From Prompt-Response to Goal-Directed Systems: The Evolution of Agentic AI Software Architecture

TL;DR

The paper tackles the problem of evolving AI from stateless prompt-response systems to autonomous, goal-directed agents capable of perception, planning, action, and adaptation within governed, auditable loops. It syntheses classical agent theories with modern LLM-driven practices to propose a governed reference architecture, a taxonomy of multi-agent topologies, and an enterprise hardening checklist, situating these within industry platforms like Kore.ai, Salesforce Agentforce, TrueFoundry, ZenML, and LangChain. The key contributions include a separable cognition–execution architecture with typed tool interfaces, a structured view of coordination topologies and their failure modes, and pragmatic governance and observability patterns that enable reproducible, auditable deployments. The findings suggest a convergence toward standardized stacks that couple orchestration with governance, implying a maturation toward web-service-like protocols and registries to support scalable, composable autonomy. Open challenges remain in verifiability, interoperability, safe autonomy, and data sovereignty, underscoring the need for formal guarantees, interoperable contracts, and scalable governance frameworks as agentic AI moves into production at scale.

Abstract

Agentic AI denotes an architectural transition from stateless, prompt-driven generative models toward goal-directed systems capable of autonomous perception, planning, action, and adaptation through iterative control loops. This paper examines this transition by connecting foundational intelligent agent theories, including reactive, deliberative, and Belief-Desire-Intention models, with contemporary LLM-centric approaches such as tool invocation, memory-augmented reasoning, and multi-agent coordination. The paper presents three primary contributions: (i) a reference architecture for production-grade LLM agents that separates cognitive reasoning from execution using typed tool interfaces; (ii) a taxonomy of multi-agent topologies, together with their associated failure modes and mitigation approaches; and (iii) an enterprise hardening checklist that incorporates governance, observability, and reproducibility considerations. Through an analysis of emerging industry platforms, including Kore.ai, Salesforce Agentforce, TrueFoundry, ZenML, and LangChain, the study identifies a convergence toward standardized agent loops, registries, and auditable control mechanisms. It is argued that the subsequent phase of agentic AI development will parallel the maturation of web services, relying on shared protocols, typed contracts, and layered governance structures to support scalable and composable autonomy. The persistent challenges related to verifiability, interoperability, and safe autonomy remain key areas for future research and practical deployment.

Figures (6)

• Figure 1: Visual Abstract: From Passive Models to Goal-Directed Systems.
• Figure 2: Comparing the Single-Turn LLM vs. The ReAct Loop vs. Multi-Agent Orchestration.
• Figure 3: Reference architecture for agentic AI systems. Cognition (LLM) is separated from control flow, memory, and tool execution; governance and observability cross-cut the stack. This separation reflects enterprise platform emphasis on orchestration, security, and tracing koreai_platform_2026truefoundry_agentic_2025zenml_platform_2025langchain_blog_2026.
• Figure 4: The Governed Reference Architecture.
• Figure 5: Taxonomy of multi-agent topologies.