Walking through the halls of Oracle AI World in Las Vegas this year, it was easy to get swept up in the usual showcase of flashy demos and broad-strokes keynotes about AI-powered reinvention. Cutting through the noise, IBM emerged as a counterweight to the hype, reminding enterprises that meaningful AI impact does not begin with models or agents, but with the discipline to build the right foundations.
At several sessions led by IBM, the message was clear: AI innovation may be urgent, but it will not succeed without baked in preparation, planning, processes, and the operational maturity to support AI at scale. Rather than promising overnight breakthroughs, IBM’s perspective reframed the week’s discussions around what must be true inside an organization before any of the promised intelligence can deliver outcomes.
Clean Core, Clear Data, Readiness
As ERP Today reports, only 31% of digital transformations succeed, as initiatives often falter when data, processes, and architecture are not ready. One of the early IBM-led sessions at the conference, “From cloud to competitive edge: Rethinking Oracle for AI success,” illustrated this point. The case of Education First, presented by its CFO and CIO alongside IBM’s global Oracle offering lead, showed how even with significant investment in SaaS and Oracle Cloud, outcomes hinge on how the technology is aligned with the organization’s business architecture, governance, and operating models. IBM’s role ensured those elements worked in concert, turning the existing cloud platform into an engine for transformation rather than assuming the platform alone would deliver value.
Education First’s leaders detailed how they modernized their data fabric, unified financial and operational data, and established a single source of truth using Oracle Fusion Applications with IBM’s help. Such steps were key in reversing the odds of failure and establishing a robust ERP foundation to support analytics as well as emerging AI and automation capabilities.
The takeaway for ERP and enterprise-application teams? Scalable AI adoption depends on strong underlying systems. Clean data, a unified core, and aligned business processes can help ensure that new intelligence enhances well-structured workflows rather than magnify existing complexity. IBM grounded this principle in its decades-long partnership with Oracle, which includes L4 delivery status, joint engineering work across Oracle Cloud Infrastructure (OCI), and the integration of IBM’s Granite models with OCI. Together, these elements form the technical basis that many Oracle customers rely on when moving from experimentation to operational AI.
The Agentic AI Advantage
Later in the week, the session “Agentic AI with real ROI: Fuel performant agents with IBM Granite Models and OCI” shifted the discussion from readiness to what enterprise-scale AI can look like once that foundation is in place. IBM emphasized that deploying one or two generative models is not enough. Enterprises must consider orchestrating a portfolio of agents, each specializing in different functions, and integrate them with business processes and hybrid-cloud infrastructure. This is where IBM’s own models (Granite) and environments (watsonx) are paired with OCI to create the architecture to support this scale.
This combination matters because multi-agent systems need to be coordinated components that interact with finance, HR, procurement, customer service, supply chain, and more. IBM explained how its governance workflows and controls extend across these agents, using watsonx orchestration to monitor decisions, manage handoffs, and apply consistent policy across domains. This approach—tying model governance to the operational reality of ERP systems—ensures that AI agents will behave predictably in regulated or high-impact processes.
To help organizations operationalize this complexity, IBM linked these agentic concepts back to its Consulting Advantage Model. IBM outlined how its architectural blueprints and governance frameworks map directly onto Oracle Fusion Applications, giving enterprises a structured way to integrate agents into cross-functional workflows without disrupting core operating models. Rather than treating agent deployment as an experimental add-on, IBM positioned it as a continuation of foundational design work that ensures agents can scale, reuse each other’s capabilities, and remain auditable over time.
This is a clearer picture of how Oracle Fusion and IBM are approaching multi-agent AI: Oracle Fusion provides the infrastructure and application fabric, while IBM supplies the model layer, orchestration logic, and governance backbone for those agents to operate with consistency. For ERP leaders, this matters because domains cannot be treated as isolated islands. AI agents trained for one workflow must be aligned with others. The preparation work is the difference between experimentation and enterprise adoption.
Realities of AI-Enabled Applications
Two other sessions drove home how preparation manifests in functional domains.
