# From AI pilots to autonomous AMS: The enterprise readiness test for SAP - IBM

*Источник: IBM*
*Дата: 2026-09-23*
*Язык: en*

**Кратко:** Enterprises are rapidly moving from experimenting with generative AI to deploying copilots, intelligent automation and agentic AI across technology and business functions. SAP application management is an obvious area of opportunity because large SAP estates contain significant volumes of repetitive, data-intensive and rules-driven work.

Enterprises are rapidly moving from experimenting with generative AI to deploying copilots, intelligent automation and agentic AI across technology and business functions. SAP application management is an obvious area of opportunity because large SAP estates contain significant volumes of repetitive, data-intensive and rules-driven work. This work exists across areas such as incident management, application maintenance, testing, enhancements and operations. AI is already assisting many of these activities and delivering meaningful productivity gains.
Yet moving from AI-assisted work to autonomous execution represents a fundamentally different challenge. An AI assistant that recommends a remediation tactic is different from an agent that identifies an incident, determines the appropriate response, executes the remediation and validates the outcome autonomously. This shift creates new questions for chief information officers (CIOs) about whether the enterprise is ready to give AI meaningful operational authority.
The distinction matters because AI adoption and AI readiness are not the same thing. An organization can run dozens of AI pilots and still lack the data, processes, integration, governance and skills required to operate AI at scale. Therefore, organizations should evaluate their journey toward autonomous AMS not by the number of AI initiatives underway, but by their readiness to let AI take on increasingly consequential work.
Traditional automation follows predefined rules and workflows. Generative AI assists people with information, analysis and recommendations. Agentic AI introduces a fundamentally different capability—allowing systems to reason, plan and execute multi-step activities toward defined outcomes.
Consider a routine but business-critical accounts payable (AP) scenario in SAP. An invoice arrives that does not match the corresponding purchase order and goods receipt, creating an exception that needs to be investigated before payment can proceed.
With AI assistance, an AP analyst can use AI to compare the relevant records, identify the likely cause, summarize similar historical cases and recommend a resolution. The AI improves analyst productivity, but the human analyst still owns the investigation, decision and execution.
Now imagine the same process supported by an agentic AI system. The agent detects the exception, gathers and correlates the relevant SAP data, investigates the discrepancy and checks business rules and historical patterns. It then determines whether the case falls within an approved resolution path, initiates the appropriate workflow and validates the outcome.
Where the case falls outside its authority, such as a policy violation, unusual transaction or potentially fraudulent activity, it escalates to a human with the relevant evidence and recommendation.
The difference is subtle but profound. AI assistance helps people perform a process. Agentic AI can perform a process within defined boundaries. The first improves individual productivity. The second begins to change the operating model itself. And this distinction becomes important as enterprises consider autonomous AMS.
The objective is not simply to apply AI to more SAP tasks. It is to progressively determine which operational activities AI can own safely, which ones require human judgment and what capabilities the enterprise needs before making that transition.
If autonomous AMS represents a fundamentally different operating model, organizations need a way to determine whether they are ready for it. The answer cannot be reduced to a single AI maturity score. Readiness depends on several interconnected capabilities across technology, processes, data, governance and workforce. Here are six questions that can provide a practical starting point.
Autonomous AI cannot operate effectively on an application landscape that is overly complex, poorly integrated or difficult to change. CIOs need to assess custom code, technical debt, architecture, Clean Core, API exposure and the ability of agents to interact safely with applications. Clean Core and API readiness are no longer just modernization priorities. They are foundational to AI-driven operations. If agents cannot reliably understand and interact with the application environment, autonomy will remain constrained, regardless of how capable the underlying AI becomes.
Autonomous agents need more than access to a large language model (LLM). They need the right information to understand what is happening and determine what to do next. SAP operations generate vast amounts of data across transactions, logs, monitoring, ITSM, configuration, knowledge and business processes. If these sources are disconnected, stale or inconsistent, agents cannot build an accurate operational picture. As a result, data readiness is a prerequisite for autonomous AMS. The goal is not more data, but the right information, available at the right time and in a form that agents can act on.
This operational reality is one of the most overlooked readiness questions. Years of operating complex SAP environments often leave behind manual handoffs, approval steps, exception paths and workarounds that work for people but are poorly suited to autonomous execution. Simply adding an AI agent to an inefficient process can automate the inefficiency rather than remove it. As a result, organizations need to identify which processes should be redesigned for AI. The question is no longer how to automate an existing workflow, but how to redesign it for autonomy.
The greater the authority granted to AI, the more important clear decision boundaries become. Leaders need to define which actions agents can perform independently, which ones require human approval and which actions organizations should not delegate. Most importantly, they need clear accountability and mechanisms to reverse autonomous actions. Organizations can progressively move from human-in-the-loop to human-on-the-loop and, where appropriate, manage-by-exception. This move toward autonomy should happen only when processes are predictable, risks are understood and controls are proven effective. The goal is not maximum autonomy, but the right level of autonomy for each decision.
