The Human-First AI Framework
Version 1.0 - Published June 10, 2025
Author: Paweł Jarmołkowicz
Full framework document: LINK
Part 1: The Manifesto
Our Guiding Principles for Human-Centric AI
This is our constitution. It is the fundamental, ethical foundation upon which our methodology, our designs, and our daily choices are built.
AI should make you more capable, not more dependent.
AI should strengthen your thinking, not replace it.
AI should create value by amplifying human potential.
AI should enhance human connection, not replace it.
Part 2: The 5 Principles of Practice
Practical Rules for Designing Human-First Systems
These five principles translate our Oath into actionable design rules. They are the tests we apply to our systems and processes to ensure they remain true to our core values.
1. Design for Deliberate Choice: We must design systems that preserve and enhance human agency, giving the user meaningful control over the collaboration.
2. Design for Complementary Partnership: We must architect our systems based on the unique strengths of each intelligence. The AI is the "Pattern Engine"; the human is the "Meaning Engine."
3. Measure What Matters: Insight over Efficiency: Our highest priority is the generation of transformative insights. We will judge success by the quality of thinking enabled, not just the quantity of work produced.
4. Design for Cognitive Ergonomics: We must design human-AI interactions to be seamless and intuitive, minimizing the extraneous cognitive load required to use the system.
5. Design Through Participation & Experimentation: The most effective systems are co-created with the people who will use them every day, built on a foundation of safe-to-fail experimentation.
Part 3: The Implementation Methodology
An 8-Week Sprint to Deliver Tangible Value
This is a lean, 8-week sprint designed to deliver a tangible win and build internal capability.
Phase 1: PINPOINT (Weeks 1-2)
Objective: To find the single greatest point of leverage for our pilot. We begin with a rigorous diagnosis, using an assessment toolkit to analyze the organization's readiness and audit a key workflow to distinguish between Craft (tasks ideal for AI) and Wisdom (tasks reserved for humans). This allows us to identify the perfect pilot project.
Phase 2: PILOT (Weeks 3-6)
Objective: To co-design and validate a new, high-performance Human-AI workflow. A small "Pathfinder Team" experiments with a new workflow, designing a seamless partnership between the human and AI. They test it on a real work product, focusing on learning and iteration to create a system that demonstrably produces higher-quality strategic insights.
Phase 3: PROVE (Weeks 7-8)
Objective: To prove the value to the wider team and create a pull for adoption. The Pathfinder Team presents their success story to the organization, sharing their results and a simple "MVP Playbook" so others can replicate their success. This creates genuine excitement and a data-driven mandate for a wider rollout.
Part 4: Governance & Trust
Our Commitments to Responsible and Ethical AI
Executing this methodology requires an absolute commitment to responsible innovation. The following commitments are non-negotiable:
Uphold Client Trust Above All: We will be transparent with clients and protect their data with stringent governance.
Ensure Data Privacy & Security: We will comply with the highest standards of data protection, including GDPR and HIPAA.
Mitigate Bias & Promote Fairness: We will actively audit AI outputs and design processes that promote equitable outcomes.
Maintain Human Accountability: The ultimate accountability for any decision or deliverable always rests with the human practitioner.
Read the Full Framework
This has been a high-level summary of the framework. The full document contains the detailed methodology, the complete Assessment Toolkit (including the Readiness Assessment, Intelligence Audit, Maturity Scale, and Transformation Map), and the theoretical grounding that underpins this work.
It is a living document, and I invite you to read it, critique it, and use it.
— Paweł Jarmołkowicz