SAP’s AI agents redefine ERP—but can governance keep up?

Shift*Academy ↗

The gist

SAP is unleashing a new era of autonomous AI agents inside ERP systems, automating business-critical workflows faster than most companies can modernize—or govern—them.

What to know

  • SAP has embedded over 50 Joule AI Assistants and 200+ specialized agents into its ERP suite, automating complex tasks across finance, supply chain, HR, and customer experience.
  • Enterprises like H&M, Bayer, and ExxonMobil are already reporting up to 80% faster vendor onboarding and major efficiency gains, but legacy infrastructure and data silos remain major hurdles.
  • SAP is doubling down on governance, agent identity, and open standards through €100 million in partnerships with Anthropic, Palantir, and Google Cloud to ensure AI is both scalable and secure.

AI-Native ERP Takes Shape

SAP’s AI agents are moving beyond chatbots to orchestrate autonomous, industry-specific workflows tightly governed for trust and compliance.

SAP’s Autonomous Enterprise vision represents a transformative integration of agentic AI deeply embedded within its ERP systems through the unified Business AI Platform and Autonomous Suite. This architecture empowers over 50 domain-specific Joule AI Assistants and more than 200 specialized agents to automate and orchestrate complex workflows across finance, supply chain, procurement, HR, and customer experience, moving beyond standalone chatbots to AI-native processes that collaborate with humans and operate within strict governance frameworks. By leveraging industry-specific AI solutions, SAP tailors autonomous workflows to sector-specific challenges, underscoring its commitment to embedding AI that acts on insights and context rather than merely generating suggestions.

Central to SAP’s AI architecture is the Joule interface, a reimagined user experience and orchestration layer that functions as a super agent, enabling outcome-driven interactions across SAP and non-SAP systems. Joule facilitates seamless collaboration among AI assistants and agents by generating dynamic visualizations and managing complex workflows, thereby enhancing user engagement and operational transparency. This approach is reinforced by SAP’s forward-deployed engineering teams who work closely with customers to co-develop explainable, reliable AI solutions deeply integrated into existing ERP environments, ensuring practical value and trustworthiness in autonomous operations.

SAP’s CEO Christian Klein and COO Sebastian Steinhauser emphasize that accuracy, compliance, and security are non-negotiable pillars for AI-driven automation in mission-critical enterprise processes, reflecting SAP’s focus on trustworthy AI embedded within governance and data quality frameworks. Steinhauser highlights that the AI layer is built atop a robust application and platform foundation, cautioning against superficial AI overlays that risk replicating past automation failures. This foundational rigor, combined with strategic partnerships with Anthropic and Palantir and a €100 million fund to accelerate AI adoption, positions SAP to deliver scalable, secure, and compliant autonomous workflows that reduce manual intervention and ERP migration complexity by up to 35%.

SAP’s autonomous enterprise strategy integrates AI assistants, trusted business data, and cloud ERP migration tools into a unified operating model that supports autonomous workflows and collaboration within industry-specific contexts. Recent data acquisitions like Dremio and Reltio enhance real-time analytics and master data governance, grounding AI agents in cleaner, connected enterprise data to improve decision-making quality. This integration facilitates a shift in SAP’s cloud revenue model toward consumption and value-based pricing, reflecting broader trends in enterprise AI adoption and underscoring SAP’s commitment to measurable business outcomes through aligned AI initiatives and comprehensive change management.

Sources

Data Hurdles and Agent Governance

Legacy systems and fragmented data threaten AI agent effectiveness, making robust agent identity and audit controls essential for secure enterprise automation.

Embedding autonomous AI agents within enterprise ERP systems demands overcoming immense data complexity and legacy infrastructure challenges, as highlighted by SAP CEO Christian Klein who pointed to harmonizing millions of data fields through knowledge graphs to capture unique process logic. However, legacy modernization remains a bottleneck, with enterprises facing the 'enterprise paradox' of urgent AI expectations clashing with slow infrastructure upgrades and manual processes. This tension is underscored by SAP and Google Cloud’s emphasis on cloud migration and modern platforms to support production-scale multi-agent orchestration, while SAP research reveals that fewer than 40% of companies effectively share customer data across CRM and CX platforms, exposing significant governance and integration gaps.

