As generative AI, agentic systems, and intelligent automation take on activities that ERP systems were not designed for, organisations are rethinking how decisions are made, executed, and governed.
In this Q&A, Sandesh Bilagi, President and COO of Ramco Systems, discusses how ERP systems are evolving in response to AI-driven workloads and what enterprises should consider as autonomous systems become more prevalent.
Which ERP functions are most vulnerable to AI-driven workloads?
Traditional ERP systems rarely fail in obvious ways. The strain tends to surface gradually, usually at the edges. The first areas to show signs of stress are those built on rigid assumptions, such as fixed workflows, tightly structured data models, and batch-oriented processing.
AI does not operate within those boundaries. It works with ambiguity, learns continuously, and depends on real-time context. When this meets a system designed for predictability, friction is inevitable. Processes that once felt efficient begin to appear restrictive. Data structures that ensured consistency begin to feel incomplete when unstructured inputs come into play.
Another challenge arises wherever ERP relies heavily on repetitive manual tasks. In such areas, users often experience fatigue and become more prone to human error. When AI takes over these activities through automation, performance improves significantly, not only in speed and accuracy, but also in overall operational resilience.
At a deeper level, this is not just a technical limitation. It is a clash of philosophies. ERP systems were built to enforce discipline and standardisation, while AI is designed to question and adapt. When these two coexist, the challenge is not just about scaling performance, but whether the underlying architecture can support a more fluid and intelligent way of operating.
How is AI changing the role of ERP?
ERP is no longer the centre of enterprise activity. It is becoming the anchor that holds everything together.
As AI systems begin to operate alongside, and sometimes outside, traditional workflows, they take on responsibilities that ERP was never designed for. These include prediction, recommendation, and, in some cases, initiating decisions. ERP, in contrast, is evolving into a system that ensures consistency, compliance, and traceability across those actions.
This evolution also changes how users interact with the system. In traditional ERP environments, users were required to learn the system and adapt their behaviour accordingly. As AI becomes embedded, the dynamic reverses. The ERP, augmented by AI, begins to learn from users and guide them intuitively toward faster, better, and more informed operations and decisions.
This creates a layered operating model. Intelligence becomes more adaptive and exploratory, while ERP remains structured and accountable. For someone in an operational leadership role, this shift is significant. It requires a move away from the idea that all decisions must originate within the ERP. Instead, the focus shifts to ensuring that decisions, regardless of where they originate, are executed with integrity and control.
In this sense, ERP becomes less visible in day-to-day interactions, but far more critical in maintaining the overall coherence of the enterprise.
Where do agentic AI systems fail in enterprise environments?
Agentic AI systems most often break down when they rely on a single agent to manage complex enterprise interactions, rather than decomposing tasks into multiple specialised agents with clearly defined goals and boundaries. This increases the risk of inconsistent behaviour and unreliable execution in real-world environments.
Another frequent failure point arises when agentic systems are not deeply integrated with systems of record, limiting their ability to complete transactions end to end and resulting in partial automation that stops short of resolution.
Issues also emerge when AI logic is embedded in rigid scripts or opaque low-code tools, making it difficult for customer success teams to update or refine business logic without relying on engineering teams when conditions change.
Breakdowns become harder to contain and resolve when enterprises lack full observability into AI decision-making. It becomes difficult to understand why a system took a particular action, complicating debugging, auditing, and regulatory compliance.
What factors determine trust in AI-led customer interactions?
Customer trust is strengthened when AI-led interactions are grounded in verified enterprise data and deterministic workflows, and weakened when systems rely on unconstrained language models that can hallucinate or provide confident but incorrect responses.
Trust also depends on whether AI interactions lead to actual resolution, such as completing refunds by leveraging integrations with backend systems to execute changes, rather than stopping at surface-level responses.
In addition, transparency plays a critical role. Enterprises that can explain how and why an AI system took a specific action are better positioned to maintain customer confidence when outcomes are questioned or disputes arise.
What are the trade-offs between flexibility and standardisation in ERP?
Flexibility is often seen as essential for innovation, but in ERP environments, it introduces complexities that cannot be ignored.
As systems become more flexible, maintaining consistency becomes more difficult. Data definitions start to vary, processes begin to diverge, and the idea of a single source of truth becomes harder to sustain. Over time, this affects not just system performance, but also the quality and reliability of decision-making.
At the same time, excessive standardisation can create systems that are efficient but rigid. They struggle to adapt to new business models or incorporate emerging technologies such as AI in a meaningful way.
The answer lies in balance, but not in a superficial sense. Organisations need a clearly defined core where standardisation is non-negotiable, especially in areas such as financials, compliance, and master data. Around this core, there must be the ability to introduce controlled flexibility, allowing innovation without compromising stability.
This discipline must extend beyond systems into organisational behaviour. Uniformity across the organisation, both between and within departments, is essential to ensure that standard operating procedures are consistently maintained, even as flexibility is introduced at the edges.
Ultimately, this is less about technology and more about discipline. The organisations that succeed are the ones that are deliberate about where they allow variation and where they enforce consistency.
How should enterprises approach accountability for autonomous systems?
Accountability should be established before autonomous systems are deployed rather than after problems emerge.
Organisations remain responsible for decisions made by AI systems and therefore need clear governance frameworks that define objectives, constraints, and decision boundaries. These frameworks should also specify when systems can act autonomously and when human oversight is required.
Comprehensive logging, audit trails, and explainability mechanisms are increasingly important. They allow organisations to reconstruct decision paths, demonstrate compliance, and investigate incidents when necessary.
Many enterprises are adopting models in which autonomous systems handle routine or low-risk activities, while more complex, sensitive, or ambiguous cases are escalated to human decision-makers. This approach helps balance efficiency with accountability and risk management.
