AI at scale starts with ERP that can evolve

Insights

• Enterprises want AI to deliver real business value, not just pilots.
• Research shows most AI use cases fall short without strong data, processes, and operating foundations.
• ERP systems are critical because they provide governed data and end‑to‑end process context for AI.
• Traditional, heavily customized ERP has slowed AI adoption and scaling.
• Modern cloud ERP with continuous innovation enables AI to be embedded, scaled, and sustained across core operations.

As AI moves from experimentation to execution, enterprises are under pressure to deploy it quickly, scale it safely, and integrate it into day‑to‑day operations. Leaders expect AI to improve decisions, automate routine work, and help organizations respond faster in a changing business environment. As a result, pressure is growing to move beyond AI pilots and show lasting business value.

Research from Infosys AI Business Value Radar 2025underscores both the progress made and the challenges organizations face with implementing AI. This study of more than 3,200 companies and over 130 AI use cases across Europe, the US, Australia, and New Zealand shows that while AI adoption is growing, only between 19% and 20% of use cases fully meet business objectives, with another 30% or so delivering partial value.

Enterprise resource planning (ERP) systems sit at the center of this readiness question. As systems of record for finance, supply chain, manufacturing, HR, and procurement, they provide the transactional data and process context required to embed AI into core business workflows, as long as they can evolve fast enough to absorb continuous change. This challenge is increasingly being addressed through modern cloud ERP platforms such as Oracle Fusion Applications.

Why ERP matters for enterprise AI

Enterprises increasingly expect AI to operate within core business processes, influencing decisions, triggering workflows, and automating execution across functions. ERP systems support this ambition because they provide two critical assets that most AI initiatives lack on their own:

Governed transactional data: ERP systems capture data from everyday business activities like sales, payments, inventory updates, employee records, and customer orders. Because this data is checked and controlled, it can be trusted and linked back to real business actions.

End-to-end process context: ERP systems show how work moves across the organization, from receiving an order to getting paid, from buying supplies to paying vendors, from planning production to delivery, and from hiring employees to managing their careers. This helps AI understand how different activities are connected and how decisions in one area affect results elsewhere in the business.

Together, these capabilities allow AI to work reliably inside core business processes. When AI uses ERP data and workflows, it can support everyday decisions, automate routine tasks, and highlight risks in areas like inventory, finance, and workforce planning. ERP therefore serves as the foundation for applying AI across enterprise operations.

Why ERP has historically held AI back

Even though ERP has the potential to support AI, many organizations struggle to make ERP work in practice. One reason is that ERP strategy is often not well aligned with business goals.

In the past, ERP systems were updated only through large, high-risk upgrades that happened every few years. These upgrades took a lot of time and effort and often disrupted normal business operations. As business needs changed, organizations added custom code and integrations to fill gaps, which made ERP systems more complex and harder to change over time.

Because of this, even small changes became slow, costly, and risky. Adding new capabilities often required changes across many parts of the system, increasing the chance of errors or disruption. As a result, organizations focused on keeping systems stable rather than introducing new ideas, and AI efforts often stayed limited to small pilots. These limitations became more obvious when organizations started exploring AI use cases that require frequent updates and clean, well-connected data.

These constraints became more visible as organizations moved ERP systems to the cloud. While cloud migration improved infrastructure and scalability, it did not automatically make ERP environments ready for AI-driven capabilities. Companies kept the same customizations, maintained fragmented data models, and continued long, infrequent release cycles. So, while cloud migration improved the infrastructure, it didn’t actually make it easier to adopt or scale AI.

This is because moving ERP to the cloud doesn’t automatically fix data quality, standardize processes, or make change easier. If master data is inconsistent, processes aren’t aligned, and releases aren’t well managed, ERP systems remain hard to improve. Adding AI on top of this requires heavy data cleaning, custom integrations, and manual oversight. This makes AI expensive to build, hard to scale beyond small pilots, and difficult to maintain over time.

ERP modernization on its own is not enough to support AI. What is needed is an operating model built for continuous evolution, one that reduces complexity, improves data quality, and enables frequent, low-risk change. Only then can ERP systems support AI at the pace and scale enterprises now expect.

Continuous innovation as the foundation for ERP-led AI

To scale AI successfully through ERP, organizations should manage ERP systems as a platform that improves regularly through small, predictable updates.

Three elements are essential to drive continuous innovation:

Standardized processes: Reducing unnecessary customizations and aligning processes to standard patterns lowers complexity and increases consistency. This makes it easier for AI to learn, reason, and act across the enterprise.