In the session, “A proven framework to build agile supply chains,” IBM and Oracle noted only 6% of organizations claim full end-to-end supply chain visibility, highlighting the wide preparation gap. The session discussed how IBM and Oracle Fusion Applications are helping organizations build integrated planning, execution, and responsiveness capabilities by first remaking the data foundation, business rules, and supply chain architecture before integrating AI-powered tools.
Next, the session “Beyond the buzz: Harnessing the power of Gen AI in HR” tackled payroll, manual processes, and compliance. Legacy systems and disconnected applications force organizations into reactive mode; the path forward requires creating unified HR platforms, defining rules around generative AI use, and ensuring real-time insights rather than batch-reports. Again, IBM emphasized that AI tools are powerful, but only once they sit on structured, well-governed data platforms.
Governance and the Human Element
One of the most consistent themes across the event was governance. In a world eager to talk innovation, the question turns quickly to “Is this safe? Is it accurate?” In IBM-led discussions, governance was not framed abstractly. IBM tied it directly to the assets it brings to Oracle Fusion customers, starting with how the Consulting Advantage Model enables AI integration without compromising compliance or auditability.
Speakers also pointed to the governance features built into the Granite models. Because Granite was developed with transparent data practices, tunable behaviors, and policy-based controls, the models can operate as agents in regulated environments with less risk of drift or unpredictability. Those controls become even more important inside watsonx, where governance workflows track lineage and evaluate outputs before applying enterprise policies consistently across multi-agent systems.
IBM connected these governance capabilities to the work on ContextForge, which enables agents to select the right context, knowledge base, or task handler automatically. For Oracle Fusion customers, this means governance extends across multiple agents as they coordinate decisions across the ERP system. And in supply chain and HR— two domains where IBM already has established frameworks—governance determines how AI recommendations influence planning, compliance, and workforce decisions.
That means ERP leaders should consider treating AI adoption not as a vendor call-out for “we can do generative,” but as a structured program with controls. IBM explained the goal of agentic AI is to elevate human work, not replace it, which in turn makes governance and oversight essential. Without that foundation, the risk is not just failed ROI, but failed trust.
Where Transformation Goes Next
Taken together, the customer-led sessions at Oracle AI World showed a unified progression across industries, with each organization tackling data quality, governance, and operating models before scaling AI. From those discussions, the following threads leading to the next phase of AI-enabled ERP transformation dominated:
- AI readiness continues to outweigh AI experimentation; the case studies presented across Education First, Verisk, DirecTV, and IBM reinforced that strong data foundations and aligned business processes remain essential prerequisites for successful adoption.
- Multi-agent architectures emerging across finance, HR, and supply chain demand enterprise discipline, with cross-domain workflows and controls becoming central to scale.
- The longstanding Oracle-IBM partnership served as a reminder that hybrid cloud alignment and coordinated vendor ecosystems are increasingly critical to delivering practical, end-to-end innovation.
- Governance considerations are becoming a focal point for organizations as they seek to manage new risks around auditability, regulatory compliance, and model oversight.
- While generative AI offers speed and automation, the supply chain and HR discussions made clear that these benefits matter only when accompanied by visibility and control.
IBM’s roadmap adds another dimension to this trajectory. Its forthcoming custom agent, built on ContextForge and the IBM Consulting Advantage Model Context Protocol servers, aims to match tasks automatically with the appropriate assistant, agent, or document collection, supporting scale as organizations move beyond pilots into live operations. This direction aligns with a broader industry shift: ERP environments are evolving from systems of record into systems of intelligence, in which AI agents operate as embedded, components of core processes.
The overarching takeaway is that the era of broad, exploratory AI projects is transitioning into one centered on value, scale, and trust. Innovation now depends less on novel models and more on organizational readiness, architectural coherence, and disciplined execution. For enterprises modernizing ERP systems or integrating AI into operational domains, the foundation laid today will determine the capacity to operationalize AI tomorrow.
What This Means for ERP Insiders
Integration discipline determines whether AI can operate reliably inside ERP systems. The sessions at Oracle AI World demonstrated that enterprises advancing fastest on AI are those stabilizing core data structures and aligning business architecture before deploying agents or generative capabilities. IBM’s work with Oracle Fusion customers shows how unified operating models, governed data fabrics, and consistent process standards form the conditions in which multi-agent systems can function without creating operational risk. For ERP stakeholders, the implication is: AI maturity will increasingly track the quality of integration, not the quantity of added features.
