Large Language Models have transformed software development, knowledge work, and human-computer interaction by allowing people to express intent in natural language. But the claim that AI will replace enterprise software rests on a category error.
LLMs are probabilistic systems optimized for plausible language generation. Enterprise software must provide reliable execution, reproducibility, governance, auditability, security, and mathematically trustworthy computation.
Three original contributions
Three Generations of Computing. The paper proposes a progression from Deterministic Computing to Probabilistic Computing and, now, Intent-Driven Computing (IDC): a paradigm in which people express business intent naturally and technology translates it into governed execution.
The AI Competence Mirage and the Computational Trust Gap. Fluent language creates a powerful perception of competence. The paper names the gap between that perceived confidence and the computational guarantees actually available behind an answer.
A three-layer enterprise architecture. The proposed model separates the responsibilities that modern AI systems too often blur:
PIL — Probabilistic Interface Layer
Interprets human intent expressed in natural language.
DIL — Domain Intelligence Layer
Encapsulates domain knowledge, business methodology, policies, and constraints.
DEL — Deterministic Execution Layer
Performs trusted computation and governed execution.
The central thesis
Large Language Models do not represent the end of deterministic software. They represent the emergence of a new computational paradigm in which probabilistic intelligence orchestrates deterministic execution.
This changes the economics of software without removing the need for software itself. AI lowers the cost and time required to build systems, while increasing the strategic importance of domain intelligence, architecture, governance, and operational trust.
Why this matters for enterprise leaders
The future is not software versus AI. Natural language will become a universal interface; domain intelligence will determine how problems should be solved; and deterministic systems will remain responsible for the outcomes organizations must reproduce, audit, secure, and defend.
For executives, investors, entrepreneurs, developers, and data scientists, the practical question is therefore not whether an AI can produce a persuasive answer. It is whether the complete system can transform intent into a result that the organization can trust.
Resumen en español
Los Modelos de Lenguaje Grandes han transformado la forma en que interactuamos con la tecnología, pero su capacidad para generar lenguaje plausible no equivale a una garantía de ejecución confiable.
El paper propone separar tres responsabilidades: una capa probabilística que interpreta la intención, una capa de inteligencia de dominio que determina cómo resolver el problema y una capa determinística que ejecuta reglas, algoritmos y cálculos de forma reproducible y auditable.
La conclusión es simple: la IA no elimina el software empresarial. Reduce radicalmente el costo de producirlo y desplaza la ventaja competitiva hacia el conocimiento del dominio, la arquitectura, la gobernanza y la confianza operacional.