a permanent limit.

🔹 Agent specialization is still an open design question. NetSuite is testing AI skills as a way to package knowledge and tools for agents, and Chess said the company is actively weighing whether narrow, specialized agents or broader generalist agents produce better outcomes across finance and operations.

🔹 Users connecting outside AI tools are showing a different kind of response. Through the AI Connector Service, which links NetSuite to tools like Claude and ChatGPT, Chess said users report feeling empowered to attempt work they previously avoided, not just faster at tasks they already did.

Chess pointed to accountability as the constant in this shift: humans keep ownership of the outcome even as AI takes on larger pieces of the work required to get there.

Watch the full video at sixfivemedia.com, and subscribe to our YouTube channel so you never miss an episode.

Disclaimer: Six Five Media is for information and entertainment purposes only. Over the course of this video, we may discuss companies that are publicly traded, and we may reference their equity share prices. Nothing discussed during this webcast should be considered investment advice or a recommendation to buy or sell any security. We are not investment advisors, and you should not rely on this content as financial advice. Six Five Media collaborates with technology companies and industry leaders to produce research-driven interviews and multimedia programming for enterprise technology audiences.

‍

‍
Brian Chess:
My first intuition when I ran into AI problems is I would look for ways to solve them out of my old toolkit. And what I've figured out is I think the AI native way to go is AI problem, you need more AI because that's the solution to all of the problems.

Keith Kirkpatrick: 

Hi everyone, and welcome to the Six Five Summit, AI Unleashed 2026. I'm Keith Kirkpatrick, VP and Research Director for Enterprise Software at the Futurum Group. I'm joined by my colleague and co-host Nick Patience, VP and practice lead for AI platforms. For this enterprise AI software and agent spotlight, we're exploring how AI is reshaping enterprise applications and why the next generation of ERP is being built AI first. Joining us today is Brian Chess, Senior VP of AI, Product, and Technology at Oracle NetSuite. Brian, welcome to The Six Five. Thanks. Thanks for having me. So, Brian, for years ERP has served as the system of record for business. How does the emergence of AI, particularly generative and agentic AI, change organizational expectations around the role of an ERP platform?

Brian Chess: 

Well, I think before we got to talk about ERP, we should talk about like what's changed about expectations for computers? Sure. So, you know, before computers were fast and they were accurate, but they were also like super duper literal and inflexible. And humans, comparatively speaking, were kind of slow and kind of unreliable. They didn't have that big a working memory, but they were really flexible and creative. And they had properties like ego and ambition. And what that meant is what we did with our systems when we put computers and people together, is we said, okay, people, you need to adapt to the way the computers work because you're way more adaptable than they are. And so now what's happened with generative AI is computers can be a lot more flexible and even maybe like a little bit creative. And so where we're going is a different balance between computers and people. So we're going to have systems that are much, much easier to use and capable of handling a lot more ambiguity than they could in the past. So the humans are still going to bring the lion's share of the creativity. They are going to continue to have more context than the computers do, although the computers are going to have a lot more context. And the humans are still going to bring all of the ego and the ambition. And so what's going to happen to ERP systems in particular then? So what have those ERP systems been? They've been systems that record information about all aspects of the business and then report it back out when people want to understand what the state of the business is or need help in trying to make a decision. So ERP and AI go together really, really well because we've already got these systems where we're accumulating a lot of context. And that AI can watch as things continue to change and evolve and can almost be continuously learning about the state of the business. So what should people expect? They should expect systems that are a lot easier to use than the ones they've used in the past. and that begin to leave the box of just recording and then spitting that information back out and actually start to play a role in the decision making of the business.

Nick Patience: 

So that's interesting. But there's obviously a distinction, Brian, between adding AI features to existing software and then building AI in the foundation level into the platform. So what would you say really means for ERP to be AI native? Here's one good witness test.

Brian Chess: 

Is the AI an essential part of the system? If it's an essential part of the system, then I think I'd say that's AI native. Or is it kind of a nice to have? Where, you know, it's got this AI feature over here, but you could just kind of ignore it and get on with your day and work just fine. I'll tell you one thing that I see in terms of mindset shift with people. I mean, certainly I was not born an AI native. And so I am learning as I go. I'll tell you a little bit about how my mindset is shifting is these days. Well, maybe my first intuition when I ran into AI problems is I would look for ways to solve them out of my old toolkit. And what I've figured out is I think the AI native way to go is AI problem? you need more AI, because that's the solution to all of the problems. So for ERP in particular, I think it's about, does the AI know how to take advantage of the data model? Does it know how to flex the functionality of the application? And does it know how to meet the user where they're at? Or does it keep asking the user to do things that the AI knows how to do already?

