← All posts
5 min read

Intent-Driven Development: How I Ship Products Without Requirements Documents

The methodology that's replaced PRDs, roadmap decks, and sprint planning for me. Start with outcomes, not features.

The Traditional Trap

I've written dozens of product requirements documents. Multi-page decks with user stories, acceptance criteria, technical specs, timeline estimates. By the time they're approved, half the assumptions are outdated. The market shifted. A competitor launched. The customer learned something new.

Meanwhile, the engineering team is building exactly what was spec'd — because that's what the PRD said. Six months later you ship a product that solved February's problem in August's market.

Sound familiar?

The Shift to Intent

I started using something different a few months ago. I call it Intent-Driven Development (IDD). It's not agile, it's not lean — it's a fundamental inversion of how products get defined and built.

Traditional approach: Define features → Build features → Hope they create outcomes.
Intent-driven approach: Define outcomes → Validate the path → Build only what serves the outcomes.

The difference sounds subtle. In practice, it's transformative.

Core Principle

Intent is the product's source code. The customer's desired outcomes and constraints define what gets built. Humans review the Intent. AI helps execute the implementation.

How IntentWin Got Built

IntentWin — the AI proposal generator I launched last week — was built using IDD from day one. Here's what that looked like:

Phase 1: The Outcome Contract
Before writing any code, I defined the customer's transformation. Government contractors need to move from spending 40-80 hours per proposal (current state) to generating compliance-ready drafts in under an hour (desired state). The product's entire purpose is bridging that gap.

Everything else — features, architecture, integrations — is downstream of that outcome.

Phase 2: The Win Strategy
Not "product strategy" in the MBA sense. This is the specific way this product wins against alternatives: manual proposal writing, hiring consultants, or using generic AI tools that don't understand government contracting.

Our win themes: verified claims (no hallucinations), compliance-first (no disqualifications), and speed through parallel generation (not just faster typing).

Phase 3: Human-Locked Intent
The Intent document — outcomes, strategy, constraints — gets locked before implementation starts. This is the contract. If market conditions change, we update the Intent first, then adjust implementation. Never the other way around.

What IDD Replaces

PRDs become Outcome Contracts. Instead of "the system shall allow users to upload documents," it's "the user transforms scattered RFP materials into structured intake in under 5 minutes." The outcome defines success. The implementation is flexible.

Roadmaps become Intent milestones. Not feature releases on a Gantt chart. Validation gates: Has the outcome contract been verified with real users? Does the knowledge system correctly enforce claim verification? Is the quality overseer catching errors before humans see them?

Sprint planning becomes phasing. Each phase has a testable hypothesis. Phase 1: Can we extract requirements from a PDF? Phase 2: Can we generate a compliant executive summary? Phase 3: Can we do full multi-section generation with quality review? If a phase fails validation, we don't proceed. The Intent stays locked until we figure out why.

The Marketing Angle

Here's why this matters for GTM strategy: Intent-driven products are easier to market because the marketing is baked into the definition.

When your product is defined by customer outcomes, your messaging writes itself. You don't need a positioning exercise to figure out what to say. The outcome contract is the positioning. "We help government contractors cut proposal time by 90% while increasing win rates" isn't marketing fluff. It's the literal product definition.

Your sales deck becomes an explanation of the outcome journey. Current state pain, desired state gain, the transformation bridge your product provides. Every feature you demo maps directly to an outcome. No "nice to have" functionality that sounded good in a roadmap session but doesn't serve the core promise.

The AI Multiplier

IDD works without AI. But AI makes it orders of magnitude faster.

With IntentWin, the AI doesn't just generate code. It generates based on a structured Intent that encodes outcomes, constraints, and quality criteria. The three-layer knowledge system (company truth, proposal intent, generated content) is the architecture that makes this possible.

I can define a new product outcome on Monday, validate the approach with real users by Wednesday, and have working software by Friday. Not an MVP that needs six months of iteration. A validated implementation that serves the defined outcomes.

The Pattern

Define outcomes → Lock Intent → AI-assisted implementation → Validate against outcomes → Ship. Repeat.

What You Can Steal From This

Start every project with an Outcome Contract. Write down the customer's current state, desired state, and how you'll bridge the gap. Be specific. Numbers help. If you can't define the transformation, you don't have a product yet — you have an idea.

Lock your Intent before building. Create a document that defines what success looks like without prescribing implementation. Share it with stakeholders. Get alignment. This becomes your north star when implementation decisions get messy.

Phase by validation, not by feature. Each phase should test a hypothesis about whether you can deliver the outcome. If you can't generate a compliant proposal section, there's no point building the export layer. Validate the core first.

Use AI for execution, not exploration. The AI works best when the Intent is clear. Don't ask Claude to "build a proposal tool." Give it an outcome contract, constraints, quality criteria, and let it orchestrate the implementation. The clarity is your job. The execution is AI's.


I've shipped more working product in the last three months using IDD than in the previous year using traditional methods. The products are more focused, easier to explain, and actually solve the problems they set out to solve.

The methodology is simple: outcomes first, Intent locked, AI-accelerated execution. Everything else is implementation detail.

Want the full AI marketing playbook?

The tools, workflows, and frameworks I use every day -- in one practical guide.

Read the AI for Marketers Guide →