SAP transformation has an economics problem.
The technology is more capable than ever; S/4HANA offers a stronger foundation for data, automation, integration and AI. Yet the work required to get there still depends heavily on scarce human capacity, manual analysis and long chains of handovers.
That model is expensive, it’s slow, and it makes progress dependent on who’s available, what they already know and how quickly knowledge can be transferred between teams.
The answer isn’t simply to move the work somewhere else – it’s to change the amount of work that needs to be done by expensive human resources in the first place.
That’s the idea behind Deshoring.
The accepted model is reaching its limits
Most SAP programmes still follow a familiar pattern:
A customer brings in a systems integrator; specialists are assigned to workstreams; analysts gather information; consultants review it; architects make recommendations; and teams produce the documents, specifications, test scripts and plans needed to move the programme forward.
The work then moves through multiple review and approval stages, often across internal teams, external consultants and offshore delivery centers. Every handover creates friction, every new consultant needs context, and every decision has to be explained again before it can be applied consistently.
This isn’t a criticism of the people involved. Most SAP professionals are working hard within a delivery model that has remained largely unchanged for years. The problem is that the model is built around human availability and billable time, even when much of the work involves analysis, comparison, documentation, knowledge recall, or the production of structured first drafts.
That distinction matters because SAP transformation is becoming more complex at the same time as experienced SAP capacity is becoming harder to secure.
S/4HANA programmes require knowledge across processes, architecture, integration, data, security, clean core principles and new development approaches. A single consultant can’t hold all of that knowledge, and a project team can’t always access the right expertise at the moment it’s needed.
As a result, some of the most expensive people on the programme spend their time answering repeated questions, reviewing documents, searching for information and correcting work that should have been right the first time.
The challenge isn’t that people are no longer valuable, it’s that too much of their time is being consumed by work that agentic AI can now accelerate, prepare or support.
What changes when AI can work continuously?
Generative AI can produce content in response to a prompt. Agentic AI goes further by supporting a sequence of activities such as analysis, comparison, drafting, checking and refinement, while using the information available to it to determine what should happen next.
That means an AI agent can review a body of material, identify relevant patterns, compare possible options, prepare a structured output and explain the reasoning behind its recommendations. It can also recall information from previous work and apply it consistently to the next task.
Unlike a human project resource, an agent doesn’t need to be staffed for a particular phase of the programme, wait for a knowledge transfer session or start again every time a workstream changes direction. When it’s connected to the right information, it can work continuously across the programme and help your team move from raw information to a useful first output much faster.
Consider the work involved in a typical S/4HANA programme. Teams may need to review custom code, assess clean core opportunities, compare current processes with standard S/4HANA capabilities, produce fit-to-standard analysis, draft functional and technical specifications, create test scenarios and test data, and maintain a clear record of risks, decisions and open questions.
These tasks still need human oversight, but they don’t always need a senior consultant to start from a blank page.
When an agent prepares the analysis and the first version of the work, your team can focus on the activities where human expertise creates the greatest value: applying business context, challenging assumptions, making decisions, managing risk and securing approval.
That’s where delivery economics begin to change.
Deshoring is not offshoring
Offshoring moved work to humans in lower-cost geographies in order to create a labor cost advantage.
Deshoring takes a different approach. It uses agentic AI to reduce the volume and cost of work that requires expensive human capacity, while building stronger capability inside your own organization.
It’s not a geographic strategy, and it is not simply another way to resource an SAP programme; it’s a different delivery model in which intelligent, always-on AI agents work directly with your team to handle analysis, documentation, comparison and other high-effort activities that traditionally create dependency on external consulting capacity.
The goal isn’t to replace your team with technology, it’s to give your team a persistent and increasingly capable resource that works alongside them, retains programme context and helps them take greater ownership of the transformation.
This changes the relationship between your organization and its systems integrator. Instead of relying on external consultants to perform every analytical and documentation task, your internal team can direct the work, review the output and make the decisions that matter, with agentic AI supporting the work in between.
The Deshoring Manifesto sets out the wider case for this approach, including the principles of customer self-sufficiency, broader knowledge, continuous learning, acceleration, independence from commercial bias and greater cost transparency.
The Fold: where delivery economics change
The most useful question isn’t, “what can AI do?” it’s, “which tasks should no longer be done in the same way?”
The Fold is the point where the effort and skill intensity of a task can be compressed.
A task folds when agentic AI can reduce the time required to complete it, reduce the level of skill needed to produce the first version, or both. The task doesn’t necessarily disappear; instead, the way it’s delivered changes.