Governance architecture is a strategic differentiator in AI-enabled ERP modernization. IBM’s Consulting Advantage Model, Granite’s embedded control features, watsonx governance workflows, and ContextForge orchestration collectively highlight how governance now determines the ceiling of AI adoption inside enterprise applications. These capabilities reflect a broader trend toward embedding policy, auditability, and safety into the model layer itself rather than treating governance as an add-on. As multi-agent systems begin influencing mission-critical workflows, ERP leaders will increasingly judge providers and partners by the sophistication of their governance stack.
Ecosystems anchored in engineering alignment are setting the pace for AI deployment. The Oracle-IBM partnership illustrates how AI transformation accelerates when platform and services strategies are built in tandem. The joint Granite+OCI architecture shown at Oracle AI World signals a market shift toward vertically integrated AI capabilities that extend from infrastructure to workflow orchestration. For ERP providers and enterprise architects, this reflects how future competitiveness will depend on partnerships that merge application intelligence with cloud performance and synchronize platform innovation with enterprise-grade delivery.
Insights from the Ground
These moments from the session rooms captured what end users are actually doing, struggling with, and learning as they bring AI into ERP-connected environments.
IBM Case Study
- IBM operates a supply chain supported by 20+ interconnected systems, making traditional “lift and shift” automation insufficient.
- Before deploying AI, IBM’s team eliminated unnecessary work and simplified processes to remove manual friction points.
- AI agents were introduced only after this simplification—agents now democratize data access, respond to natural-language questions, and surface insights without dashboards or reports.
- IBM reported $316 million in cost savings over three years, alongside major reductions in inventory losses.
- AI adoption drove cultural change: the company saw a new generation of supply chain talent attracted by AI-enabled roles.
- Decision cycles accelerated as teams learned to sense, decide, and act faster, supported by continuously improving agents.
- IBM frames 2024 as the year AI became embedded in daily operations; the current state is projected as agents taking on more complex tasks autonomously.
DirecTV Case Study
- DirecTV is building an entirely new tech stack and supply chain platform on a deadline that requires go-live by July 2026.
- The transformation is anchored in a definition of supply chain agility: responsiveness, flexibility, visibility, and speed of change.
- DirecTV inherited a tangled ecosystem across CRM, logistics, ERP, and supplier networks, and cited unifying these as a major early challenge.
- The company’s AI strategy begins with a unified data fabric as the enabler to forecast improvements, AI simulations, and real-time supplier collaboration.
- AI is being used for scenario testing before operational decisions are made, supporting a shift from responsive to predictive operations.
- DirecTV is overhauling its success metrics: measuring scenario-building speed, activation performance, demand–supply alignment, and agent-adoption levels.
- A key insight shared: AI agents’ success depends not only on outputs, but on how often humans need to override their decisions.
Verisk Case Study
- Verisk is replacing a 26-year-old on-premises PeopleSoft system with Oracle Fusion Cloud Enterprise Resource Planning (ERP) and Oracle Fusion Cloud Human Capital Management (HCM) across finance and HR.
- The shift is CFO-sponsored, driven by cost-reduction mandates, KPI improvement, and the need for standardized processes in a highly regulated environment.
- With Oracle Fusion Applications live, Verisk now has core HR and finance data in a single, unified ecosystem, enabling new AI-driven value opportunities.
- The organization is building a multi-year roadmap, combining platform capabilities, process efficiencies, and agentic AI; 2026 is the target for broader deployment.
- Verisk is treating AI adoption as a governance-first journey, requiring architecture reviews, business-unit alignment, and C-suite approval.
- Use cases span error correction, predictive analytics, variance analysis, and proactive fraud detection, each handled as a mini-program with strict ROI criteria.
- Verisk is evaluating conversational agents from Oracle, IBM, and internal teams, noting that impact is as much about upgrading workforce capability as cost savings.
