Context Distillation
New signals can change the project state.
Context Distillation checks their relevance, proposes atomic deltas, and incorporates only human-confirmed deltas into Context.
How Context Distillation worksContext Driven Project Management (CDPM)
So AI can understandthe whole project.
CDPM complements existing project methods with living project context.
This enables AI to analyze connections, evaluate next steps, and provide Guidance—while humans remain accountable.
Two functions. One loop.
Context Distillation keeps project context current.AI analyzes connections and provides guidance.
Distillation maintains the foundation. From that foundation, AI evaluates the project as a whole and proposes reasoned next steps.
Function 01 · Maintain context
Function 02 · Analysis & Guidance
With isolated content
AI processes individual inputs
- AI summarizes emails and documents
- AI drafts minutes or status reports
- AI extracts action items from an individual signal
With current project Context
AI analyzes the project as a whole
- AI evaluates next steps by goal contribution and urgency
- AI identifies bottlenecks and critical dependencies
- AI assesses scenarios and forecast ranges in the confirmed project state
Four building blocks, one loop
How it works:the CDPM loop
Confirmed Context is the foundation for Analysis and Guidance. New Signals trigger distillation; only human-confirmed deltas update Context.
Context → Analysis → Guidance → Signals → Context
Continuously maintained Context connects goals, decisions, risks, and dependencies as a shared foundation.
- Goals & Success Criteria
- Timeline & Milestones
- Budget & Resources
- Team & Stakeholders
- Dependencies & Risks
- Action Items
From current, confirmed Context, AI evaluates project health, bottlenecks, critical dependencies, and forecast ranges.
- Plan vs. Actual Comparison
- Identify Context Gaps
- Trend Analysis
- Resource Utilization
- Bottleneck Analysis
- Critical Dependencies
Based on Analysis, AI proposes goal-aligned options that move the project forward. The accountable human decides.
- Context-based Action Item Creation
- Goal-aligned Prioritization
- Action Item Re-evaluation
- Trade-off Decisions
- Resource Distribution Optimization
- Context Adjustments (Timeline, Budget)
New emails, documents, messages, and conversation notes can change the project state. Context Distillation reviews them; only confirmed deltas then update Context.
- 1Signals from multiple sources
- 2Relevance check & atomic deltas
- 3Human review and decision
- 4Traceable context update
Context Distillation
From signal toconfirmed project context
AI checks relevance and proposes atomic deltas. A human confirms what changes Context. This keeps the foundation for Analysis and Guidance current.
01
Signal
Email, document, message, note
02
Relevance check
Compare signal with context
AI as catalyst
03
Delta proposal
Atomic, sourced change proposals
04
Human decision
Confirm, defer, or reject
Human in the loop
05
Updated context
Confirmed deltas only
01
Atomic Distillation
Complex signals are split into individually reviewable change proposals with sources and assumptions.
02
Selective Confirmation
The accountable human can confirm, reject, or defer individual deltas for clarification.
03
Audit Trail
A versioned history records who confirmed which delta, when, and for what reason.
AI proposes.The accountable human decides.Only then does the project context change.
Time · Budget · Scope
Trade-offs become visiblebefore anyone decides.
A living context enables AI to assess next steps and possible effects across time, budget, and scope. Assumptions and uncertainty remain visible before a human decides.
- Time
- 6 months
- ±0
- Budget
- €500k
- ±0
- Scope
- 100%
- ±0
BaselineSelected scenario
What-if Scenarios
Original plan with balanced time, budget, and scope targets.
Illustrative figures. Assumptions and dependencies are confirmed before any decision.
Contribute
The whitepaper followsits own principle.
The whitepaper is a rolling SSOT: contributions are collected, evaluated, and distilled directly in the text.
- 01
Contribute live in the reader
Click directly on a passage and submit your proposed change plus rationale in contribute mode.
- 02
Automatic validation
Filters and AI triage check quality and relevance. Only valid suggestions become visible.
- 03
Community prioritization
Suggestions stay anchored in context and are evaluated per location via up/down votes.
- 04
Distillation by the Project Owner
Meaningful suggestions are accepted, integrated into the markdown SSOT, and then published.
Collaboration happens live in the reader. Add your improvement exactly where it belongs.
Documentation
The framework,openly documented.
The whitepaper, toolkit, and glossary are developed in the open in the CDPM repository and licensed under CC BY 4.0.
WHITEPAPER
CDPM Meta-Framework Specification
The complete framework description, including the four building blocks, the role model, and implementation principles.
Open ReaderTOOLKIT
CDPM Implementation Toolkit
Practical templates, checklists, and guidelines for implementing CDPM in your organization.
Download PDFGLOSSARY
CDPM Terms & Concepts
Comprehensive glossary defining all CDPM-specific terms, roles, and concepts.
Download PDF
Context First
Context first.Artifacts follow.
Read the whitepaper or experience CDPM in practice with TensorPM, the AI project agent built on CDPM.