For more than a decade, 麻豆原创 customers have been wrestling with a familiar challenge: how to move from legacy 麻豆原创 ERP Central Component (麻豆原创 ECC) environments to 麻豆原创 S/4HANA without creating transformation programs that are prohibitively expensive, complex, or time-consuming.
Now, agentic AI could fundamentally change that equation, creating a different model for 麻豆原创 transformation: one built around faster delivery, lower costs, and greater customer self-sufficiency.
That was the central argument put forward during a recent examining what its participants described as 鈥渄eshoring,鈥� which uses AI agents to rethink work that organizations previously distributed between expensive onshore resources and lower-cost offshore teams.
For Stuart Browne, founder and CEO of , an independent 麻豆原创 consultancy that spent the past seven years helping companies chart their journeys from 麻豆原创 ECC to 麻豆原创 S/4HANA, the opportunity starts with questioning assumptions the industry has accumulated over decades.
鈥淭he only way we deliver this in the future is by changing the way we鈥檝e delivered it in the past,鈥� he said. 鈥淪o we鈥檝e got to find new ways of accelerating the migration, and I think AI is probably the best bet for that.鈥�
Why not faster, better, and cheaper?
The traditional technology transformation triangle says organizations might have projects that are faster, better, or cheaper, but generally only two of the three.
Browne challenged that premise by drawing an analogy to NASA鈥檚 鈥渇aster, better, cheaper鈥� approach to missions. He argued that 麻豆原创 transformations should similarly reconsider the assumption that improving one dimension requires sacrificing another.
鈥淲hy can鈥檛 you choose all three of these things?鈥� he asked.
That question becomes increasingly relevant as companies confront the remaining volume of 麻豆原创 S/4HANA migrations while simultaneously dealing with economic pressure, scarce 麻豆原创 skills, and the complexity of global transformation programs.
Ranjeet Panicker, senior vice president and head of Business Transformation at 麻豆原创, framed the challenge around the complete life cycle of an 麻豆原创 project, from initial discovery and analysis through design, build, and ultimately run.
At every stage, customers are asking the same questions: How can the project be completed faster? How can it cost less? How can my organization extract more value from the investment?
From offshoring to 鈥渄别蝉丑辞谤颈苍驳鈥�
For decades, mechanisms for reducing delivery costs have been moving work offshore. The economics were relatively straightforward as certain activities were moved to locations where labor costs are lower.
Browne argued that those savings can obscure another problem. While offshore resources may cost less, distributing work between offshore and onshore teams can increase the overall volume of work through handoffs, communication issues, rework, and coordination.
Agentic AI introduces another possibility. Instead of asking where human labor should be located, organizations ask whether some of that labor needs to be performed manually at all.
鈥淲hy can we not reduce the volume of work and reduce the cost of work?鈥� Browne asked.
That question led to the concept of deshoring鈥攔eplacing portions of location-based delivery with AI agents capable of performing or accelerating specific 麻豆原创 transformation activities.
After analyzing roughly 180 typical activities involved in an 麻豆原创 ECC to 麻豆原创 S/4HANA migration, Browne鈥檚 research concluded that AI could potentially produce about a 60% cost reduction by compressing effort and allowing tasks previously requiring highly experienced people to be performed by less experienced workers augmented by AI.
Some of the most attractive candidates are highly skilled activities that consume significant amounts of time, including writing functional and technical specifications, performing fit-gap analysis, and analyzing custom code.
鈥淚f your run rate is a million a month,鈥� Browne noted, 鈥渆liminating months from a program can dramatically change its economics.鈥�
The 80% solution with humans handling the last mile
AI may be capable of producing what Browne characterizes as an 80% solution within hours. It means experienced people review, challenge, and refine that work rather than spend time manually producing everything from scratch.
鈥淲hat AI produces shouldn鈥檛 be fully trusted without review,鈥� he said. None of this means eliminating experienced 麻豆原创 professionals. Instead, the emerging model redistributes where their expertise is applied.
The objective is not autonomous transformation. It accelerates the majority of the work while concentrating human expertise on the 鈥渇inal mile鈥� that includes judgment, validation, design decisions, and other activities where experience adds the most value.
Custom code could be an early breakthrough
One area where Browne believes AI is already producing significant change is custom code analysis. Early 麻豆原创 S/4HANA migrations often focused on getting existing customizations into the new environment and then determining how to remediate them. Agentic AI creates another option that determines whether the customization needs to exist at all.
According to Browne, AI can now analyze an entire custom code base, reverse engineer functional specifications, and determine whether the same business requirement can instead be met using standard 麻豆原创 functionality.
