Presented by SAP
AI can generate code quickly, but for that code to run reliably inside a large enterprise — integrated with live systems, governed for compliance, and maintainable over years — foundational work is required that most organizations underestimate.
Michael Ameling, CPO of SAP Business Technology Platform, notes that while 81% of organizations have a detailed AI strategy, only 12–16% reach AI‑driven execution. The barriers rarely trace back to the quality of the generated code.
"Generating code is one thing," Ameling says. "Enterprise customers, including multinationals and large organizations, need to ensure there are no compromises in compliance or security. Code that runs reliably for ten or twenty years, as it does at many of SAP's largest customers, also has to be maintained, patched, and understood by whoever inherits it. Life cycle management, in other words, does not generate itself."
Why AI‑generated code fails in enterprise production environments
The productivity gains from AI code generation are real and well documented, but the ease of prototyping has given many organizations a misleading sense of readiness.
Teams often build something compelling, then discover they lack access to the data it depends on, the integrations it assumes, or the permissions required to run it in a real environment. AI amplifies an organization's existing data and process maturity, but it cannot substitute for it.
This dynamic grows more intense as AI moves from producing code to executing actions. Latency, cost, and system load rise when logic runs continuously against live data rather than producing a one‑time output. The performance requirements for an autonomous agent operating across a multinational's transaction systems are categorically different from those of a developer copilot.
Connecting AI‑generated logic to fragmented enterprise systems
The most underestimated architectural challenge in enterprise AI projects is integration. Real enterprise landscapes are not clean slates: they combine cloud services, legacy on‑premise infrastructure, fragmented data stores, and dozens of business applications that were not designed to interoperate. To make AI‑generated logic operate reliably across these systems requires a layer that unifies data access, process context, and governance — and that layer must be in place before any agent begins execution.
Organizations that view AI as a reason to defer infrastructure modernization are mistaken. "The question is not whether to modernize or not. Of course you need to modernize," Ameling says. "But the value you get on top of this is much higher with AI. Federated data access and harmonized process layers are not alternatives to upgrading a fragmented landscape, they're what make the upgrade worthwhile."
At a platform level, that translates into practical requirements: structured data integration, end‑to‑end process visibility, and the ability to discover and connect to APIs across modern and legacy systems. SAP's approach with the Business AI Platform draws on tools including Joule Studio, Integration Suite, Business Data Cloud, and the SAP AI Agent Hub enterprise architecture layer to provide that context. The goal is to give AI‑generated logic accurate, current knowledge of what a business is doing and how, rather than just access to raw data.
AI agents solve large problems by dividing them into smaller, autonomous tasks; each agent handles a specific domain while coordinating toward a shared outcome. A financial close, for example, involves dozens of discrete sub‑processes. Agents working in parallel on defined constraints can dramatically compress cycle times — but only if the underlying systems they interact with are coherent and accessible.
Governance and oversight for AI agents in production
When AI moves from assistant to operational actor, governance questions loom large: agents that trigger workflows, update records, and interact with live business systems need the same accountability framework applied to human employees — identities, defined privileges, and auditable behavior.
There are two distinct models:
- Principal propagation, where an agent acts on a user’s behalf and inherits that user’s permissions and scope.
- System‑triggered agents, where the agent operates under its own identity and role‑defined privileges, functioning more like an automated role than a personal assistant.
Both models require the same underpinning infrastructure: an agent hub where operators can see which agents exist, the APIs they can access, and what they are authorized to do. Observability must be operationalized for AI and combined with both technical and business evaluations.
"In production, openness is very important," Ameling says. "We use OpenTelemetry as a framework, so we can integrate with other solutions, for end‑to‑end observability of the tool, third‑party agents and the like."
Standard technical evaluations that test whether an agent produces consistent outputs are necessary but not sufficient. Business evaluations must determine whether an agent actually moves the performance indicators it was deployed to improve, and these assessments must be end‑to‑end. Where testing occurs is equally important: the traditional dev/test/prod cycle breaks down when a model produces different outputs against test data versus live data. Achieving trustworthy AI in production means accepting that validation looks fundamentally different from long‑standing engineering practice, including live environment testing and even A/B/C experiments to ensure reliable outcomes.
How AI‑driven code generation reshapes software engineering roles
Developers' roles are not disappearing, but their center of gravity is shifting. The productivity multiplier is substantial when developers run multiple coding agents in parallel across open terminals, each solving a different problem and each taking minutes to complete. But this introduces a new cognitive demand: humans must remain in the loop, track context across concurrent workstreams, evaluate outputs spanning large codebases, and make architectural judgments no agent should be trusted to make alone.
"The more specific and complete the prompt, the less intervention is required, and developers are learning that bringing more context upfront pays dividends in reduced back‑and‑forth," Ameling says. "But the output still needs to be understood, not just accepted."
The competitive edge will remain domain knowledge and intellectual property, not tooling alone. Companies that win will be those that most effectively encode their domain expertise into the systems they build. "A manufacturer's process expertise, a financial institution's risk logic, a logistics firm's routing intelligence — these are the assets that AI can accelerate, but only if the organizations that hold them do the work to make them accessible and usable," Ameling adds. "Protect that, and apply AI to accelerate your differentiation."
Closing
AI‑based code generation delivers fast results for prototyping and development tasks, but production‑grade, enterprise‑scale deployment requires integration, data readiness, governance, observability, and lifecycle management. Without these foundations, generated code will not become a long‑term, secure, compliant, and maintainable solution.
(Sponsored content — SAP. Sponsored articles are produced by a company that pays for the post or has a business relationship with the publisher and are clearly labeled.)
