How PBA Works
Product Behavior Architecture: the AI-native product operating system.
Product Behavior Architecture (PBA) decides how a product must behave across market, user experience, operations and revenue before AI tools touch the build. It turns those decisions into behavior cards that coding agents and engineers build from and verify against. Product, design and engineering leaders run on it.
Decide market, experience, operations and revenue before the build.
Where PBA Sits
Above the spec, at the business decisions.
AI strategy advisors give you a roadmap and stop before the build. Spec-driven development tools start at the specification and make engineering faster. PBA starts at the business decisions, writes the specification from them, and carries them through to what teams and agents build.
1
Business decisions
Market, experience, operations and revenue, decided explicitly before anything is built. This is Phase 0: the 4 P’s.
2
Behavior cards
Each decision becomes behaviors: what the product does, what it must never do, why, and how to check it.
3
Teams and agents build
Coding agents build from the cards. Engineers and QA verify against them. Nobody has to translate design into code.
Spec-driven development works. But do the specs maximize business value, or just engineering efficiency? PBA makes sure they serve the business.
Phase 0
It starts with four business decisions.
Before a single feature is named, PBA forces four decisions: market viability, user experience, operational feasibility and revenue model integrity. Every behavior the product ends up with traces back to them.
Spec-first tools
Documents requirements
Forces decisions
Generates outputs
Prioritizes what matters
Start at the spec
Start at the business decisions
Spec once
Keep decisions current through delivery
Describe what to build
Define what it must never do
M
Market viability
Can the product reach, win and keep the customers it needs?
U
User experience
What must it do, and never do, for the people who use it?
O
Operational feasibility
Can the business deliver, support and govern it at scale?
R
Revenue model integrity
Will the way it earns money survive the first contract?
How It Runs
Five stages. One unbroken chain of decisions.
Claude produces and challenges: it drafts and builds at volume, then red-teams the work in separate conversations. A person decides: every ruling is numbered, dated and approved by the product owner.
The sequence matters. Decisions made in Groundwork prevent failures that would otherwise surface in the build, in testing or with customers.
The Constructs
The mechanisms that make PBA executable.
Each one forces a decision, removes noise, or makes a rule checkable.
Prohibition Set
What the product must never do.
Most product methods say what to build. Few write down what the product must never do, in business, user and clinical terms. PBA does, and gives every prohibition a check that an agent and a tester can run.
Turns tacit judgment into rules agents follow.
Lane Test
Product decisions stay out of engineering territory.
Every item passes one question: would a coding agent work this out from the behavioral intent anyway? If yes, it stays out of the spec. WHAT and WHY belong to the business. HOW belongs to engineering.
Removes document bloat. Preserves engineering autonomy.
Decision Test and Impact Triage
Every step forces a decision, and only material ones survive.
Before a detail is added, it must force a new decision rather than elaborate an old one. Findings are scored HIGH, MEDIUM or LOW, and LOW is cut.
Keeps specs decision-dense and material.
Behavioral Contracts
How the product must behave, not how it is built.
The output agents and engineers execute against: testable behavior, written as behavior cards. Not a PRD and not a wireframe.
Removes interpretation. Makes every rule verifiable.
Who It’s For
One voice, three directions.
The same operating system answers a different question at each level of the company.
For CEOs and CPOs
AI’s speed exposes how slowly decisions get made.
Execution got fast. Work now waits on the business, or moves ahead on assumptions. The bottleneck moved from building to deciding.
For Boards and Investors
Leaders are asked for AI results. Most organizations can show AI activity.
AI-native companies replace old business models rather than improve them. Oversight belongs on the intent AI systems are given, and on measuring what they deliver.
For Product and Engineering Leads
Spec-driven development works. Do your specs serve the business, or just engineering?
Context engineering starts too late if it isn’t written to maximize business value. Agents get most of the way, then stall on behavior nobody specified.
Not a faster workflow. An operating system.
A roadmap and a list of use cases leave the build to someone else. PBA writes the decisions into the behavior cards agents build from.
Most methods start with the product: personas, features, sprints. PBA starts with whether the business can reach, serve and earn from its market, then writes the spec from those answers.
Agents read stale documents as current truth. PBA separates what is decided from what is still open, resolves conflicts across the corpus, and keeps the cards current as rulings change.
Book a Discovery Consult
Free. No commitment.