A new study from researchers at Chalmers University of Technology and the Volvo Group challenges the prevailing narrative that artificial intelligence (AI) is rendering software developers obsolete. Instead, the paper posits that AI agents are expanding the landscape of software engineering through what they describe as “semi-executable artifacts.”
These artifacts include prompts, workflows, policies, escalation rules, and decision routines, which directly influence system behavior but depend on human or probabilistic interpretation for execution. This perspective emphasizes a broader understanding of software engineering beyond just traditional coding, suggesting that AI systems are reshaping the field in significant ways.
At the core of the paper is the “Semi-Executable Stack,” a diagnostic model that is structured in six concentric rings. The innermost ring consists of classic code, followed by prompts and natural language specifications, orchestrated agent workflows, control systems like guardrails and monitoring, operational organizational logic, and, finally, social and institutional considerations such as frameworks like the EU AI Act. This model illustrates how software engineering has historically prioritized the innermost rings while the outer layers are becoming increasingly crucial.
The authors highlight that while traditional software engineering has focused primarily on rings one and two, current trends indicate that rings two through five are now critical engineering objects. Moreover, ring six, which addresses social fit, is becoming a determinant of practical effectiveness in AI deployment.
The researchers identify a significant gap in engineering methodologies applicable to the outer rings, particularly five and six. While established methods for coding and debugging exist within the first three rings, frameworks for managing decision-making routines and regulatory compliance are still underdeveloped. This gap underscores a shift in focus for software engineers as they adapt to new AI capabilities.
To support their argument, the researchers present three key observations. Firstly, AI does not need to match the output of the top engineers to effectively change team dynamics; it only has to be sufficiently competent. Secondly, the scale of AI deployment is often more valuable than achieving peak performance, as everyday applications may collectively offer greater organizational benefits. Lastly, as domain experts increasingly utilize natural language to build their own systems, maintaining clean engineering practices becomes even more essential.
Critiques regarding AI’s reliability and code quality are reframed by the researchers as engineering challenges rather than insurmountable barriers. For instance, when AI experiences “hallucinations,” the emphasis on testing and monitoring becomes increasingly significant. Additionally, the phenomenon of “prompt drift,” where adjustments to a prompt lead to unforeseen alterations in system behavior, highlights the complexities that arise as AI-generated code proliferates.
The transition to incorporating AI into engineering practices transforms into its own engineering challenge, especially as nuanced judgment becomes more valuable amid the cheapening of low-level tasks. The authors assert, “The scarce skill shifts from building faster to deciding what is worth building or changing, which ring is actually being changed, how that change will be validated, how it will be governed, and how it will be maintained over time.” This shift suggests a broader organizational redesign rather than mere efficiency in traditional software tasks.
This paper was presented as part of a keynote by Robert Feldt at the Agentic Engineering 2026 Workshop in Rio de Janeiro and reflects insights drawn from industrial collaborations in the automotive sector with Volvo partners. The findings indicate a pressing need for the software engineering community to adapt to this evolving landscape, as AI continues to reshape roles and responsibilities within the industry.
See also
DeepSeek Launches 1.6 Trillion Parameter V4 Model on Huawei Chips Amid U.S. IP Theft Claims
Tesseract Launches Site Manager and PRISM Vision Badge for Job Site Clarity
Affordable Android Smartwatches That Offer Great Value and Features
Russia”s AIDOL Robot Stumbles During Debut in Moscow
AI Technology Revolutionizes Meat Processing at Cargill Slaughterhouse
