Agentic AI and S/4HANA migration: Rethinking the economics of ERP transformation

For many enterprises, SAP S/4HANA migration is no longer a question of intent. They understand the need for a cleaner, more intelligent ERP. Where the transformation hits a bottleneck is the path to get there without turning the program into a multi-year exercise in cost escalation, delayed value, and post-go-live firefighting.

Data shows that S/4HANA transformations take, on average, 30% longer than planned, while more than 60% of companies report deviations in budget, schedule, and result quality.

Most CIOs and transformation leaders know why this happens. Years of customization, ABAP code, Z-programs, legacy interfaces, fragmented master data, process exceptions, and manual workarounds all surface during migration. Traditional delivery models handle this complexity through large teams, sequential workstreams, manual analysis, and repeated review cycles. As the landscape gets more complex, this model means the effort expands almost linearly.

Agentic AI has the potential to change that equation by accelerating parts of the transformation lifecycle that still depend heavily on manual analysis and execution. However, there is an important qualification. The transformation economics cannot be viewed only through the hours AI might remove from a migration plan. Enterprises also need to understand the platform, infrastructure, and consumption model required to run agents reliably at scale.

That question has become even more relevant with the evolution of SAP’s own platform strategy.

At SAP Sapphire 2026, SAP renamed the SAP Business Technology Platform (SAP BTP) to SAP Business AI Platform (BAIP). The name change heralds the advent of a unified foundation for building, contextualizing, and governing AI agents, applications, and workflows. The erstwhile BTP capabilities are now a core part of this broader BAIP platform, alongside the data, AI, and business context required for enterprise AI. For S/4HANA transformation, this makes the underlying platform more central to the Agentic AI conversation.

Agentic AI starts with readiness

In discussions around Agentic AI, it is easy to move quickly to the use case: identify what the agent should do and estimate the effort it could save. However, before we talk about what agents can accelerate, we have to ask whether the enterprise has the infrastructure required to enable it, and what that readiness means for the economics of the transformation.

Agents that plan, query, act, and self-correct need compute and orchestration, access to appropriate models, and APIs through which they can interact with S/4HANA and surrounding applications. Also, an agent working on an S/4HANA transformation cannot rely on generic SAP knowledge alone. It needs access to the customer’s own landscape: custom code, configuration, business processes, functional specifications, integrations, data structures and business rules.

Once agents begin acting on enterprise systems, identity, security, auditability, observability and execution controls all become part of the architecture.

Within the SAP ecosystem, that can involve capabilities across SAP AI Core, AI runtimes, model access, SAP HANA Cloud, integration and event services, security services, and governance tooling. The exact architecture will differ by enterprise and use case, but the principle is the same: autonomous workflows have infrastructure dependencies that need to be understood early.

This has direct implications for the business case.

An enterprise may budget for an Agentic AI-led transformation based on expected productivity gains, only to discover later that additional platform services, model consumption, compute, storage, or integration capacity are required. Those costs can be problematic if they are not designed into the transformation.

Therefore, decisions around agent deployment must start with the right questions: What capabilities already exist? What needs to be provisioned? What will be consumed as agents scale? And where will the investment continue to create value after the migration is complete?

Where Agentic AI can change the delivery model

Once that foundation exists, the productivity opportunity becomes much more tangible across the transformation lifecycle.

• Custom code is a good example. Most large ECC landscapes contain thousands of custom objects with very different levels of relevance. Agents can support classification, compatibility analysis and first-pass remediation, allowing SAP specialists to focus their time on the objects where architecture or business judgment is required.
• The same principle applies further upstream. Functional specifications, business requirements, and existing documentation contain information that traditionally has to be interpreted and translated repeatedly as work moves from design into development and testing. AI-first delivery models can ingest those artifacts and help generate downstream specifications, code, and test assets while retaining more of the original context.
• Testing is another major opportunity. Regression testing typically expands rapidly as processes cut across modules, geographies, local rules and non-SAP systems. AI can support test design, generate scripts, execute repeatable tests and identify failure patterns much earlier in the lifecycle.

At ITC Infotech, this is an area we are already exploring through SAPFabrik and our partnership with Humanize. Humanize brings an AI-first delivery model into the SAP lifecycle, with agents that can ingest artifacts such as business requirements and functional specifications and help generate downstream technical specifications, code and test scripts. This can reduce the repeated interpretation and hand-offs involved as work moves from design through development and testing.

This is a more meaningful productivity opportunity than automating isolated tasks. A significant part of transformation effort sits in the transitions between teams: understanding an existing artifact, recreating context, producing the next artifact, reviewing it and then handing it over again. Agentic AI can begin to compress those transitions.

SAPFabrik combines Humanize with AI assets being developed on SAP Business AI Platform. It can also integrate with ITC Infotech’s broader platforms where their capabilities are relevant to the transformation. For example, an S/4HANA program can draw on iQStudio, our quality engineering and testing suite, for broader testing requirements, while OmniFabrik can bring AIOps capabilities into the SAP environment where required.

Rethinking the economics means looking at both sides of the equation

On one side is delivery productivity. AI can reduce repetitive analysis, increase parallelism, automate portions of documentation and testing, improve reuse and allow scarce SAP expertise to concentrate on decisions that require judgment.

There can also be downstream benefits. Better code classification, earlier quality checks and stronger clean-core discipline can reduce the amount of legacy complexity that gets carried into S/4HANA, improving maintainability and upgradeability after go-live.

On the other side is the economics of running Agentic AI itself. Unlike conventional application transactions, agents may reason through multiple steps, call models and tools, evaluate outputs and retry before completing a task. Each of those interactions consumes tokens and compute, so cost depends not only on what an agent does, but on how efficiently it gets there.

Poorly bounded loops or unnecessary model calls can quickly increase consumption. Token budgets, iteration limits, model selection and runtime monitoring therefore become part of the operating economics of Agentic AI, not simply technical safeguards.

The business case for Agentic AI therefore needs to account for both the productivity gains and the cost of enabling them. How much implementation effort can be compressed? What new platform and consumption costs will be introduced? Which capabilities can be reused across the transformation? And how much value will the same foundation create once S/4HANA is live?

The investment should outlive the migration

A transformation program is temporary; the SAP environment it creates will operate for years.

The platform and AI capabilities established during an S/4HANA program can therefore have a much broader role than migration acceleration. The same foundation can support AI-enabled application management, intelligent testing, AIOps, and business agents across functions such as finance, supply chain, and sales.

Our SAP AI roadmap takes a long-term view by considering: AI in how we deliver transformation, AI in how we run the SAP environment, and AI in how the business performs on top of it.

A different CXO decision

For CXOs, the S/4HANA migration discussion can no longer revolve only around how many people a program requires or how quickly an implementation partner can complete the work.

The more useful questions are becoming: Which parts of the transformation can genuinely be compressed through AI? What enterprise context will agents need to do that work effectively? Is the underlying platform ready? What will that capability cost to build and operate? And what can the enterprise continue to do with it after migration?

S/4HANA transformation remains one of the few opportunities an enterprise gets to rethink its ERP foundation at scale. Agentic AI can make that transformation more productive, but realizing the value requires treating AI readiness, architecture and economics, as part of the transformation design from the beginning.

Author:

Jayabrata Maitra,
Technology Practice Head – ERP