Strategy & Outcomes
Define the institutional problem, desired outcome, decision rights, constraints and evidence of success.
One connected architecture
Full-Stack Intelligence, From Infrastructure to Real-World Use
AFA connects the layers that are often separated across vendors and departments: strategy, governance, data, compute, models, knowledge, copilots and agents, Marbles, Physical AI, people, creative experience and continuous improvement.
The canonical AFA stack
Nine layers from institutional purpose to continuous improvement.
Not every organization needs every layer at the same time. The model exists so individual projects do not accidentally create dependencies or risks elsewhere in the system.
Data, Knowledge & Governance
Classify information, protect IP, establish authoritative knowledge, permissions, policies, records and human accountability.
Sovereign Compute & Infrastructure
Choose cloud, hybrid, private or on-premises compute, networking and storage based on workload, sensitivity, scale and resilience.
Models & Intelligence
Select, evaluate, compare, ground, adapt and govern models according to actual organizational requirements.
Copilots, Agents & Marbles
Turn intelligence into useful digital work through Microsoft Copilot, agents, workflows and persistent organizational understanding.
Physical AI & Robotics
Connect intelligence to perception, simulation, edge systems, machines and real-world operations under bounded authority.
People, Training & Adoption
Build the human judgment, skills, role clarity and change capability required for responsible use.
Studio, Simulation & Experience
Make complex systems understandable through film, visual design, immersive environments, training media and creative proof.
Measurement, Support & Continuous Improvement
Evaluate outcomes, monitor model/system behaviour, preserve learning, support users and improve the environment over time.
Sovereignty runs vertically through every layer
Control is not a single infrastructure box.
Data, compute, model, knowledge, cognitive and operational sovereignty cut across the stack. Different workloads require different combinations of control.
Data Sovereignty
Who controls organizational information, where it resides and where it is permitted to travel.
Compute Sovereignty
Where intelligence is processed, who controls the infrastructure and what external dependencies exist.
Model Sovereignty
Which models are used, how they are evaluated and whether the organization can replace or adapt them.
Knowledge Sovereignty
Which sources, laws, policies, standards and institutional records the AI should treat as authoritative.
Cognitive Sovereignty
Whether the AI interprets information through an appropriate institutional, jurisdictional, linguistic and cultural frame.
Operational Sovereignty
What AI, agents, machines and robots are authorized to do—and when people must remain in control.
How AFA enters the stack
Assess → Architect → Deploy → Integrate → Align → Evaluate → Secure → Govern → Support
AFA can begin with the actual decision in front of the organization and expand only where evidence justifies it.
Assess
Understand the work, people, information and constraints.
Architect
Design the environment and governance before scale.
Demonstrate
Use pilots, labs or prototypes to test assumptions.
Deploy
Integrate the chosen capability into real workflows.
Improve
Measure, learn, support and evolve the system.
Different doors into one company
Start where the value or responsibility is clearest.
Institutional AI
Strategy, governance, sovereign architecture, Copilot, agents and knowledge systems.
Solutions →Physical Intelligence
Robotics, perception, simulation and the governance of real-world action.
Physical AI →Human & Creative Intelligence
Academy, Studio, training, storytelling and immersive experience.
Academy →Continue the journey
Related pathways
You do not need to buy the whole stack.
Start with the layer where a real institutional decision exists.
Start with a Workshop