Governance emerges as a cornerstone in scaling AI agents, requiring robust frameworks that encompass agent identity, audit trails, and strict access controls to ensure compliance and security in high-stakes environments like finance. PwC and SAP stress that every agent must have a unique identity and operate within governed data contexts, while Microsoft’s Agent 365 and SAP’s AI Agent Hub provide centralized governance layers to monitor and manage multi-agent ecosystems. This governance complexity extends to balancing autonomous agent operation with human oversight, as organizations grapple with defining decision rights, segregation of duties, and exception handling to prevent compounding operational risks from AI-driven transactions.

Multi-agent orchestration is critical for managing complex workflows that exceed the capabilities of single AI systems, necessitating sophisticated coordination models and integration across diverse enterprise systems. Experts like Martin Bufi emphasize that successful deployments rely on narrowly scoped agents orchestrated with patterns tailored to workflow dynamics, supported by emerging standards such as the Model Context Protocol (MCP) and AGENTS.md from the Agentic AI Foundation to reduce integration friction. Case studies from Southeast Asia and companies like General Assembly demonstrate how chains of specialized agents improve accuracy and efficiency, while platforms like Harnyss and Microsoft Dynamics 365 embed agents directly into ERP workflows to maintain operational control and enable seamless handoffs within governed environments.

Successful scaling of AI agents hinges on foundational data governance, process standardization, and cross-functional collaboration to create a dependable contextual environment that feeds AI with deep, current data and aligned business rules. Vanessa Padiachee of Standard Bank highlights the necessity of high-quality, trusted, and accessible enterprise data alongside strong partnerships between business and technology teams to reduce project risk. Moreover, phased deployment strategies that start with data readiness and targeted low-risk use cases, as practiced by SotaTek and PromptCare, help enterprises iteratively refine AI integration while maintaining human-in-the-loop oversight to ensure compliance, trust, and exception management.

Sources

Enterprise Wins and Lessons Learned

H&M, Bayer, and ExxonMobil show that rapid AI-driven gains depend on foundational data cleanup and phased ERP modernization, not just technology upgrades.

Leading enterprises such as H&M, Bayer, and ExxonMobil exemplify how autonomous AI agents are delivering measurable business value by optimizing operational efficiency and accelerating ERP modernization. H&M enhances its omnichannel retail experience through personalized AI-driven upselling during returns, while Bayer automates repetitive tasks in shared services centers. ExxonMobil’s greenfield approach, focusing on master data cleanup and minimal customization, lays a robust foundation for powerful AI layers, underscoring that successful AI adoption hinges on foundational modernization and data harmonization.

The transition from AI pilots to production-scale autonomous workflows is increasingly driven by executive demands for clear business outcomes and measurable ROI, moving conversations beyond technology capabilities to tangible impacts like revenue growth, cost reduction, and customer satisfaction. Southeast Asian logistics firms slashed vendor onboarding times from five days to under four hours using multi-agent AI workflows, while SAP and Microsoft emphasize governance frameworks and human-in-the-loop models to ensure AI agents operate securely and transparently within core business processes.

Large-scale ERP migrations and AI integrations, such as those at Standard Bank and TCL SunPower, highlight the critical role of strong governance, detailed process mapping, and foundational technical readiness—including Unicode conversion and data quality—in realizing operational resilience and accelerating transformation timelines. TCL SunPower’s deployment of SAP GROW and the Joule AI co-pilot accelerated financial close by three days, improved inventory accuracy by up to 15%, and consolidated multiple finance systems, demonstrating how phased rollouts and AI-driven automation can scale from pilots to enterprise-wide impact.

Agentic AI is reshaping the economics and execution of SAP transformations by automating complex, multi-step workflows and reducing reliance on highly experienced labor through AI-augmented task execution. Stuart Browne’s research suggests up to a 60% cost reduction by compressing effort and enabling less experienced workers to handle tasks previously reserved for experts, while experienced professionals focus on high-value judgment and validation. This shift is evident in accelerated delivery of functional specifications and fit-gap analyses, where AI produces 80% solutions within hours, dramatically shortening project timelines and enabling enterprises to scale autonomous AI agents beyond experimentation into sustained production use.

