ARTIFICIAL INTELLIGENCE MAY SOON EMULATE PICASSO AND TOLKIEN - IS PORTER NEXT? AN OPTIMIZATION OF ORGANIZATION STRATEGIC DECISION-MAKING WITH AI TECHNIQUES

Abstract

This dissertation explores improving strategic decision-making, an intractable and often abstract type of optimization problem, via the application of artificial intelligence (AI) techniques. The latest AI techniques based on deep learning (DL) and artificial neural networks (ANNs) have powerful capabilities including generating stories ad Tolkien, art ad Picasso, and hold great promise against more complicated problem classes. However, it is still unclear as to how and how well, these techniques can help improve strategic decision-making. I propose several novel management models and approaches to deploy AI techniques as tools against the problem of organization strategy, account-for and mitigate trade-offs, and achieve more optimal outcomes. A combination of inductive, deductive and abductive reasoning approaches were employed to advance two sequential studies, with the results from the first informing the second. Study one is a review of nearly 500 pieces of peer-reviewed literature, books, databases, empirical case studies and other sources from 1945 to 2025. A set of seven largely uncaptured trade-off dimensions exist, with relevance in both an organization strategy and computational context: accuracy, explainability, fairness, privacy, reliability, security, and speed; and, twenty-one pairwise trade-offs between them. These trade-offs persist despite the advent of several AI techniques examined herein, including Large Language Models (LLMs), Generative Adversarial Networks (GANs), Retrieval-Augmented Generation (RAG), Mixture of Experts (MoE) and GFlowNetworks (GFNs). AI techniques such as Language Models (LMs) and their requisite training also have not addressed important trade-offs such as accuracy vs. speed. Contributions to the literature and theory, include: 1) a comprehensive examination of AI techniques within broad organization strategy contexts spanning industries and functional areas, 2) explication and extensive exploration of trade-off dimensions and pairwise trade-offs between them that have relevance within both organization strategy and computational optimizations, 3) finding that these trade-off dimensions persist despite advancements in and performance of AI techniques, and 4) finding that different AI techniques are better suited for different problems/goals. Contributions to the practice, include: 1) a categorization of trade-offs by industry and functional area, a common way that strategists and consultants delineate their expertise and problem domains, and 2) a multi-level mapping that organizations can use to choose AI technique(s) for their specific problem/activity. Study two takes the above, and developed and introduced several management models and approaches for improving strategic decision-making and outcomes through the deployment of AI techniques in these organization processes. Simulations were used to assess AI technique performance against different trade-offs, yielding the following results: RAG was the most effective technique, while LLMs had the widest applicability. Combinations of AI techniques, such as LLM+RAG, outperform individual techniques with respect to both colloquial attributes, e.g. parameters and cost, and the set of seven trade-off dimensions. MoE-based algorithmic architectures exhibit gains in the broadest set of desired outcomes while mitigating unfavourable trade-off dimensions. Contributions to the literature and theory, include: 1) a unifying multi-step approach to maximize fitness of a production function using the NK Model and strategic fit with computational resources, and 2) proposition that AI Technique is the most accurate way to refer to anything AI. Contributions to the practice, include: 1) a conversion from colloquial attributes such as number of parameters, to the set of seven trade-offs dimensions, and 2) a hybrid process of humans and AI techniques, that organizations can utilize to maximize outcomes while minimizing unintended trade-offs, modelled to improve the accuracy vs. speed trade-off. This research study has the potential to enhance organization strategic decision-making and outcomes, via computational approaches such as AI techniques. Improving management models and processes to address issues that have thus far contributed to suboptimal outcomes, may lead to increased economic value to organizations, their stakeholders and customers, the economy, and society at large.