Keith Kirkpatrick: 

You know, Brian, you raise an interesting point. I mean, when we think of like, The popular narrative around AI native, a lot of times people just think of that as being very surface level, human to computer interface type agents. How much of NetSuite's AI investment is actually going into the data layer to really provide additional context about workflows, about processes?

Brian Chess: 

We're in the process of launching NetSuite Next. It's the next generation of NetSuite. We needed to make some significant changes, and I'll tell you why. Actually, the data layer, I think we were already in a fantastic system because we had a suite where we really did combine different sources of data and get everything together in one place. But all of a sudden, we needed a new form of collaboration. That collaboration is now not just a user and a computer, it's a user, an application, and the AI. And what we needed to have happen is that the AI needed to be able to talk directly to the user, but the AI also had to be able to talk to the application. Because when the user says, hey, here's what I want to do, we don't want the AI just making up a whole new piece of software. We want it leveraging what the application can already do to get the application to meet the user's needs. So a lot of the work that went into NetSuite Next was getting AI embedded in such a way that it could relate to the application. And we get a lot of benefits out of it. One of them I've already talked about some, it's that the user doesn't have to understand as much about what the system expects, because it can just ask for it at once. And either the AI can translate that into something it can do, or it can explain to the user, here's what I need from you in order to meet your request. And so now they can negotiate a solution to the problem, whereas before people were just starting to thumb through the docks. So I think the big trick here, and we're going to see a lot of this over the next couple of years, is not just how do we make these really capable, amazing new AI powered systems, how do we get people there? How do we get people transitioned out of the old way of working? Show them what's possible and get them comfortable in that new world. And so I think that's an exciting opportunity and it's probably a lot of what we're gonna see over the next couple of years.

Nick Patience: 

Brian, one of the other things we expect we're going to see a lot more over the next couple of years, and we've been hearing a lot about in the last 18 to 24 months, obviously, is agents. And you're talking about talking to the software and that kind of metaphor. So that sounds to me quite agentic, and it's obviously become such a kind of a massive issue. What role do you see agents, intelligent agents, specifically playing within ERP?

Brian Chess: 

I think that the power is immense. I think that the number of things that you can reduce to a set of natural language instructions, hands to an agent, and then automate a portion of the business, there's tremendous power there. But there's also a real problem to be solved. And that is, ask somebody, hey, how does this process work in your business? What are they going to tell you? They'll tell you, they'll give you a list of things, in most cases. How closely does that list correspond to what actually happens in the business? A lot of times it's imperfect. Why? Because these businesses are complicated things. And in many cases, no one person knows exactly all of the things that happen in all the scenarios. So in order to make agents successful, we have to have a way for people to explain what they want to do and then change what they want to do as they see the outcomes of their decisions. So I think we've got a tremendous amount of technology available to us right this minute, but we've got a lot of work to do on how do we use that technology to enable this next level of automation that we can see.

Nick Patience: There's also a kind of issue, I guess, around that next level automation, how much autonomy we choose to give to those agents. You were talking at the beginning about if AI is all about having to rely on people to tell it what to do, then there's not much point in some sense. But again, with agents, it could go the other way, isn't it? Where are we in that kind of journey, do you think, in terms of being able to trust autonomous agents? Is it still extremely early?

Brian Chess: 

Well, I think that the word you used towards the end there was maybe the most important word to me, and that is trust. How do these systems, how do they earn their place? What should you do in order to establish trust? I think that's a very, very important question. And I think it's going to be a lot like the way we establish other trust relationships. It's not going to be, okay, let me close my eyes and push the button. It's going to be, let me work with this thing. Trust is not binary. I may trust the system for some things and not for others. I need to have a way to express that. Well, it turns out we actually have been doing this with employees for a very long time. We trust them to do some things and we don't trust them to do others. And how do we refine that view? We work with them. maybe giving them more responsibility over time. And so I expect that that is going to continue with agents. We see this when people are adopting AI. The very first things they ask are analysis questions. because they don't trust that AI to write. The first thing they trust it to do is read. So they'll ask it to analyze something for them. And then they'll do a little more analysis. And they'll get the sense like, oh, it really knows what it's talking about in this area. OK, now I trust it more. And we'll get to the point where, all right, now you can start writing some data too. And that's establishing that trust relationship. I will tell you this kind of as a rule of thumb. I think that the more sophisticated the AI becomes, the more apt comparing the AI to how we deal with humans makes sense. And so that I think it's another thing that gives people very good intuition about these systems where sometimes software could be difficult for people to develop intuition about. The AI is going to be not necessarily easy, but easier, especially as it becomes more sophisticated.