Sources

Ecosystem Partnerships Drive Scale

SAP’s €100 million fund and alliances with Anthropic, Palantir, and Google Cloud are turning open standards and deep integrations into a competitive edge for AI-powered business automation.

By mid-2026, SAP had strategically launched a €100 million partner fund alongside open standards initiatives to accelerate AI adoption across enterprise workflows, underscoring a collaborative approach to innovation. Partnerships with leading technology firms such as Anthropic, Palantir, Google Cloud, and SK AX form the backbone of SAP’s multi-agent commerce architectures, enabling scalable autonomous AI agents that automate finance, supply chain, and ERP processes. This ecosystem-driven strategy not only fosters joint growth but also integrates AI deeply into core business functions rather than treating it as an add-on productivity tool.

SAP’s collaboration with Google Cloud exemplifies the critical infrastructure and governance challenges inherent in scaling agentic AI for enterprise use. PwC highlights that every AI agent must have a distinct identity, traceable actions, and governed, contextualized data access to support production-scale marketing and retail operations. Despite 78% of businesses recognizing AI’s importance for customer retention by 2026, data-sharing remains a bottleneck—with only 37% and 39% of companies sharing customer data across CX and CRM platforms respectively—an issue SAP aims to address through its ecosystem standards and partnerships.

SAP’s commitment to open standards and interoperability is evident through its Gold Membership in the Agentic AI Foundation and active participation in projects like the Model Context Protocol and AGENTS.md. These efforts are crucial for integrating large language model-driven agents with external data sources and tools, ensuring seamless multi-agent workflows across diverse enterprise environments. This standardization work complements SAP’s cloud transition strategy, which emphasizes trusted operational data and governance as key competitive differentiators for both cloud and select on-premises customers.

The partnership with SK AX illustrates SAP’s ambition to build an AI-native enterprise by embedding agentic AI directly into ERP business processes. SK AX’s AXgenticWire ERP Suite, featuring specialized capabilities like 'my Finance' for automating financial tasks and anomaly detection, demonstrates how AI agents can be tailored to legacy systems and security policies. Leveraging SAP’s AI Agent Hub and global ecosystem, SK AX plans to scale these AI-driven ERP solutions internationally, using SK Group as a real-world test bed to validate early use cases before broader market expansion.

Sources

Modernization or Missed Opportunity

Without accelerated ERP upgrades and harmonized data, even the best AI agents risk falling short—prompting SAP and Palantir to bet on usage-based models and open platforms to future-proof enterprise transformation.

SAP CEO Christian Klein underscores the critical urgency for enterprises to accelerate AI value realization, warning that prolonged modernization timelines—such as multi-year ERP upgrades—delay the deployment of transformative agentic AI capabilities. He emphasizes that the ERP system, as the 'brain of the enterprise,' must be preserved and leveraged to provide the deep process logic and semantically rich data context essential for effective AI integration. This foundational complexity, involving over 7.5 million data fields harmonized over five years, necessitates a nuanced approach that balances innovation speed with the intricate realities of enterprise data and process governance.

SAP and Palantir are strategically repositioning themselves from traditional software vendors to business AI companies and AI infrastructure providers, reflecting a broader market shift toward agentic AI that transcends conventional applications. SAP’s approach integrates technical and domain expertise to manage the profound impact of AI on enterprise workflows and culture, embedding nearly 50 AI assistants and over 400 Autonomous Suite agents across core functions like finance and supply chain. This shift is accompanied by a move away from seat-based licensing toward consumption- and value-based pricing models, anticipating that usage-driven revenue will dominate cloud growth by 2030 as AI adoption scales.

Despite the promise of agentic AI, SAP leadership acknowledges significant challenges including AI accuracy limitations, governance complexities, and the necessity of strong foundational systems to avoid failures akin to 'RPA 2.0 disasters.' CFO Dominik Asam stresses that AI must be rigorously governed in high-stakes processes like finance to mitigate risks from errors or hallucinations, highlighting that reliable returns stem from process-specific AI rather than generic large language models. SAP’s open platform strategy allows customers to select any LLM, including open source, ensuring control over data and business context while reinforcing SAP’s unique value proposition rooted in deep domain knowledge and embedded governance.

