Manish Goyal, Co-Founder of Dynamics Square, drives digital transformation via AI & Dynamics 365 with 17+ years of proven experience.
getty
Most conversations around agentic ERP focus on what AI can do. But in many ERP modernization discussions I have been part of, the real problem appears much earlier. Companies often chase AI capabilities before defining the operational problems they are actually trying to solve.
McKinsey’s latest State of AI research found that 88% of organizations now use AI in at least one business function. Yet only about one-third have begun scaling AI initiatives at the enterprise level, with most companies still operating in experimentation or pilot stages.
Leadership teams get excited about intelligent automation features while ERP providers demonstrate AI-driven workflows and copilots. In that environment, AI does not solve operational inefficiencies. It tends to magnify the operational problems already present in the business. In many cases, the technology is not the primary issue. The bigger problem is that organizations expect AI to optimize workflows that were never operationally consistent to begin with.
During one manufacturing ERP modernization discussion, I saw how years of process layering had gradually complicated even routine procurement decisions. Every operational exception introduced another approval step, eventually making routine purchasing activities dependent on multiple managerial checkpoints before action could move forward.
The workflow appeared controlled, but it slowed procurement and created dependency on a small group of decision-makers. The larger issue was not the absence of AI capabilities. The real constraint was the operational structure surrounding the process itself.
Many organizations are investing in agentic AI capabilities while still relying on workflows built around manual oversight, layered approvals and process dependency. According to McKinsey & Company, organizations that redesign workflows alongside AI adoption are significantly more likely to realize measurable business value. The firm also found that only 21% of companies using generative AI have meaningfully redesigned workflows so far.
Another challenge that surfaces frequently during ERP modernization is excessive customization carried forward from older on-premise systems. In many organizations, layers of process modifications gradually accumulate around departmental preferences, operational exceptions and legacy working styles.
As a result, different teams often end up following slightly different versions of the same workflow across the business. Leadership teams may still expect AI systems to operate consistently in these environments, but fragmented process structures rarely produce reliable outcomes.
I have also seen organizations remain heavily dependent on Excel even after ERP modernization because workflows and data structures remained inconsistent across departments. When AI capabilities are introduced into environments like these, user trust quickly becomes a challenge. Employees hesitate to trust AI recommendations when operational structures already lack consistency.
This is why I believe many companies approach agentic ERP in the wrong sequence. They focus heavily on intelligent features before addressing operational readiness, process standardization and governance discipline. Deloitte recently noted that layering AI agents onto broken processes does not resolve operational inefficiencies—it often magnifies them.
The organizations seeing the strongest AI results are not necessarily the ones deploying the most features. Three areas, in particular, deserve far greater leadership attention:
In many organizations, operational delays are less about technology limitations and more about approval structures that became increasingly layered over time. As workflows evolve, additional checkpoints tend to get introduced to manage exceptions, improve oversight or reduce risk. Over time, however, those same structures can slow routine decision-making and create unnecessary process dependency.
One valuable exercise during ERP modernization is identifying where work consistently stalls without adding business value. In several transformation discussions I have been involved in, simplifying approval structures created faster operational improvements than introducing additional automation capabilities.
When organizations reduce unnecessary dependency within workflows, autonomous systems are better positioned to support execution, responsiveness and decision-making at scale.
AI systems generally perform far more effectively in environments where operational rules, workflows and data structures are consistent across the business.
While that may sound straightforward, many organizations underestimate how much process variation gradually develops between departments. I have seen situations where procurement, finance and operations teams were all using the same ERP platform while still following different workflow structures, approval paths and reporting practices.
Before expanding AI-driven execution, leadership teams should focus on simplifying process structures, reducing unnecessary customization and improving data consistency across functions.
That also helps explain why many enterprise AI initiatives continue struggling to produce measurable business outcomes. According to MIT, researchers found that 95% of generative AI implementations failed to create meaningful P&L impact because AI capabilities were introduced into fragmented operational environments without sufficient workflow integration.
In many ERP modernization initiatives, the more difficult challenge is often organizational rather than technical.
Many managers still associate operational control with approvals, manual checkpoints and direct oversight of day-to-day decisions. But as AI capabilities become more embedded into ERP environments, leadership responsibilities gradually shift toward governance, exception management and outcome monitoring instead of transaction-level supervision.
That transition can be difficult for many organizations. I have seen operational teams resist standardization efforts because managers were concerned about losing visibility into decisions or reducing their influence over existing workflows. In many cases, the resistance reflects uncertainty around changing responsibilities and decision ownership.
Technology alone does not resolve those concerns. Leadership alignment, operational clarity and change management usually matter far more during the transition.
Most organizations will eventually incorporate AI capabilities into their ERP environments. The larger differentiator will be how effectively they reduce the operational friction surrounding those systems.
From what I have seen during ERP modernization discussions, organizations rarely struggle because AI lacks capability. The larger challenge usually comes from fragmented processes, workflow dependency and approval-heavy operational cultures. Conversations around AI in ERP frequently focus on new features and automation capabilities. But the organizations creating measurable operational value are usually the ones improving process discipline and operational consistency before scaling AI initiatives.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。