AI excels at coding, yet strategic context and architectural design remain inherently human.
Understandably, concerns about AI’s impact are escalating, particularly within the engineering community. Software engineers are frequently encountering news suggesting that AI poses a threat to their careers.
However, this apprehension, though valid, doesn’t accurately represent the current capabilities of these systems or their probable short-term development.
Despite pervasive discussions, agentic AI remains limited to rule-based, deterministic environments. It effectively handles coding, refactoring, validation, and pattern recognition. Yet, it struggles when faced with ambiguity, shifting human priorities, complex non-binary trade-offs, or situations demanding empathy and nuanced interpretation.
Authentic engineering extends beyond deterministic tasks. Product development involves more than just writing code; it encompasses strategic, human, and situational context, an area where AI currently lacks proficiency.
Current State of Agentic AI
Presently, agentic AI demonstrates considerable skill within specific, confined parameters. It performs exceptionally well in settings with explicit expectations, stringent rules, and stable objectives. For tasks like code analysis, test generation, or identifying bugs based on historical data, it proves highly effective.
These AI systems function akin to trains on fixed rails: swift, efficient, and able to traverse any laid path. However, should business objectives or strategic directives change, AI agents will continue on their predetermined path, oblivious to the altered destination.
Strategy: An Open and Evolving System
Engineering is not an isolated discipline. It is directly influenced by business strategy, which in turn shapes product direction and technical priorities. Each of these interconnected levels introduces unique biases, interpretations, and human judgment.
Furthermore, these decisions are dynamic, evolving with shifting urgency, leadership directives, and customer demands. A strategic pivot rarely disseminates through an organization as a clear, deterministic instruction. Instead, it surfaces through various informal channels: a leadership brief, a client conversation, casual discussions, internal messages, or private meetings.
Currently, agentic AI is not designed to operate in this manner.
Agentic Systems Lack Strategic Context
For agentic AI to truly advance, it must transcend static logic, incorporating dynamic, evolving strategic and directional context.
This implies moving beyond merely defining a function’s purpose to evaluating its continued relevance, confirming the priority of its parent initiative, and ensuring it aligns with recent changes in customer needs or product strategy.
Current AI tools operate detached from this critical layer. They fail to absorb the intuitive signals that product managers, designers, or tech leads inherently process. They cannot assimilate organizational realignments and adapt their actions appropriately.
Until this capability is developed, these systems will remain limited to deterministic assistance rather than becoming genuine collaborators.
Our Future Development Aims
To be explicit, the objective is not to replace human talent. Instead, it’s about enhancing human potential—not solely through task automation, but by valuing the essential human perspective inherent in every impactful product.
As agentic AI assumes more of the routine, laborious, and repetitive aspects of engineering, it liberates human professionals to concentrate on higher-value activities: creating innovative solutions, tackling complex challenges, and designing with meaningful impact.
Allow AI to provide foundational support, highlight insights, and validate work. Reserve for humans the roles of interpretation, strategic direction, and creative development—executed with purpose, drive, and diligence.
Achieving this requires agentic systems that not only function within codebases but also grasp broader contexts. We need systems capable of understanding not just explicit commands, but also evolving conditions. These systems must be able to adapt their understanding as priorities shift.
Ultimately, the aim extends beyond mere automation. It encompasses improved alignment, more effective time utilization, and superior results for the end-users of our creations.
This necessitates developing tools that go beyond code interpretation, truly comprehending the purpose of our creations, their intended audience, the implications, and their overarching significance.
—
New Tech Forum offers a platform for tech industry leaders, including vendors and external contributors, to delve into and debate cutting-edge enterprise technologies with extensive coverage. Content selection is editorial, focusing on technologies deemed crucial and highly relevant to InfoWorld’s readership. InfoWorld strictly prohibits the submission of marketing materials for publication and retains full editing rights for all contributed content. For all inquiries, please contact doug_dineley@foundryco.com.