Keith Kirkpatrick: 

Is there a specific agent right now that NetSuite has shipped or is perhaps close to shipping that actually is allowed to make some sort of a decision or take an action without a human in the loop? Or is that even desirable at this point? I mean, do we need to go through that process that you just discussed about taking sort of a very stepwise approach to rolling out autonomous agents?

Brian Chess: 

I don't think that there's any agent that we ship today or will ship in the near future that's just going to start pulling the levers for you. We think that developing that trust relationship still means you begin with a human in the loop, which means that we will come and suggest things. I'll just give you an example. we're pretty good at spotting anomalous transactions, things that just don't look like the other stuff. And in many cases, we can actually tell you probably like, hey, here's what everything else looks like. This is probably what this thing should look like too. but we don't go fix it for you. We make it really, really easy to fix, but we still expect that we want a human in the loop when they go and they review exceptions in their transaction set.

Keith Kirkpatrick: 

Brian, that's a really interesting point in terms of making sure that these agents are doing what they're supposed to be doing based on how humans are essentially overseeing it. I'm curious what you see in the future when it comes to something that we're hearing a bit about now, which is multi-agent orchestration, where you have one agent perhaps starting a process and then others picking up the ball. Do you see a future in that and what needs to happen in order to do that responsibly and in a way where we can still trust these processes?

Brian Chess: 

So how many agents should you have? Is it desirable for you to have 101 agents or do you want one great big agent? I would say that eventually is a technology question, but it's another one that we grapple with with employees too. When do you want a specialist and when do you want a generalist? And so I think there are a lot of good reasons and it's remarkably simple to do to get agents to start collaborating. But where do we put those boundaries? Is it important that we should have one agent centered on financials and we should have another one centered on the warehouse? Or should we be blending those two things together? Because eventually it is one big optimization problem. I think we're still in an exploratory state there where we're trying to say, well, if we have this agent specialized, does it over-specialize? Does it over-optimize on one aspect? If we want a team of agents, do we want to model that like we would model a human team? Or do we really want to slice this very, very finely and have specialist agents in very, very micro categories? I think one of the ways that we're exploring this right now is with AI skills, where we're figuring out how to package up units of knowledge, instructions, tools that agents can use. And then we're figuring out, is it better to give an agent 10 skills or to have it specialize in one skill? But I think this encapsulation of intelligence into that skill has been a really important step forward in letting people experiment and also optimize for their use cases.

Nick Patience: 

Brian, if we think about those skills and how we evolve with those and you look forward, say, three, maybe even five years, what does a NetSuite user in finance or operations, what does their job actually look like day to day inside the kind of AI native ERP system?

Brian Chess: 

At the beginning of our discussion, I talked a little bit about what I think the humans still have a really good lock on. I think the humans are still more creative because they have more context. We have supplied a lot of context to the AI. The humans still have more, but they also have the ambition and the ego. And so I think that's going to have humans doing fewer pulling of the levers, but more guidance of the business about what is the thing that we want to see happen next. I'll make a prediction. Over the next couple of years, Maybe you think it's happening a lot already, but I think it's going to happen even more. We're going to use the word outcome a lot because we're still going to hold the human accountable for that outcome. We don't get to pass it off on the AI, but we are going to expect AI to deliver larger and larger components of that outcome once the human has directed the system about what it wants.

Keith Kirkpatrick: 

That makes a lot of sense, Brian. Building on that, final question for you. What does that mean for NetSuite moving forward in terms of what you're building, what you're bringing to the market?

Brian Chess: 

Well, I'll tell you something that's really interesting that we've seen happen as we roll out Next, as we have people using our AI Connector Service to connect like Claude and ChatGPT. We see something that I have not seen before in business software, and that is we see users with a very emotional response to their experience. And I say that because it's not just that they feel more productive in their jobs, it's that they feel empowered in their jobs. And so they all of a sudden are exploring things. Of course, everybody says, so it tells their boss like, Oh, I'm more productive now. But the truth of the matter is they're beginning to do things that they would not have attempted before. So it's not just a matter of, I got the work done in an hour instead of a day. It's a matter of, I have now explored paths that I knew about before, but they were basically unavailable to me. So I think we're going to see for NetSuite users that they are making more advantage of what NetSuite gives to them, because now it's more accessible to them, and that they're going to build on that feeling of empowerment to achieve more. So I think it's an incredible time to be running a business or to be making business software.

Nick Patience: 

That's great. Well, thanks very much, Brian. And thanks, Keith. Thank you. That wraps up this enterprise AI software and agent spotlight at the Six Five Summit AI Unleashed 2026. Thanks for joining us. Don't forget to subscribe. Follow us on socials and visit SixFiveMedia.com for more conversations and insights from across the summit. We'll see you next time.

‍