Socratic Intent Engineering

Adapted from The Socratic Intent Engineer: Mastery of Intent and the Human Edge by Martin Montero — the field manual this site is the companion system for. PIE is how you cook; this is how you decide what’s worth cooking, and keep the agent aimed at it.

The mistake that changes the question

In 2024 a fintech announced its AI support agent was doing the work of ~700 people — 2.3M conversations, resolution time cut from eleven minutes to under two. A year later the CEO admitted they’d cut too deep and began rehiring humans. The AI hadn’t failed. It succeeded at the wrong thing. “Resolve tickets fast” was a measurable proxy; the real goal was lasting customer relationships.

That gap — between what an AI can do and what you actually need it to do — is where the real work now lives. When capability is abundant, the scarce resource isn’t compute. It’s wisdom: clear intent, and the judgment to know what’s worth building at all.

Why this pairs with PIE

The PIE build flow gets you from intent to a shipped artifact — the doing. Socratic Intent Engineering is the thinking that runs first and alongside it, so what you ship is aimed at the right thing. Start most projects here, then cook. The Build Studio walks you through both.

The four foundations

The Socratic method

Ask, examine the assumption, probe for contradiction, refine — before you write a single instruction. Nearly every useless deliverable traces to a question that was never asked.

First-principles thinking

Decompose to fundamental truths instead of copying what others do. Don’t start with “I need a report” — start with “three people must understand three risks in five minutes.” The format follows.

Intent engineering

Encode what an agent should want, not just what it should know — goals, trade-off hierarchies, decision boundaries — so it optimizes for the real objective. This is what the Studio bakes into your project constitution.

The human edge

Taste, judgment, and the courage to bet on a direction while the map is blank. Output is cheap; choosing is expensive. The AI can’t decide what matters — that’s your job, and the valuable one.

Three eras of AI engineering

Era	What it does	Enough?
Prompt engineering	crafts a single instruction	personal, session-bound
Context engineering	tells an agent what to know	necessary, not sufficient
Intent engineering	tells an agent what to want	the layer that makes AI strategically coherent

The frontier models are all extraordinarily capable; the gap between them matters far less than the gap between builders who give an agent clear, structured, goal-aligned intent and those who don’t. The race is an intent race.

→ BUILD: where the method becomes an artifact

The Ten Questions → The integrated method + the complete 8-step workflow — the practical core.

Build Studio → The Ten Questions, operationalized: your answers become a constitution, spec, and agent prompt.

Kitchen Prep (PIE) → Name the problem before reaching for a tool — the PIE stage this method sharpens.

Agent skills → Encode judgment, escalation, and intent into reusable skills — not just instructions.

The human stays in the driver's seat

Use AI for everything that can be systematized, so you can spend yourself on everything that can’t — your perspective, your values, your read on what your community actually needs. Don’t be a cover band. Write your own music; let the AI play the instruments.