For example, a senior consultant may no longer need to spend two days producing a detailed first draft of a functional specification. An agent can prepare the draft from the relevant process information, business requirements, previous decisions and technical constraints, leaving the consultant to review the logic, challenge the assumptions and make the decisions that require experience.
The same principle applies to testing. An agent can create a structured starting point for test scenarios and scripts using the process design, requirements and expected outcomes. Business users and functional experts can then refine the scenarios, confirm that they reflect the way work needs to happen and approve them for execution.
The Fold doesn’t mean that every activity can or should be automated. Complex stakeholder alignment, organizational change, leadership decisions and accountability remain human responsibilities. The opportunity is concentrated in high-effort, repeatable and knowledge-intensive tasks where people currently spend significant time analyzing information, comparing options, preparing documents or recalling what has already been decided.
The Fold
The Fold is the extent to which agentic AI can compress the time and skill intensity required to complete SAP programme tasks. It’s where work becomes faster to produce, less expensive to review and easier for your internal team to own.
The human last mile still matters
Deshoring isn’t about removing people from transformation, it’s about putting people where their judgment has the greatest impact.
Your team should remain responsible for context, challenge, decisions, governance and approval because they understand the commercial priorities, operating model, risk appetite and organizational realities that an AI agent cannot determine on its own.
An agent can identify possible clean core opportunities, but your team must decide whether a change is practical, valuable and acceptable.
It can compare a current process with S/4HANA standard functionality, but business leaders must decide what the organization is willing to change.
It can draft a test scenario, but business users need to confirm whether the scenario reflects how work must happen in reality.
This is the human last mile, and it’s a critical part of a controlled delivery model. The value comes from combining continuous machine capability with informed human judgment, rather than expecting either one to deliver transformation alone.
Your programme context is the multiplier
An agent becomes significantly more useful when it understands your programme.
That means connecting it to the knowledge your team is creating every day, including:
- Programme risks and mitigations
- Design decisions and their rationale
- Process information
- Business requirements
- Configuration and architecture documentation
- Open questions
- Agreed answers
- Defect history and previous analysis.
Without this context, an agent can only provide general SAP information – with it, the agent can support decisions in the context of your landscape, your processes and your programme history.
This is where the economics become more compelling. Your knowledge is no longer trapped in meeting notes, project folders or the memory of individual consultants. It becomes a living source of programme capability that can be searched, compared and applied continuously.
The longer the programme runs, the more valuable that capability becomes because each decision, issue and outcome improves the context available for the next task.
Where should you start?
You don’t need to redesign your entire SAP delivery model on day one. Start with tasks where the baseline is clear, the output can be reviewed and the value of a faster first result is easy to measure.
Custom code analysis is a strong starting point because it involves large volumes of technical information and requires teams to identify what existing code does, where it creates risk and which elements may be replaced, retired or redesigned in line with clean core principles.
Fit-to-standard analysis is another practical candidate because an agent can compare current processes with S/4HANA capabilities and help identify where the organization can adopt standard functionality rather than carry forward unnecessary complexity.
Specifications can also fold when an agent creates consistent first drafts from agreed requirements, process information and design decisions, giving functional and technical experts more time to review the content and resolve the issues that need their attention.
Testing offers a further opportunity because test scenarios, test scripts and test data can be produced more quickly while maintaining clear links to requirements and expected outcomes.
These tasks share three characteristics: they require significant effort, they involve specialist knowledge and they follow structures that can be reviewed. That makes them useful places to establish a baseline and test an agentic workflow.
The practical next step
Choose one task and measure the baseline. Record how long it takes, who needs to be involved, what information is required, how much rework is typical and where delays occur.
Then apply an agentic workflow using the relevant programme context. Ask the agent to analyze the information, produce the first output and explain its reasoning. Have your team review the result, record the time saved and assess the quality of the work before deciding whether the workflow should be expanded.
That evidence will tell you more than a broad AI strategy ever could because it will show exactly where your programme can fold, where human review remains essential and what needs to change before the approach can be applied at scale.
The future of SAP delivery will not be defined by whether organizations use AI; it will be defined by whether they use AI to change the economics of transformation.
The first step isn’t to replace your people, it’s to stop asking your most expensive people to spend their time on work that an agent can prepare continuously, consistently and at a lower cost.
Start by putting a number on the opportunity.
You can use our Agentic AI calculator to estimate how much you could save by adopting Deshoring in your SAP programme.
