The promise of self driving enterprise systems is immense, but the challenge of scaling this technology is proving to be a major roadblock for CIOs.
The Unfulfilled Promise of Enterprise Resource Planning
Despite colossal investments spanning decades, enterprise resource planning (ERP) platforms have a notorious track record of underperformance. Data from a recent Bain & Company benchmarking survey indicates that over 80 percent of ERP transformations fail to meet their projected budget, timeline, and value targets. The root of this widespread shortfall often lies in implementations that result in overly customized, difficult to maintain, and intelligence deficient systems. The original vision of an agile, standardized, and harmonized backbone for the enterprise frequently devolves into a complex, sprawling, and bespoke technological landscape that is resistant to necessary upgrades.
Generative AI’s Potential to Deliver Touchless Systems
The emergence of generative AI provides a compelling potential solution to this persistent inefficiency. While initial applications have focused on streamlining the migration process itself, through automation of testing, code remediation, and documentation, the truly disruptive potential resides in creating “touchless platforms” powered by a more advanced paradigm: agentic AI.
Defining the Next Evolution of Enterprise Automation
Agentic AI systems are fundamentally intelligent, event driven digital components capable of executing actions and initiating decisions autonomously, without human intervention. This could include automatically rerouting complex workflows or making real time process adjustments. This capability shifts the ERP from a passive repository of data to a dynamic engine for both decision making and execution, promising unprecedented efficiency and agility. The ultimate goal is an environment where human employees interact directly with AI agents, bypassing traditional software user interfaces. Platform vendors are now aggressively developing off the shelf and semi customizable agentic tools, signaling a clear market shift toward more sophisticated, automated enterprise processes. However, despite this vendor activity, most current products remain in their early stages, often limited to narrow or highly specific use cases.
The Five Obstacles Hindering Scalability
While the direction of travel is certain, many corporations find themselves stalled in the pilot phase of agentic AI adoption. This stagnation is largely attributable to five critical roadblocks. Organizationally, there is a lack of clear operating models for human agent collaboration and a deficit in the necessary internal technical skills. Technically, the immaturity of current agentic tooling is an issue, with essential orchestration frameworks and robust customization options only just beginning to emerge. Data quality presents a significant barrier, particularly where information is siloed or governance is weak. Strategically, senior executive sponsorship can be elusive, and concerns over future vendor lock in are common. Finally, quantifying the return on investment (ROI) remains difficult, given the lack of clear pricing models, and the challenging nature of predicting productivity gains and ultimate business outcomes.
The Areas Poised for the Greatest Impact
Despite these adoption challenges, the market consensus is clear. According to the Bain study, a substantial 78 percent of IT leaders anticipate that some degree of ERP functionality will be augmented or replaced by agentic AI within the next three years. The most profound and immediate impacts are expected to materialize in core financial and planning processes. Surveyed executives specifically identified procure to pay, record to report, and forecast to plan as the ERP functions most likely to see early and substantial gains from agentic automation.
Laying the Foundational Structure for Agentic AI
To move beyond isolated pilot projects and achieve enterprise wide scale, CIOs must systematically address four interdependent strategic questions. The first question centers on defining the necessary foundations. This begins with a disciplined prioritization of use cases, focusing on the three to five areas that offer the highest measurable business value for rapid returns. Scaling also requires the development of new operating models to explicitly define the framework for human agent interaction, adjust workforce roles, and establish transparency in agent performance metrics. Robust governance is essential to manage compliance and data risks and is a prerequisite for any widespread deployment. Crucially, organizations must move beyond merely bolting agents onto existing structures and commit to fundamentally redesigning core processes to be agent first. This necessitates clear data triggers, defined data structures, guardrails, and auditability, rethinking the entire decision making and execution flow. An agentic transformation must be conceived as a system capable of supporting dozens or hundreds of agents, not just a handful of prototypes. Without this comprehensive groundwork, successful scaling is highly unlikely.
