Last Updated on 15. September 2026
For years, companies looking to cut the cost of their SAP development had an obvious answer: near- or offshoring. The logic was simple. If an hour of development is significantly cheaper at another location, direct development costs go down. So software development increasingly moved to wherever developers were available at lower cost.
AI-powered SAP development is changing that math. That’s because AI doesn’t just make the hour of development cheaper. It changes how many human development hours a project needs in the first place. When an experienced SAP developer can use AI agents to accelerate or automate large parts of code generation, analysis, testing, and documentation, a pure comparison of daily rates starts to lose its meaning. What matters instead is a different question: how much reliable output does a given budget actually produce?
This is exactly where we come in at mgm: experienced SAP developers, supported by AI agents, with a direct line to you, deep system understanding, and accountability that stays with people.
How AI Is Boosting Productivity in Software Development
That generative AI can boost productivity is now well documented. PwC’s Global AI Jobs Barometer 2026 shows that industries with high AI exposure are increasing their productivity more than less-exposed sectors. Since generative AI became widely available starting in 2022, this effect has only grown stronger.
The shift is visible in software engineering itself, too. McKinsey describes a move in 2026 away from the classic coding assistant toward agent-based development models: AI no longer just supports individual lines of code — it increasingly takes over entire work packages. It generates code, writes tests, analyzes bugs, prepares documentation, or implements changes across multiple files. In the companies McKinsey observed, this is even changing the size of development teams in some cases. Tasks that used to require larger teams are increasingly being handled by smaller teams working alongside AI agents. That’s the key point: AI isn’t just making the same development organization somewhat faster. It’s changing the division of labor between humans and machines.
Google’s DORA research reaches a similar conclusion. In the State of AI-assisted Software Development Report 2025, more than 80 percent of surveyed technology experts report productivity gains from AI. At the same time, DORA shows that AI doesn’t work in isolation — it amplifies the development organization that’s already in place. Strong engineering structures benefit more, while weak processes actually become more visible under the added speed.
“Using AI” on its own, in other words, is not yet a high-performing development model.
Why Experienced Developers Don’t Become Less Important as AI Improves
A common assumption is that as AI gets better at programming, fewer experienced developers will eventually be needed. Current research points instead toward a shift in their role. A systematic literature review by Mohamed, Assi, and Guizani evaluated 37 independent studies on the use of AI assistants in software development. The results show clear benefits for repetitive, well-defined tasks, but a more mixed picture for more complex work and for code quality.
A widely discussed study by METR made this especially clear. Experienced open-source developers worked on real tasks in codebases they knew very well. Using the AI tools available in early 2025, they took about 19 percent longer on average than without AI. That’s surprising at first glance, but it mainly shows one thing: generated code is not yet a finished result. Suggestions still have to be placed in the existing context, checked, adjusted, or discarded. In mature, complex systems especially, it isn’t enough for a solution to merely look plausible.
That’s what makes METR’s February 2026 update interesting. Newer data now points to positive productivity effects from modern AI agents, though METR notes that methodological limitations still make it hard to reliably determine their magnitude. Still, the direction is notable: as AI gets better at producing software, the developer’s task shifts. Less “how do I write this code,” more “is this the right solution for this system.”
McKinsey describes a similar shift. Systems thinking, architecture, breaking down complex requirements, evaluation, and ownership are all becoming more important. As a result, developers need — earlier in their careers — skills that used to be typical only of very senior colleagues.
For us at mgm, the conclusion is this: AI doesn’t reduce the economic value of an experienced SAP developer. It reduces the share of their time spent on work where their experience is, frankly, too expensive to use. Our senior developers don’t have to write every line of code themselves. They make sure that, in the end, the right line of code lands in your system.
Why Context Remains Critical in AI-Powered SAP Development
Enterprise software rarely consists of isolated programming tasks. A single change can touch data models, authorizations, interfaces, existing processes, extension concepts, release readiness, and future maintainability — all at once. In mature systems, there are also years or decades of customer-specific decisions layered on top.
AI can generate code. But it doesn’t automatically know why a particular process was built exactly that way for you, which interface is especially critical, or which architectural decision might cause problems two years from now.
That’s why control remains essential. Fraunhofer IESE points out that AI-generated code creates new requirements for verification and validation: results that can’t be fully traced, flawed assumptions, or hallucinated references can pose new challenges for existing quality assurance processes. And DORA shows that higher development throughput doesn’t automatically mean higher stability. Speed without robust engineering processes can actually amplify existing weaknesses.
For our approach, that means: AI takes on work, and our senior developers keep the responsibility. They define the solution approach, provide context, review results, and decide what gets adopted. The human role isn’t a downstream quality check — it’s part of the development process itself.
