Dominic Jainy has spent the better part of his career dissecting the friction between human intent and machine execution. As an IT professional deeply embedded in the worlds of machine learning and blockchain, he has watched ERP systems evolve from the rigid, clunky databases of the early 2000s into the sophisticated, AI-driven ecosystems we see today in 2026. With major players like SAP, Oracle, and Infor recently unveiling groundbreaking agentic frameworks, the conversation has shifted from “what can AI do” to “how do we govern what AI is already doing.” In this discussion, we explore the collapse of static user interfaces, the resurgence of “unlimited” customization through agentic layers, and the high-stakes battle between established enterprise giants and the frontier AI models threatening to turn them into mere background APIs.
The dialogue covers the four major trends currently reshaping the industry: the completion of the AI-driven user experience, the tension between SaaS standardization and AI-led personalization, the competitive moat created by deep business logic, and the critical importance of governance in an autonomous enterprise.
We are seeing a massive shift where ERP systems are moving away from static interfaces and toward purpose-driven, natural-language engagement layers. How does this transition fundamentally change how a typical employee interacts with the company’s “system of record”?
The shift we are witnessing is nothing short of a total pivot in the psychology of work. For years, employees derided these systems as complex labyrinths where you had to go to a specific location, click through five tabs, and enter data in a very rigid format just to complete a single task. Now, with the expansion of tools like Joule Work, we are moving into a purpose-driven environment where the software comes to the user rather than the user hunting for the software. Instead of a static set of screens, you have an engagement layer that translates your native language into direct action. It feels less like operating a machine and more like collaborating with a specialized colleague who already knows the context of your request. It’s an event-based experience—if a supply chain disruption occurs, the interface adapts to that specific situation immediately, rather than forcing you to pull up three different reports to understand what happened.
Oracle recently made waves by unveiling its Fusion Claw software, which includes 25 agentic AI applications right out of the box. What is the significance of “agentic” AI compared to the chatbots we’ve been using over the last few years?
The distinction is critical because it marks the move from “talking” to “doing.” While early generative AI assistants could summarize a document or draft an email, agentic AI like Fusion Claw actually possesses the agency to execute multi-step business processes within the system’s built-in governance and security. These 25 applications aren’t just there to answer questions; they are designed to perform tasks like reconciling accounts or managing skills gaps autonomously. The real genius here is that these agents are built with the same authorizations and authentication protocols as human users, meaning they can navigate the metadata of the enterprise without getting lost or misdirected. This level of autonomy is what allows an organization to scale AI across the enterprise because the system isn’t just suggesting an answer—it is actually pulling the levers and turning the gears of the business.
Infor has taken a slightly different path by focusing on industry-specific “micro-verticals.” How does their “Industry AI” architecture solve the problem of generic AI models that don’t understand the nuances of a niche business?
Infor’s strategy is a direct answer to the frustration many C-suite leaders feel when they realize a general-purpose model doesn’t know the difference between a bolt for an Aston Martin and a component for a cooling system at Team Air Distributing. By spotlighting these industry leaders, Infor is proving that their AI agents are “born” with specialized expertise baked into the architecture. Their Velocity program gives customers essentially unlimited access to a suite of technologies designed to work with third-party applications, which is vital because no business exists in a vacuum. It’s about building a UX that adapts to individual user needs while ensuring that every action taken by an agent passes through a rigorous governance, risk, and compliance layer. This micro-vertical focus ensures that the “hallucinations” common in frontier models are filtered out by the sheer weight of industry-specific logic.
There is an interesting tension emerging regarding customization. SaaS was supposed to end the era of messy, hard-to-maintain custom code, but now vendors are pitching “unlimited” AI customization. Are we at risk of repeating the mistakes of the past?
That is the million-dollar question, and there are two very vocal sides to this debate. On one hand, you have leaders who believe there is no such thing as “out-of-the-box” AI, arguing that every company’s logic is a competitive advantage that must be reflected in the code. They see AI as a way to build extensions and change the user experience without being tethered to expensive consulting services. On the other hand, skeptics warn that this could lead to “vibe coding,” where employees build overlapping or redundant apps that ignore established best practices. We’ve already seen cases where a customer used an AI agent to build a custom procurement tool that actually duplicated a standard feature already available in their software. If we lose the discipline of the “clean core”—keeping the ERP untouched while building extensions on a separate platform—we risk creating a digital landfill of unmanageable, customized agents that will be a nightmare to maintain in the years to come.
You’ve mentioned that ERP vendors are now competing more with frontier model makers like Palantir and Anthropic than with each other. Why is it so difficult for these “outsider” AI companies to provide a true alternative to a native ERP system?
The outsiders are running into what I call the “metadata wall.” Companies like Palantir or Anthropic can offer incredibly slick user experiences, but they often lack deep, native access to the business logic and historical metadata that lives inside an ERP system. SAP’s leadership has been very vocal about the fact that they don’t want to become just an “API call” for someone else’s AI. When you try to “screen scrape” your way into an enterprise, you miss the deep nuances of how data and authorizations interact. I’ve seen customer situations where these frontier models play fast and loose, promising quick results, but they ultimately fail because they don’t understand the “why” behind the data. An ERP vendor is steeped in the historical understanding of what makes a business tick, and that is a moat that is very hard for a general-purpose model to cross.
Security and governance seem to be the final hurdles for widespread adoption. How are vendors like SAP convincing their most cautious clients, such as Novartis or Morgan Stanley, that an autonomous enterprise is actually safe?
It comes down to treating agents with the same scrutiny we apply to human employees, or perhaps even more. SAP brought their Chief Security Officer, Marielle Ehrmann, onto the main stage specifically to drive home the point that agents without proper governance can do significant harm without any malicious intent. The governance promise now includes the end-to-end lifecycle of every agent, whether it was provided by the vendor, built by a partner, or customized by the client. These agents operate within the existing security frameworks—the same logins, the same permissions, and the same audit trails that have protected the system of record for decades. Organizations are being told that they must get their data “houses” in order first; you cannot build a skyscraper of agentic AI on a foundation of messy, unmanaged data.
What is your forecast for the evolution of the autonomous enterprise over the next few years?
The next two to five years will be defined by the “Great Foundation Building,” where the hype of agentic AI meets the reality of data discipline. We will see a consolidation of interfaces, where the traditional ERP menu disappears entirely, replaced by a conversational layer that predicts user needs based on real-time events. However, the real winners won’t be the ones with the flashiest AI; they will be the companies that maintained a “clean core” while successfully layering in custom, governed agents that enhance their unique business processes. We will likely see a significant shakeout among frontier AI providers as businesses realize that “vibes” and general logic cannot replace the deep, structured intelligence of a system that actually owns the data. Success on any timeline is going to depend on how well an organization can balance the urge for unlimited AI customization with the cold, hard necessity of centralized governance.