The Critical Decision: Build, Buy, or Partner
Once the foundational readiness is established, the second question addresses the delivery model: should the organization build custom agents in house, acquire pre existing solutions from vendors, or enter into co development partnerships? The optimal strategy is highly dependent on the specific use case and must balance seven key factors. These include the strategic relevance of the capability, the functional fit of vendor applications, ease of integration, total cost of ownership, internal talent capacity, regulatory risks, and the potential for future vendor lock in.
Standard, noncritical workflows that remain confined within a single system might be best suited for partnering with the platform publisher for rapid, cost effective implementation. Conversely, acquiring solutions from third party specialists is sensible for commodity, non complex capabilities that allow for quick traction without major internal investment. Building in house is most appropriate for high impact, cross system processes that provide a genuine competitive differentiation and where maintaining full intellectual control justifies the extended timelines and higher costs. The right approach will rarely be singular, demanding a dynamic and flexible strategy that acknowledges the rapid pace of innovation in this sector.
Leveraging Strategic Alliances for Accelerated Maturity
The third critical question concerns the role of external partnerships in transitioning from pilot to scale. At the exploration stage, platform providers are essential, offering the tools, templates, and support needed for safe experimentation, often via low code studios and prepackaged agents. As organizations mature, business integrators become vital strategic partners, helping to prioritize use cases, define the new operating model, redesign core processes, and establish the frameworks for measuring ROI. System integrators are increasingly developing deep agentic expertise, primarily assisting clients with the customization and deployment of agents across complex system landscapes. These partnerships are dynamic, and as an organization gains maturity, it will absorb partner experience and be able to internalize more of these functions over time.
The Consequences of Waiting for the Ideal Moment
The final question addresses the strategic risk of inaction. While many enterprises continue to observe from the sidelines, a growing number of industry leaders are successfully scaling agentic AI and realizing significant benefits. Bain research indicates that leaders who have scaled AI across workflows are already achieving 10 to 25 percent EBITDA gains. Those who delay risk not only being outpaced by their competitors but also by the aggressive roadmaps of their own technology vendors, limiting their influence over feature development.
Further, prolonged delay increases the risk of vendor lock in. As enterprise platforms evolve into comprehensive orchestration hubs that govern both the agents and their interconnections, switching costs will escalate dramatically. Companies that fail to develop their own internal “AI muscle” risk becoming passive consumers of proprietary automation ecosystems. Delaying also raises the ultimate cost of transformation. Organizations that separate their agentic journey from existing IT modernization efforts may find themselves re engineering processes twice: once for the ERP migration and again for AI embedding. To maintain cost efficiency, the agentic transformation must be an integral component of the ongoing enterprise technology overhaul. Finally, there is a critical people implication. Agentic AI is becoming a factor in talent recruitment and retention, as high potential professionals are drawn to organizations working on future facing technologies, putting those who delay at risk of losing their competitive edge in talent.
The Emerging Battle for the Orchestration Layer
Agentic AI is no longer a theoretical concept; it is now an embedded reality within leading enterprise platforms. The technology is proven, and the tools are available, yet many CIOs remain in a state of stasis. Breaking this deadlock hinges on a deliberate focus on the four strategic questions outlined above. A pivotal decision yet to be fully determined across the industry is who will ultimately control the orchestration layer. Platform providers are positioning themselves as the central hub, often promoting proprietary interoperability that implicitly nudges customers toward their ecosystems. CIOs must remain vigilant. Success with agentic AI depends not just on acquiring the right tools and talent, but also on making architectural decisions that secure long term flexibility. As multi agent systems become increasingly complex, the ability to effectively direct and monitor them will become a core source of enterprise advantage. The future leaders will be those who successfully navigate the treacherous transition from localized experimentation to disciplined, enterprise wide execution. The right framework for decision making can be found by consulting the latest research and analysis in this space (https://www.bain.com/insights/is-agentic-ai-the-inflection-point-for-scaling-erp-transformations/).
