As businesses integrate artificial intelligence into core operations, the difference between 90 percent and 100 percent accuracy is not a minor gap. According to SAP global president of customer success Manos Raptopoulos, it is existential. Enterprise AI governance, he argues, secures profit margins by replacing statistical guesses with deterministic control.
Consumer-grade language models often miscount words in a document by roughly ten percent. In production environments, that margin of error is unacceptable. Raptopoulos states that the evaluation criteria for large language models in business have formally shifted toward precision, governance, scalability, and tangible impact on revenue.
The Core Governance Challenge
Corporate boards now face a pressing transition: AI systems are evolving from passive tools into active digital actors. These agentic systems can plan, reason, coordinate with other agents, and execute workflows autonomously. Because they interact directly with sensitive data and influence decisions at scale, Raptopoulos warns that failing to govern them exactly like a human workforce exposes an organization to severe operational risk.
He compares the uncontrolled proliferation of AI agents to the shadow IT crises that plagued businesses a decade ago, but notes the stakes are higher. Establishing agent lifecycle management, defining autonomy boundaries, enforcing policy, and instituting continuous performance monitoring are now mandatory requirements.
Engineering Constraints for Agentic AI
Integrating modern vector databases with legacy relational architectures demands significant engineering capital. Teams must restrict the agent’s inference loop to prevent hallucinations from corrupting financial or supply chain execution paths. Setting strict parameters increases computational latency and hyperscaler compute costs, which alters initial profit and loss projections.
When an autonomous model requires constant, high-frequency database queries to maintain deterministic outputs, token costs multiply quickly. Governance therefore becomes a hard engineering constraint, not merely a compliance checklist.
Three Baseline Issues for Boards
Raptopoulos argues that corporate boards must resolve three baseline issues before deploying agentic models: identifying who holds accountability for an agent’s error, establishing audit trails for machine decisions, and defining the exact thresholds for human escalation. Geopolitical fragmentation complicates these questions.
Sovereign cloud infrastructures, AI models, and data localization mandates are regulatory realities in major markets including New York, Frankfurt, Riyadh, and Singapore. Enterprises must embed deterministic control directly into probabilistic intelligence. Raptopoulos views this requirement as a C-suite mandate, not an IT project.
The Data Foundation Moment
Commercial AI systems remain entirely dependent on the quality of the data and processes they operate upon. Fragmented master data, siloed business systems, and over-customized ERP environments introduce dangerous unpredictability at critical moments. If an autonomous agent relies on fragmented foundations to generate a recommendation affecting cash flow, customer relations, or compliance, the resulting operational damage scales instantly.
Extracting tangible enterprise value requires advancing beyond generic large language models trained on internet-scale text. Raptopoulos states that true enterprise intelligence must be grounded in proprietary corporate data: orders, invoices, supply chain records, and financial postings embedded directly into business processes. Relational foundation models optimized for structured business data will continually outperform generic models in forecasting, anomaly detection, and operational optimization.
The operational friction of making an over-customized ERP environment intelligible to a foundation model halts many deployments. Data engineering teams spend excessive cycles cleaning fragmented master data simply to create a baseline for AI ingestion. When a relational model needs to interpret complex, proprietary supply chain records alongside raw invoice data, the underlying data pipelines must operate with zero latency. If data ingest fails, the model’s predictive capabilities degrade instantly, rendering the agent functionally dangerous.
Integrating legacy architecture with modern relational AI requires overhauling deeply entrenched data pipelines. Engineering teams face indexing decades of poorly classified planning data so that embedding models can generate accurate vector representations. Boards must therefore evaluate whether their current data estate is genuinely prepared, rather than layering probabilistic intelligence over disjointed foundations.
Designing Intent-Based Interfaces
Enterprise application interaction is transitioning from static interfaces to generative user experiences. Instead of manually navigating complex software ecosystems, employees will express their intent to the system. Raptopoulos offers an example: a user instructs the software to prepare a briefing for their highest-revenue customer visit that week. The AI agents then orchestrate the necessary workflows, assemble surrounding context, and surface recommended actions.
However, adoption among the workforce remains conditional upon trust. Employees will embrace these digital teammates only if they feel confident in the system’s accuracy and governance.
Looking ahead, Raptopoulos plans to present these governance frameworks at the AI & Big Data Expo North America. The expected next steps for enterprise AI involve tighter integration of deterministic controls, ongoing audits of agent behavior, and development of standardized accountability models across global markets.