How Good Code Standards Make AI Agents More Efficient
One factor that’s often underestimated in discussions of AI-driven development is the quality of the development environment itself. In a controlled experiment with 540 agent runs in 2026, Sonar examined how cleanly structured code affects the work of a coding agent. On the cleaner codebase, the agent needed 7.2 percent fewer input tokens, 8.5 percent fewer output tokens, and roughly a third fewer repeat accesses to files it had already worked on. Estimated reasoning effort dropped by 11.1 percent, while the success rate stayed nearly the same. Clean code didn’t make the agent smarter — but it made it significantly more efficient.
That’s exactly the mechanism we take advantage of. Our agents don’t work with just the general knowledge of a language model. We embed them in a defined engineering context: mgm development guidelines, architectural principles, quality standards, and the SAP technology relevant to the task at hand — including RAP, CAP, Fiori with OData V4, ABAP Cloud, and the SAP Business Technology Platform. That’s what turns a generic AI tool into part of a structured development process.
How mgm Puts AI-Powered SAP Development into Practice
We pursue two paths in AI-powered SAP development:
- For encapsulated solutions, we carry out a large part of the development within our own tool and AI environment. Architecture, development guidelines, testing, and quality assurance are entirely in our hands. For you, the effort is mainly around integration and customization. Our eRechnung (e-invoicing) tool is a good example of this principle: the solution doesn’t need to be built from scratch for every customer. A substantial part of the development work is already done, and you benefit from a proven, KoSIT-validated solution. Visit the “eRechnung for SAP” product page
- The second path is custom development directly in your SAP system. Here, our senior developers work alongside AI agents. The agents take on suitable parts of development, analysis, testing, or documentation. Our developers provide system context, architectural decisions, and subject-matter judgment, and they own the result. Put simply: machine capacity for creation and analysis, human experience for context, architecture, and accountability.
This model is the foundation of our AI-powered SAP development out of Germany: not more people thrown at a problem, but higher productivity from experienced developers working with AI.
How AI Is Changing the Offshoring Math in SAP Development
The economic advantage of classic offshoring models is based on a lower daily rate. As long as a task takes roughly the same number of human development days, that advantage is hard to beat. AI changes exactly that assumption. If a senior developer, working with agents, delivers more development output in the same amount of time, the number of development days needed goes down. Going forward, it’s no longer just two daily rates competing with each other.
Put simply, two models stand opposite each other:
- On one side: senior developers plus AI capability, direct communication, and low coordination overhead.
- On the other: greater development capacity at a lower daily rate, but with additional coordination and management effort.
The lower daily rate can still be an advantage. But it’s no longer automatically the same thing as the cheaper project.
This shift is now being discussed academically as well. An article in the journal Wirtschaftsinformatik & Management describes an approach called “AI-Shoring,” which combines qualified specialists with specialized AI agents. The goal: less human resource input, shorter turnaround times, and, at the same time, high quality and availability. Software development is explicitly named as one of the fields where this applies.
That adds a new dimension to the classic make-or-buy, or shoring, question. It’s no longer just: where is an hour of development cheaper? It’s also: how many human development hours do you actually still need, and what’s the quality of the human time that remains? For complex SAP development in particular, we consider this the decisive question.
Why a Senior Developer with AI Delivers More Economic Value
Our approach, then, isn’t about replacing developers with AI. It’s about focusing their experience on the tasks where it delivers the most value: architecture, system understanding, subject-matter context, evaluation, and accountability. AI increasingly handles the underlying production work.
That makes experience scalable — and that’s exactly where the economic potential lies. The cheapest developer isn’t necessarily the one with the lowest daily rate. What matters is how much reliable, integrable, and maintainable software a given budget actually produces. For suitable SAP development tasks, a small team of experienced mgm SAP developers, supported by specialized AI agents, can already be an economically viable alternative to larger near- or offshore teams today.
And the more the actual generation of code can be automated, the less it will matter, going forward, where code is written. What will matter more is who understands which code is the right code.
Want to know what AI-powered SAP development could look like in your system environment? You’ll find an overview of our SAP offering on our website “AI-Powered SAP Development from Germany,” or reach out to us directly.
Sources
- PwC: Global AI Jobs Barometer 2026
- McKinsey: AI-powered software development. Wie Technologie die Regeln neu schreibt, 2026
- Google Cloud / DORA: State of AI-assisted Software Development Report 2025
- Mohamed, A., Assi, M., Guizani, M.: The Impact of LLM Assistants on Software Developer Productivity. Eine systematische Literaturübersicht, 2025
- METR: Developer Productivity Study sowie Update vom Februar 2026
- Sonar: A cleaner codebase results in less token usage, 2026
- Fraunhofer IESE: KI in der Softwareentwicklung. Zwischen Produktivitätsschub und Vertrauenskrise
- Springer, Wirtschaftsinformatik & Management: AI-Shoring als Zukunftsstrategie. Mit KI-Agenten dem Fachkräftemangel begegnen, 2025
