Chief AI Security Officer
SAP Labs India
Chief AI Security Officer Sudhakar Singh discusses autonomous agents, governance frameworks, and the future of intelligent business operations.
As enterprise artificial intelligence evolves beyond basic automation, SAP Labs India is developing agentic AI systems—autonomous agents capable of executing complex, multi-step business processes with minimal human intervention. Sudhakar Singh, Chief AI Security Officer at SAP Labs India, leads efforts to balance AI autonomy with enterprise security and governance requirements.
In this interview, Singh explains how SAP’s agentic framework enables agents to collaborate across business functions, the technical considerations for scaling autonomous systems, and the company’s approach to maintaining human oversight while maximizing operational efficiency.
CIO&Leader: How do you define Agentic AI in the context of enterprise software and workplace operations?
Sudhakar Singh: Agentic AI in an enterprise context refers to autonomous systems that can understand and execute complex, multi-step business objectives. A defining feature is deep business grounding; these agents are tightly integrated with an organization’s data and process landscape, allowing them to operate with contextual awareness rather than simply performing predefined tasks.
For instance, a procurement agent might not only identify the lowest-cost component but also evaluate supplier reliability and production timelines by referencing data within the ERP system.
Another key characteristic is controlled autonomy and customizability. These agents are not black boxes; they are built with frameworks that allow developers to define their scope, constraints, and logic. The most advanced form of agentic AI supports inter-agent collaboration. In this scenario, specialized agents, such as a sales agent and a logistics agent, can coordinate seamlessly to fulfil a complex order, dynamically resolving issues across functional silos without human intervention.
Joule agents exemplify this evolution, moving beyond automation to enable outcome-driven operations while maintaining transparency, accountability, and trust at every step.
CIO&Leader: What kinds of tasks or workflows within SAP systems are ideal for agent-based automation?
Sudhakar Singh: We are currently productizing pre-built reasoning agents suited for high-volume, transactional tasks. These include generating standard purchase requisitions in S/4HANA from material requirements planning runs or creating service tickets in SAP Service Cloud from incoming emails.
More advanced agents are being developed to handle complex, multi-step processes that require customization and control. A good example is automating significant portions of the procure-to-pay cycle. An agent can validate an incoming vendor invoice against a purchase order and goods receipt, identify discrepancies, and post the invoice for payment, flagging only exceptions for human review.
As capabilities evolve, agent collaboration will support cross-functional processes. Consider a lead-to-cash scenario: a sales agent in SAP Sales Cloud hands off a qualified order to a logistics agent in S/4HANA, which then coordinates with a finance agent for billing. The effectiveness of these agents is amplified by their deep integration with SAP’s data models and process flows, enabled through Business Data Context (BDC) and knowledge graphs.
CIO&Leader: How is Agentic AI enhancing decision-making and user productivity in complex enterprise environments?
Sudhakar Singh: A significant benefit of agentic AI is the shift from reactive task execution to proactive process orchestration. This enables better decision-making and significantly boosts user productivity.
SAP agentic systems are leveraging deep integration with SAP’s data and process models. This is achieved through tenant-specific data grounding, which allows agents to synthesize insights across modules. For example, suppose a user queries the status of a sales order in the Sales and Distribution module. In that case, the agent can proactively flag a related credit block from Finance, providing a complete, contextual view that would otherwise require manual effort to piece together.
Productivity gains also come from the autonomous execution of multi-step workflows. An agent managing the procure-to-pay process, for instance, can validate an invoice against Goods Receipt /Invoice Receipt records in Materials Management and post it, requiring human input only for approvals or exceptions.
In more complex environments, inter-agent collaboration automates cross-functional workflows, streamlining processes and enhancing efficiency. A supply chain agent might detect a potential stockout, trigger a purchasing agent to raise a purchase order, and notify a sales agent about potential delivery delays, automating both tasks and the associated decision logic.
CIO&Leader: What are the key technical and governance considerations when deploying autonomous agents at scale?
Sudhakar Singh: Scaling autonomous agents requires a robust framework that strikes a balance between autonomy and oversight. In general, the degree of agent autonomy must be inversely proportional to the risk associated with the task, making human oversight essential in enterprise AI deployments.
A critical implementation is the Human-in-the-Loop (HITL) mechanism, which SAP-developed agents use extensively. For high-risk activities, HITL acts as a checkpoint, pausing execution until an authorized user provides approval. This ensures segregation of duties and transparent accountability, which is especially important in regulated industries or for business-critical functions. For example, an agent might detect a supply chain disruption and analyse vendor data, but a human would still make the final procurement decision.
SAP’s agentic framework incorporates stringent security guardrails, including content moderation, injection prevention, prompt hardening, and other cloud-native security protocols. Developer enablement is also key: rather than deploying all-powerful “superadmin” agents, we encourage task-specific agents with clear privilege boundaries and authorization checks. Each agent’s operational scope and capabilities (the allowlist) are explicitly defined.
All agent interactions, decisions, and actions are immutably logged. This traceability is essential for debugging, explaining outcomes, and complying with regulations such as the GDPR, the EU AI Act, and other country-specific frameworks. Agent autonomy should always be matched to the risk profile of the use case via threat modeling, ensuring that equally strong technical and ethical safeguards accompany higher levels of automation.
CIO&Leader: How is SAP Labs India approaching R&D in Agentic AI to shape the future of intelligent enterprise systems?
Sudhakar Singh: At SAP Labs India, we are leading the charge in Agentic AI to shape the next era of intelligent enterprise systems. Our R&D approach is grounded in the belief that AI should not just augment decision-making but evolve into collaborative, autonomous agents that deeply understand business context and can act on it with purpose and precision. We are investing in next-generation AI capabilities, ranging from machine learning and natural language processing to deep learning and reasoning engines, to build flexible agentic frameworks. These frameworks enable the deployment of intelligent agents that are context-aware, ethically grounded, and capable of orchestrating complex business processes across industries.
A cornerstone of this vision is SAP Business Data Cloud (BDC), a unified data foundation launched in 2025 that breaks down data silos across SAP and non-SAP systems. When integrated with platforms like SAP Datasphere, SAP Analytics Cloud, and Databricks, BDC allows our Joule AI agents to operate on real-time, semantically enriched business data. These agents can now detect patterns, understand relationships—such as the link between supplier delays and inventory gaps—and take proactive action. This moves AI from probabilistic to more deterministic behavior, delivering higher accuracy and business value. To foster innovation at scale, SAP Labs India offers a robust ecosystem of learning and enablement, including initiatives like the AI Learning Summit and Learning Fest. These programs empower our talent across various roles, including developers, consultants, and business users, to upskill in emerging technologies and build AI-powered solutions with confidence and creativity.
We are also working closely with partners and customers through design-led innovation and strategic pilots to co-create AI agents that address real-world challenges while upholding high standards of responsibility, transparency, and fairness. In essence, our holistic approach to Agentic AI encompasses not only technology but also talent, trust, and transformation. We envision a future where intelligent agents are embedded into the DNA of enterprises, acting as trusted digital teammates to drive productivity, compliance, and innovation across the board.