Super-prompts

Introduction

The book AI-Augmented Decisions: A Practical Guide provides a clear and powerful real-world application of the super-prompting methodology. It moves the concept from theory to practice by building its entire decision-making framework around a series of five distinct, process-specific super-prompts.

This approach demonstrates how to make a complex, high-stakes domain of knowledge (high-complexity, high-impact decision-making) both AI-legible and AI-actionable.

A Five-Phase Process Driven by Five Super-prompts

The book introduces a robust, five-phase process for navigating high-impact, high-complexity decisions. Each phase is guided by its own unique super-prompt, which configures the AI to act as an expert facilitator for that specific stage of the journey.

The AI, armed with the super-prompt, guides the user through the structured activities of each phase, ensuring a thorough, repeatable, and high-quality process.

Phase 1: Decision Scoping

• Super-prompt’s role: Configures the AI to help the user define the decision, identify the work to be done, and specify acceptance criteria.
• Output: A clear and concise Decision Brief.

Phase 2: Decision Preparation

• Super-prompt’s role: Guides the user through a divergent exploration of potential ‘decision candidates’ and ‘decision differentiators’ before converging on a refined set of options and the evidence needed to evaluate them.
• Output: A comprehensive Record of Decision Preparation.

Phase 3: Decision-Making Workshop

• Super-prompt’s role: Helps design and facilitate a structured workshop to evaluate the prepared options, manage deliberations, and arrive at a consensus.
• Output: A formal Decision Proposal.

Phase 4: Decision Validation

• Super-prompt’s role: Acts as a critical partner to stress-test the proposed decision, checking it against the original brief (Scope Validation) and the supporting evidence (Evidence Validation).
• Output: The approved Final Decision.

Phase 5: Decision Adoption

• Super-prompt’s role: Assists in translating the final decision into a human-centric, agile plan for action, focusing on stakeholder engagement and governance.
• Output: A pack of Decision Adoption Resources.

Case Study in Action: “Choosing the Holiday of a Lifetime”

The book illustrates this entire process with a tangible, end-to-end case study: “choosing the holiday of a lifetime”. This example powerfully demonstrates how the super-prompts work in practice.
For instance, during the Decision Preparation phase, the AI, guided by its super-prompt, helps the user brainstorm eleven potential destinations. It then assists in identifying eight key “differentiators” (like ‘Wow-Factor of Wildlife’ and ‘Level of Adventure’). Finally, it helps narrow the list to six finalists for a “South America versus Africa comparison.”

In the Decision Validation phase, the AI raises a critical challenge, questioning the human decision-makers’ choice to “neutralise” the scores for health and safety risks between two finalists. This AI-driven challenge doesn’t overturn the decision; instead, it forces the humans to clarify their underlying values, ultimately strengthening the rationale for their final choice.

This case study proves the power of the super-prompting model. The AI is not making the decision. It is acting as a tireless, structured, and insightful partner that facilitates a more robust, transparent, and effective human-led decision-making process. It is the ultimate demonstration of AI-augmentation in action.

Addressing the Practical Challenges

The design of the AI-Augmented Decisions process directly addresses the key technical hurdles of super-prompting:

1. Solving the ‘Lost in the Middle‘ Problem: The process cleverly avoids a single, massive prompt where instructions could be lost. Instead, it breaks the whole process into five distinct phases. Each phase begins a new conversation with its own super-prompt, ensuring the critical instructions for that stage are always placed at the very beginning of the context window, where the AI’s attention is strongest. This modular, phased approach is a powerful architectural solution to the ‘Lost in the Middle’ problem.
2. Embracing the ‘Comprehension-Generation Asymmetry’: Asymmetry arises because AIs are currently better at understanding long, complex instructions than they are at generating long, complex outputs. The AI-augmented decisions process is built around the principle of augmentation, not automation. It leans into the AI’s strength (comprehending complex instructions) while compensating for its weakness (generating equally complex outputs). The AI’s role is to guide the human through a structured process and generate intermediate, focused outputs like summaries, lists and drafts. The human user is always responsible for the final synthesis, critical judgment and authoring of the key deliverables, such as the final Decision Proposal. This division of labour perfectly aligns with the known asymmetry of LLM capabilities.

You may also be interested in…