The growing popularity of vibe coding—using AI to generate code from natural language prompts—has been hailed as the end of the road for developers. After all, if up to 40 percent of companies are producing apps this way today, and product managers can spin up a working prototype or a financial analyst can create a single-use application to get through year-end close, what need will there be for a computer engineering degree? But according to Muhammad Alam, SAP board member, Product and Engineering, that’s not the future (or the present) that SAP envisions for devs: It’s only the beginning.
“Developers don’t just code anymore—[they] design intelligent workflows and supervise AI agents to shape real business outcomes,” Alam said from the SAP TechEd Berlin stage this past November. Even with the added resources provided by AI tools and databases, teams are stretched thin and are delivering on tighter deadlines. On another panel, he noted that “even though I have 35,000 or 40,000 colleagues, there is about 200,000 people’s worth of [development] backlog.” So it’s no surprise that with the firm’s announcement of its latest enterprise application, data and AI releases, SAP has made it clear that their solutions aren’t rooted in doing more with less, but developing smarter, more secure tools so devs can do more—faster.
Over the past two years, the speed gains of using generative AI to ship products has been a double-edged sword for large organizations. The combination of developers’ tendencies to experiment and the rapid pace of AI coding tools hitting app repositories has meant that devs are adopting generative AI coding assistants faster than security and governance teams can evaluate them. In fact, one industry survey found that more than half of developers polled didn’t use AI tools provided by their IT department. And still another survey found that only 18 percent of respondents said that their organization has a list of approved tools for vibe coding.
The solution isn’t restricting access, which would either stifle innovation or drive users to further obfuscate their methods, but to furnish approved systems that provide security by being tightly integrated into an organization’s core systems, and natively include access to developers’ preferred coding tools.
A Strong Foundation
“Our strategy at SAP is very simple,” said Alam. Rather than bolting AI on top of their applications, SAP has embedded it into the tech stack as a new integral layer. By bundling apps that cover the entire spectrum of business processes—finance, procurement, supply chain, HR, customer experience—companies produce harmonized, managed, and governed data that can then be fully accessed by AI, which then feeds data back into the apps. Harmonization means standardizing data across systems so different departments can work together—same product codes, same date formats, same customer IDs. “[We believe that] the seamless app, data, and AI flywheel creates unique and exponential value for organizations,” said Alam.
This “AI-native foundation” means that teams can build and speed up iteration throughout the SAP Business Suite without the security or governance risks. For instance, in addition to incorporating the ability to natively use their preferred coding assistants, advancements within SAP Build will allow developers to build and deploy fully custom agents in Joule Studio. These agents—autonomous software programs that can understand business intent, make decisions, and take action across multiple applications with or without human intervention—are being implemented across enterprise environments, and they’re at the core of SAP Business Suite. Joule’s role-aware assistants draw on applications and data across SAP Business Suite to partner with workers in their specific roles, allowing the workers to tap into the right agents for the job, set up the workflows, and manage the agents based on predefined rules and actions. And because the AI, apps, and data live within a single ecosystem, agent deployment can be streamlined.
For instance, deployed in HANA Cloud, SAP’s database-as-a-service, these agents support Model Context Protocol (MCP), allowing them to access multimodel data in a single query. Put simply, MCP is a “master key” to different data types, so Joule agents can answer complex questions without juggling multiple tools. This allows them to access the full context of a data request: navigating relationships across customers and suppliers, jumping databases, and querying multiple AI models at once. To facilitate this access, developers can now automatically generate knowledge graphs from SAP HANA Cloud metadata. To provide developers and customers more flexibility in their cloud architecture and data storage while maintaining critical business details and without duplicating data, SAP has announced partnerships with data management and analytics partners Snowflake and Databricks.
These features culminate in SAP’s new foundation model for predicting business outcomes from structured data, SAP-RPT-1 (pronounced rapid one), announced at SAP TechEd Berlin. While businesses run on tables, spreadsheets, and databases—what’s known as structured data—large language models (LLMs) have historically struggled to understand tabular layouts and the relationships between data points. Traditionally, you’d need to implement specialized machine learning models, training a separate model for each specific task—this is expensive and time-consuming (typically barriers to things getting done). SAP-RPT-1 combines all of these capabilities into one pretrained model that understands relational business data, keeping business context readily available for developers and specialized teams. SAP has even released an open source version that can run locally for testing.
Developers as Orchestrators
Fully weaving a business’s data with an AI layer makes the data more accessible to teams and developers, but returning to Alam’s point, it means that developers take on a new role: They’re the conductors of an agentic orchestra, setting up agents to code, pull data, or autonomously complete tasks. But this role shift, unlocked by generative AI, creates ripples throughout the product org chart. Now, rather than generate spec sheets and briefs, product managers can vibe code prototypes and deliver them to engineering to complete or build out with a clear sense of mission and execution.
Similarly, scrum team structure can be rethought as new tools are integrated into data platforms. SAP is putting this into action with a selection of “front-runner teams” piloting various AI tools, and in some cases, removing headcount and technical constraints to test out different processes and role definitions to discover what gains could be had at the edge. And the results spoke for themselves. “Some of the productivity examples we’re getting are pretty phenomenal,” with some teams seeing throughput increases of sevenfold and twelvefold, Alam said.
These results highlight what is possible when organizations take the time to rethink their structures; net-new technologies require net-new ways of working. The rapid rise of generative AI and the future-state evolution of agents as assistants to agents as fully autonomous actors in a no-code environment will require some upskilling at the organizational and worker level. To better equip the workforce as the nature of … well, work, shifts, SAP is expanding its certification programs to reach more people in more places. Through a partnership with a global online learning platform, SAP is committing to upskill 12 million workers with AI skills by 2030. To get there, SAP is implementing AI-driven accessibility features and transforming its testing approach to incorporate practical assessments and hands-on training.
As the nature of work shifts and developers are unconstrained in their access to harmonized data and the tools they need to deliver results, the developers and managers that will thrive are the ones that are comfortable at the conductor’s podium, guiding teams of agents, and learning to do more with an ever-evolving toolbox.
Learn more about SAP’s AI initiatives at SAP.com
.jpg)
