Decision Spine

About Decision Spine

Decision Spine is an independent data and AI architecture practice focused on AI Analytics Reliability - making AI-generated business answers reliable enough to use in real decision-making. AI analysts can generate valid SQL and plausible numbers while still answering the wrong business question. Decision Spine works on the systems underneath them: semantic definitions, metric governance, analytical models, evaluation frameworks and runtime guardrails. The work combines applied research, open-source tooling and hands-on architecture to help data teams benchmark AI analytics, locate silent failure modes, repair the underlying data and semantic foundations, and retest the system. Current work includes the open-source AI Analytics Harness, a reproducible lab for studying grounding, reliability, protocol, repair and ambiguity, and Preflight, a static-analysis tool for detecting ambiguous analytics definitions before an AI agent or human selects the wrong one. Focus areas: AI Analytics Reliability · Data & AI Architecture · Semantic Layers · dbt · Metrics · Evaluation Systems · Guardrails · Agentic Analytics · Data Quality