Clean, governed data: Harmonized data models, clear ownership, and built-in rules and controls embedded directly into the ERP processes and data structures ensure AI operates on reliable inputs. This reduces risk and increases trust in AI-driven decisions and sets up a foundation for decision-making.

Predictable, low-risk change: Predictable release cycles allow organizations to absorb innovation in manageable steps. Smaller, more frequent changes reduce disruption and create organizational resilience in adjusting to future changes. It also helps organizations plan better for user adoption of these incremental changes.

This operating model is already being applied in modern cloud ERP platforms. Oracle Fusion Applications provide a practical example of how a continuous innovation ERP model works. The platform is built so that it is updated regularly through predictable 90-day release cycles. Each release introduces small, incremental improvements, which organizations can adopt without going through large, disruptive upgrade projects or significant additional effort.

Over time, this approach reduces custom code and keeps ERP systems up to date. Instead of waiting for major upgrades, organizations can choose when to turn on new features, including functional improvements, analytics, or AI, based on what the business is ready to adopt.

This operating model creates an ERP foundation that is designed to support AI over time. Embedded and agent-based AI capabilities can be introduced gradually, aligned with business priorities and user adoption readiness. As a result, AI is introduced as part of regular ERP evolution, rather than as a separate initiative that disrupts existing systems and processes.

Case study: Scale AI through continuous ERP innovation

This operating model is illustrated through a global ERP transformation at a large US utility company, delivered by Infosys and leveraging Oracle Fusion Applications as the core ERP platform.

Like many large enterprises operating traditional ERP models, the organization found its ERP environment difficult to change. Enhancements were delivered through large, infrequent projects, making it costly and risky to introduce new capabilities such as AI. As a result, early AI efforts were limited in scope and difficult to scale across regions and functions. The organization struggled to keep up with the latest capabilities in ERP and supply chain due to these long and cumbersome upgrade cycles.

Rather than continuing with disconnected AI experiments, the organization focused on changing how ERP itself was modernized and operated. The transformation established continuous innovation using Oracle Fusion’s quarterly release updates, allowing new capabilities to be introduced incrementally as part of regular ERP updates rather than disruptive upgrades.

AI adoption began with Oracle Fusion’s embedded generative AI features in procurement and supply chain, focusing on reducing manual effort and improving everyday decisions. Later, AI was extended to support buyer and supplier interactions, such as negotiation messaging and supplier discovery. These capabilities were introduced gradually through regular releases, aligned with business readiness.

The next phase introduced AI agents within Oracle Fusion. Early agents focused on improving access to information and reducing reliance on manual support, while later, more advanced agents were able to take actions, not just provide answers. Using AI agents during the costing-period close helped the company make the month-end close process smoother and faster, while also spotting inventory and costing issues early — before they became bigger problems.

What organizations need to do to implement ERP-led AI adoption

Turning intent into a successful transformation demands clear goals and careful planning:

Treat AI as an outcome of ERP modernization: Position AI as the result of sustained ERP evolution rather than a separate technology program. AI scales more reliably when it is introduced on top of standardized processes, high-quality governed data, and stable core systems that are already designed to absorb continuous change.

Operate ERP as a continuously evolving platform: Move away from infrequent, large upgrades toward predictable release cycles. Limit customizations, retire obsolete extensions, and actively reduce technical debt to minimize both risk and disruption.

Strengthen data and process foundations: Establish consistent master data, aligned process definitions, and clear governance across functions and regions. These foundations ensure AI models operate on reliable inputs and produce outputs that are explainable, auditable, and trusted by the business.

Introduce AI incrementally: Deploy AI capabilities in manageable steps through regular ERP releases, starting with narrowly scoped use cases. Align adoption to business priorities and organizational readiness to avoid change fatigue and improve long-term sustainability.

Focus AI on operational impact first: Prioritize use cases that improve day-to-day decisions, automate repeatable activities, and increase responsiveness within existing workflows. Embedding AI where work already happens accelerates value realization and drives user adoption. These operational impact areas also support stronger user adoption and help build a base of users who can advocate for future changes.

Conclusion

ERP systems have long been viewed as slow-moving backbones of enterprise operations. Yet in a world where AI is becoming central to competitiveness, ERP’s role is being redefined, beyond a system of record When governed data, process context, and continuous innovation come together, ERP becomes a system of intelligent action that aids with real-time decision-making. The organizations that succeed with AI will be those that modernize ERP not as a one-time project, but as a sustained capability.