Flagship assessment method
Sovereign Test and Evaluation AI Assessment Method
A structured approach to answering a critical question: will this AI capability improve the mission without creating unacceptable operational, human or sovereign risk?
STEA²M begins with the operational task, intended outcome, user responsibilities and deployment environment. It then examines the evidence required to judge whether AI is appropriate, effective and controllable in that context.
Evaluate the effect of AI on human work, situation awareness, decision quality, supervision and net cognitive workload. Count verification effort as work, not as a free benefit.
Key questions
Evaluate capability under credible operational constraints, including degraded communications, limited compute or power, incomplete data, faults and loss of external dependencies.
Key questions
Evaluate control over data, models, dependencies, interfaces, governance and the ability to operate within existing systems and coalition or multi-agency environments.
Key questions
The pillars are equal and gated: a strong result in one does not automatically compensate for a critical failure in another.
Define the mission and operator task
Establish constraints and critical risks
Agree evaluation criteria and evidence
Test the three pillars
Identify limitations, mitigations and residual risk
Present decision-ready findings
Illustrative pathway. Assessment scope and evidence requirements are tailored to the use case.
A capability that performs well in ideal conditions may still be unsuitable if it increases operator burden, fails under realistic constraints or introduces dependencies that cannot be controlled. STEA²M makes these tensions visible rather than concealing them behind a single headline metric.
Specific deliverables depend on the agreed scope. STEA²M is an assessment method; it does not issue accredited certifications.