The economic and strategic implications of agentic AI adoption are tightly intertwined with ERP modernization, as enterprises face significant financial and operational hurdles migrating to cloud platforms to unlock full AI capabilities. SAP’s AI-powered ERP migration tools promise to reduce deployment time and costs by up to 30%, yet CIOs grapple with budget constraints where migration and AI investments compete for the same funds. Furthermore, SAP and Oracle’s pricing models, which combine base subscriptions with metered AI consumption fees, complicate ROI calculations and risk management. While SAP’s entrenched position and extensive installed base create market inertia against rapid disruption, competitive pressures from Oracle and Salesforce, alongside geopolitical and cybersecurity risks, temper near-term earnings momentum and adoption timing.

Agentic AI is poised to rewrite the economics of SAP transformation by enabling a 'deshoring' model that automates complex migration tasks, potentially cutting costs by approximately 60% and compressing effort timelines. Research by Stuart Browne reveals that AI can produce up to 80% of functional and technical specifications within hours, allowing less experienced workers augmented by AI to handle tasks traditionally reserved for highly skilled professionals. This redistribution of human expertise focuses seasoned SAP consultants on high-value judgment and validation activities, accelerating delivery while maintaining quality. Given the scarcity of SAP skills and the complexity of global programs, agentic AI adoption emerges as a critical lever for enterprises seeking faster, cheaper, and higher-value SAP S/4HANA migrations.

Sources
DiginomicaCloud Wars Live with Bob EvansTRYahoo FinanceYahoo FinanceBloomberg Talks

Agents Become Business Backbone

Multi-agent AI is set to quietly underpin enterprise operations, shifting control from humans to autonomous actors while demanding new levels of governance and interoperability.

By mid-2026, multi-agent architectures have transitioned from a stopgap solution to a foundational element of AI systems, becoming so embedded that they are expected to fade into the background of enterprise technology. This evolution reflects a dual landscape where open-source frameworks like Langgraph and Pydantic agents prioritize model interchangeability, while hyperscalers and major AI providers push vertically integrated stacks combining compute, models, agent harnesses, and interfaces into cohesive platforms. This coexistence caters to diverse enterprise needs, balancing flexibility with opinionated, end-to-end solutions.

The paradigm shift from AI as static applications to autonomous actors is reshaping enterprise workflows, with AI agents increasingly taking center stage as independent decision-makers. Predictions indicate that within three years, 90% of AI control tower users will be agents rather than humans, a milestone underscored by recent data showing agent-generated internet traffic surpassing human traffic. This transformation extends to multi-agent, multi-enterprise workflows where AI agents autonomously interact across organizational boundaries, potentially bypassing traditional human-mediated processes such as regulatory submissions, heralding a new era of direct, data-driven enterprise operations.

While the rise of agentic AI platforms like Harnyss promises autonomous business operations that coordinate agents across marketing, sales, finance, legal, and engineering, the critical challenge remains governance and control. Harnyss’s flexible control levels—from human approval to full autopilot—allow enterprises to calibrate autonomy against risk, ensuring trusted workflows run independently without sacrificing oversight. Moreover, their bring-your-own-model-key approach empowers companies to maintain control over underlying AI usage, while extensive integrations with existing enterprise systems like Salesforce, Slack, and Google Analytics emphasize interoperability over replacement, preserving the integrity of current workflows.

The emergence of autonomous AI agents introduces what some experts describe as 'alien intelligence,' necessitating robust trust frameworks and governance mechanisms to align these agents with human intent and societal values. Innovations like OpenClaw represent leaps far beyond earlier models such as ChatGPT but also highlight the pressing need for policy alignment and control to steer these powerful agents toward beneficial outcomes. Companies like Trustwise are pioneering solutions to this alignment challenge, recognizing that without effective governance, the transformative potential of agentic AI could be undermined by misaligned or uncontrolled autonomous actions.

Sources

Get the stories behind the trends

Deep-dive reporting and the weekly brief, in your inbox.