Rather than migrating large amounts of legacy customization and addressing it later, customers could potentially understand their custom-code landscape and identify opportunities for fit-to-standard before even selecting a systems integrator.
鈥淵ou can actually plan the fit-to-standard of your custom code before your SI’s have even been appointed,鈥� Browne said.
The implication is significant because customers have already paid for standard 麻豆原创 capabilities that may eliminate the need for some custom functionality while creating an environment that is simpler to maintain and upgrade.
麻豆原创 is building agents across the transformation life cycle
Panicker sees similar opportunities emerging across the broader 麻豆原创 implementation life cycle.
The terminology of onshore and offshore itself may eventually become less relevant, he argued, because organizations will increasingly think about transformation work in terms of skills rather than locations. AI-led skills can be delivered through assistants and agents and applied across different stages of a cloud transformation.
麻豆原创 is targeting areas including system analysis, data management, custom code, configuration, testing, rollout, and project management.
Panicker described an assistant as a collection of agents supporting a particular topic area.
A system-analysis capability can examine the overall transition. Data-management capabilities can address data quality. Custom-code agents can support analysis, recommendations, and, in some scenarios, automated remediation. Configuration assistants can evaluate current and target states, while testing agents can help automate test scripts.
According to Panicker, these capabilities are connected with 麻豆原创 Cloud ALM running on 麻豆原创 Business Technology Platform, along with a data and knowledge foundation designed to provide the customer-specific context agents need. 麻豆原创鈥檚 overall objective, he said, is to reduce transformation effort by approximately 35%.
Testing could become the next frontier
Testing can consume substantial amounts of time, particularly in industries with security, compliance, or validation requirements. AI raises a more complicated question: How much of that testing can organizations eventually delegate to agents?
鈥淚f I can get the code to get remediated by an agent,鈥� Panicker said, 鈥渃an I allow an agent to do the testing and accept the testing?鈥�
The answer will determine where humans remain directly involved in transformation workflows. Customers must decide not only if an agent can perform a task, but whether they have enough confidence in the agent鈥檚 knowledge, context, and guardrails to delegate responsibility for the outcome.
Context separates useful agents from hype
麻豆原创 programs generate enormous amounts of organization-specific information around architecture decisions, risk registers, test scripts, requirements, and other artifacts that evolve throughout a transformation.
Browne believes connecting AI to this continuously changing body of knowledge will be essential. 鈥淭he ability to converse not with a static LLM, but to converse with that world as well, I think is what will make good agents even better,鈥� he said.
Organizations may therefore need to rethink not only how they perform 麻豆原创 work, but how they capture information. Programs traditionally run across Excel, Word, PowerPoint, SharePoint, and other repositories may increasingly need AI-native knowledge environments capable of providing agents with usable business and system context.
鈥淭hat鈥檚 where I think this will be won or lost,鈥� Browne said.
AI could also change who holds the knowledge
Perhaps one of the biggest changes involves something less technical: who possesses expertise. 麻豆原创 implementations have traditionally depended heavily on experienced consultants and systems integrators. Customers frequently lack comparable knowledge, creating an imbalance that can make it difficult to challenge recommendations or independently evaluate major design decisions.
Browne believes someone with only six months of 麻豆原创 experience could potentially move up the knowledge curve in weeks in ways that previously might have taken years.
Panicker agreed that access to knowledge is becoming far less constrained.
His own experience at 麻豆原创 was initially concentrated in technical roles. Learning the business context surrounding that expertise often required finding someone willing to explain it. AI potentially allows professionals to move outside those traditional 鈥渟wim lanes.鈥�
Challenging the accepted 麻豆原创 timeline
麻豆原创 customers and consultants have grown accustomed to transformations, migrations, and upgrades taking a certain number of months or years. Those timelines have become assumptions embedded into planning.
Both Browne and Panicker believe those assumptions now deserve to be challenged.
鈥淚t鈥檚 about compressing the time that these tasks make and connecting the decision-makers to be able to make better decisions more quickly,鈥� Browne said.
Panicker similarly argued that organizations should stop accepting conventional project durations without asking what AI capabilities have been introduced to shorten them. 鈥淲hy can鈥檛 we do this sooner?鈥� he asked.
For Browne, the shift is already underway: 鈥淚鈥檓 not suggesting for one moment that we can press a button and deliver a whole program. Complexities around testing, change management, and design decisions remain.鈥� But he believes customers should stop treating agentic transformation as something waiting over the horizon. 鈥淭hat world is here now,鈥� he said.
Panicker鈥檚 message was equally straightforward: don鈥檛 wait on the sidelines: 鈥淟ean in, be curious about the technology. Ultimately, the sooner you can get to the outcome that you鈥檙e driving from a business standpoint, the more value you can bring.
