AI Specifics
The deepest evaluation domain. Model lineage, grounding, drift, fallback, confidence scoring, self-improvement, and self-healing. The seven AI-specific points that close the gap between proposal claims and operational reality. GAO-26-107859 identified these exact concerns as agencies' top AI acquisition risks.
7 scored points inside this domain.
Each point sits on the EFAIRA 30-point scorecard. Each point has a defined evidence baseline and a color-coded readiness mark on the Encore Readiness Receipt.
Model lineage and data provenance
Origin of the underlying models. Training data composition. License compliance for federal use. Supply chain transparency end to end. GAO-26-107859 cites vendor training data quality as a top federal AI acquisition concern.
Hallucination posture and grounding
How the system handles cases where it does not know the answer. Whether outputs are grounded, sourced, and verifiable against authority.
Drift detection and human-in-the-loop
Whether the system detects its own degradation over time. Where human-in-the-loop checkpoints exist and how they trigger. GAO-26-107859 cites model performance degradation as a top federal AI acquisition risk.
Fallback and failure-mode design
What happens when the AI is wrong, slow, or unavailable. Whether mission-critical workflows degrade gracefully or stop.
Confidence scoring and uncertainty quantification
Does every AI output carry a confidence score or uncertainty interval? Can evaluators tell when the system is sure versus guessing? Confidence transparency is the single largest gap between consumer AI and federal-grade AI.
Self-improvement and documented update plans
Does the system get better with use, or does it plateau? Is there a documented model refresh cadence, capability evolution roadmap, and patch discipline? Federal sustainment depends on knowing how the system evolves over the contract period.
Self-healing and resilience design
Does the system detect its own degradation and route around it automatically, or does failure cascade to operations before a human catches it? Production-grade resilience separates federal-ready systems from demos that survive only happy paths.
Score Domain III on your federal pursuit.
Encore engages with contractors and federal agencies through separate paths. Pick the right one for your role.