As AI takes on more routine execution, human expertise can shift to where it creates greater value. This transition enables a shift-left model, with experienced architects and engineers involved earlier in requirements, design and architecture to prevent issues and focus on defining AI decision boundaries and improving downstream execution.
In this forward-deployed model, human specialists work alongside digital workers across design, build and manage. Forward-deployed architects (FDAs) provide architectural judgment and governance, while forward-deployed engineers (FDEs) orchestrate agents and handle judgment-intensive work. Together, they can form scalable forward-deployed units (FDUs) that combine human expertise with AI-driven execution.
Organizations need to distinguish between AI adoption and AI value. For SAP AMS, CIOs should connect AI activity to outcomes such as faster incident resolution, fewer recurring incidents, safer releases and greater automation. They should also measure reduced technical debt, stronger application resilience and more capacity for higher-value work. This shift is critical. Autonomous AMS should not become an AI program searching for a business case. It should be an operating model designed around measurable business outcomes. The model should move beyond task-level productivity. The goal is to capture the broader value AI enables across the enterprise.
AI autonomy cannot be treated as a switch that allows an organization to simply move from a collection of gen AI pilots to fully autonomous SAP operations. Effective autonomy is best described as a progression in which each stage introduces new capabilities and new responsibilities:
1.    At the first level, AI assists people by providing information, recommendations and content. Humans remain responsible for execution.
2.    At the next level, AI becomes embedded into operational workflows. It can perform defined tasks and accelerate human decisions.
3.    The organization can then begin to delegate specific activities to agents. Agents can execute defined workflows independently, with humans intervening for exceptions or decisions outside defined boundaries.
4.    As confidence increases, multiple agents can coordinate across multi-step processes. The organization moves toward agentic operations, where AI can plan and execute work toward an outcome rather than simply completing isolated tasks.
5.    The final stage is increasingly autonomous operations, where processes can self-orchestrate and optimize with humans primarily managing exceptions, governing outcomes and continuously redesigning the operating model.
This progression matters because every step requires stronger foundations. The technology cannot become capable of higher autonomy before the organization’s processes, data, governance or workforce are ready to support it. The goal should not be to reach the highest level of autonomy as quickly as possible, but to activate each level of autonomy only after the organization can operate the previous stage.
This variability also means that readiness will not be uniform. For example, one SAP process can be ready for autonomous execution while another still requires substantial human involvement. The maturity journey needs to be evaluated at the process and use-case level, rather than assuming that the entire SAP estate can move forward at the same pace.
AI delivers the greatest value when it is embedded into the way an organization already creates and delivers business outcomes. In a business-aligned, product-centric operating model, AI should be integrated directly into application squads across areas such as customer and consumer, supply chain and manufacturing, enterprise systems and data, AI and analytics. This approach helps ensure that AI models, agents and insights are focused on real business priorities and product backlogs.
At the same time, shared capabilities such as SAP platform management, DevOps, testing and automation provide the reusable foundation needed to scale AI safely and consistently across teams. Centralized governance then provides the guardrails needed to keep AI adoption responsible, compliant and aligned with strategic objectives. Together, this structure can help organizations move beyond disconnected AI experiments and turn AI into a scalable capability for delivering continuous, business-aligned value at greater speed.
Making autonomous AMS real requires an integrated ecosystem of capabilities rather than a single AI technology. IBM brings together AI, automation, engineering, observability, enterprise context and orchestration to help enterprises progressively move from AI-assisted operations toward autonomous SAP AMS.
IBM Consulting® AIOps provides the operational intelligence layer, connecting monitoring, ITSM, CMDB and other enterprise data to detect issues, support triage and enable proactive operations.
IBM Consulting Application Management Suite for SAP (ICAMS) and specialized agent suites apply AI across operations, enhancements, quality and testing, transition and platform management, enabling agents to execute increasingly complex AMS activities. We use a combination of Skill.md approach and agentic orchestration to drive business value and outcomes.
IBM Bob™ brings AI-powered engineering into SAP AMS, supporting code analysis, development, testing, documentation, remediation and continuous application improvement.
IBM Context Studio provides enterprise context and memory, helping AI understand customer-specific processes, architecture, engineering standards and accumulated operational knowledge.
IBM Consulting Advantage platform provides the orchestration and governance layer, coordinating agents and AI capabilities while helping enterprises manage autonomy across their broader technology landscape.
Together, these capabilities enable IBM to help enterprises build autonomy progressively, integrating with existing SAP and enterprise investments while maintaining human oversight where judgment and accountability remain essential.
Learn more about IBM SAP Managed Services. Discover how the overall IBM SAP AMS proposition brings together AI, automation, analytics and human engineering into bespoke solutions that meet the unique requirements of each client.
Easily design scalable AI assistants and agents, automate repetitive tasks and simplify complex processes with IBM® watsonx Orchestrate™.
Create breakthrough productivity with one of the industry's most comprehensive set of capabilities for helping businesses build, customize and manage AI agents and assistants.
Achieve over 90% cost savings with Granite's smaller and open models, designed for developer efficiency. These enterprise-ready models deliver exceptional performance against safety benchmarks and across a wide range of enterprise tasks from cybersecurity to RAG.

[Оригинал](https://www.ibm.com/think/insights/from-ai-pilots-autonomous-ams)