v0.9 · Early Access Project Management Framework

True north for Human + AI projects...

NorthStar is the fourth project management methodology, built specifically for human–AI collaborative work. It defines guardrails, documentation, and validation so AI agents can move fast without breaking your architecture, your process, or your trust.

For PMs, engineering leaders, and AI platform teams.

Human in the loop

Defined ownership

Clear roles for PMs, reviewers, and domain experts across every sprint.

Documentation

Structured, living

NorthStar files keep rules, queues, and decisions aligned with reality.

Drift & risk

Prevented, not chased

Gates and scorecards catch hallucinated docs and unapproved changes early.

Project · AI-refactored learning system

ZERO DRIFT - ACTIVE

Guardrail coverage NS

100%

Critical flows operating under NorthStar rules and gates.

Human validation

Consecutive sprints with completed human validation scorecards.

Architecture drift 0

Zero

No unvetted schema or API changes across the last release cycle.

NorthStar defines how work moves: from rules and task queues, through sprint gates, into human validation. Use it alone in offline environments or pair it with ZeroDrift Cloud for measurable AI drift analytics.

The problem

Traditional PM methodologies (predictive, agile, hybrid) were built for human teams. They assume shared context, organic communication, and disciplined documentation. AI agents don't automatically do any of those things—so drift, hallucinated docs, and misaligned teams become the default.

AI drift

Agents ignore architecture and processes, slowly diverging from the intended design and business goals, often without anyone noticing until late.

Hallucinated docs

Documentation that doesn't match reality—outdated, invented, or misaligned with what is actually in production—erodes trust and slows every release.

Misaligned teams

Human PMs, developers, and AI agents work at cross-purposes with no shared language, no clear gates, and no consistent way to accept or reject AI output.

The NorthStar solution

NorthStar is a dual-layer system that combines methodology with enforced governance. Humans define rules and validate output; AI agents execute inside clearly defined guardrails. The framework is designed to prevent problems instead of chasing them.

NorthStar file – Mission, guardrails, out-of-bounds rules, and system constraints live in one canonical artifact.
Parking lot – A structured queue of tasks and prompts that keeps agents focused on approved work only.
Sprint gate – Pre/post checks that block changes without required artifacts, tests, or approvals.
Human validation – A scorecard that captures acceptance, exceptions, and follow-up work for every iteration.

System architecture

The framework drops into your existing tool stack. NorthStar defines how work moves; your repos, CI, and task systems handle where it lives.

Inputs

Guardrails from your NorthStar file; approved work from your task queue; human acceptance criteria.

Process

AI agents execute work inside guardrails; sprint gates block unapproved changes; human reviewers validate output.

Output

Governed changes; validated documentation; auditable approval trails; zero hallucination.

All work is logged. Every gate pass or failure is a data point. Connect to ZeroDrift for real-time analytics.

The NorthStar starter kit

Everything you need to govern your first AI project with NorthStar. Templates, examples, and step-by-step onboarding.

Framework Templates

Pre-built NorthStar files and parking lot structures. Copy, customize, and start using within hours.

Gate & Validation Guides

Runbooks for implementing sprint gates and scorecard validation. Tested with real AI teams.

Integration Guides

How to wire NorthStar into GitHub, Linear, Jira, Claude, or any tool your team already uses.

Get early access

Join the NorthStar program. Get templates, live guidance, and early access to ZeroDrift integration.