Back to Foresight

Foresight Introduction

How Foresight is built:
the functions it runs on, the prompt design behind it, and how its output is evaluated.

What Foresight does
Prompt design

Foresight runs on 3 system prompts (6,300 words) and 3 user prompts (2,105 words).
Foresight's design separates system prompts from user prompts: the system prompts are reusable across corporate-finance cases, while the specific focus area for a given case goes in the user prompts. The design also incorporates three concepts:

Blind evaluation

To measure quality objectively, deliverables are compared under blind, source-neutral labels (Result A versus Result B) and run through three independent AI evaluators. Each scores both work products on Accuracy, Completeness, and Actionability.

AI models deployed
Foresight
Claude Opus 4.8 + High Thinking, with no memory from prior chats
AI evaluators
ChatGPT 5.5 + High Thinking
Grok Expert
Gemini 3.5 Flash + Extended Thinking