This glossary defines the core vocabulary of decision intelligence — from AI, BI and DI to decision models, context health, autonomy envelopes, digital twins, physical AI and decision observability — in plain English.
Terms A–Z
A
Abstention
A system’s deliberate refusal to recommend or act when a situation is outside its known domain, context is missing, or evidence conflicts. Learn more →
Agentic AI
AI systems that pursue multi-step goals by planning and taking actions through tools and APIs. In decision intelligence, each agent capability is treated as a decision type with its own authority, budget and audit trail. Learn more →
AI (artificial intelligence)
Software that performs tasks associated with human intelligence, such as prediction, classification, language understanding and generation. In decision intelligence, AI outputs are inputs to a decision, not the decision itself. Learn more →
Automation bias
The tendency of people to accept automated recommendations without enough scrutiny. Learn more →
Autonomous mobile robot (AMR)
A robot that navigates facilities without fixed guides, commonly used for material movement in plants and warehouses. Learn more →
Autonomy envelope
The explicit conditions — speed, zone, value, confidence, reversibility — within which a robot, fleet or system may decide on its own; anything outside escalates to a human. Learn more →
B
BI (business intelligence)
Reporting and dashboards that show what happened and where, from historical data. BI informs a person; it does not make or execute the decision. Learn more →
Business intelligence (BI)
Tools and practices for reporting and analyzing what happened, typically through dashboards, reports and KPIs. Learn more →
C
Confidence calibration
The agreement between a system’s stated confidence and its actual success rate over many decisions. Learn more →
Context health
A score showing how complete and fresh the inputs required by a decision are. Learn more →
Customer intelligence (CI)
Analysis that explains who customers are and what they are likely to do — segments, propensity and churn scores, 360° profiles. Learn more →
CX intelligence
Customer-experience intelligence: analysis of journeys, interactions and sentiment that shows how the experience feels and which fixes matter most. Learn more →
D
Decision context
The minimum sufficient data, rules, history, constraints and human knowledge a specific decision needs. Learn more →
Decision execution
The controlled write-back of an approved decision into operational systems with confirmation, rollback and audit. Learn more →
Decision framework
The five questions every decision must answer — Signal, Context, Options, Evidence, Decide — followed by Act and Learn to execute and improve. Learn more →
Decision intelligence (DI)
The discipline of designing, modeling, executing and continuously improving decisions by combining data, analytics, AI, business rules and human judgment, and connecting each decision to its outcome. Learn more →
Decision intelligence platform (DIP)
Software that models decisions explicitly and recommends, automates, executes, governs and audits them by composing data, analytics, knowledge and AI. Learn more →
Decision latency
The time from a triggering signal to an executed action. Learn more →
Decision log
A record of each decision’s inputs, options, evidence, confidence, approver, action and outcome. Learn more →
Decision model
An explicit representation of a decision’s trigger, inputs, options, constraints, objective, authority and outcome measure. Learn more →
Decision Model and Notation (DMN)
An Object Management Group standard for modeling decision requirements and decision logic. Learn more →
Decision observability
Logging every decision end to end and measuring quality, latency, calibration, intervention, compliance and economic impact. Learn more →
Decision owner
The named role accountable for a decision type’s model, policy, autonomy level and outcomes. Learn more →
Decision replay
Reconstructing a past decision with the inputs, model versions and policies in force at the time. Learn more →
Decision ROI
Decisions per year × improvement per decision × value of that improvement, minus the full program cost, measured against an agreed baseline. Learn more →
Decision safety
Controls that keep AI-assisted decisions within acceptable risk: calibrated confidence, abstention, disagreement handling, envelopes, reversibility and escalation. Learn more →
Decision twin
A digital twin designed around a decision: it forecasts how findings evolve, prices acting now versus waiting, and routes the recommended option to the person with authority. Learn more →
Digital twin
A live digital model of a physical asset, process or system that reflects its current condition and can be used to test decisions. Learn more →
E
Edge intelligence
Models and rules that run on devices or on site, so decisions happen where the data is, in milliseconds, without a round trip to the cloud. Learn more →
Electric submersible pump (ESP)
A downhole pump used for artificial lift in oil wells; its failure is a common, costly production decision. Learn more →
Embodied AI
AI that perceives and acts through a physical body such as a robot; closely related to physical AI. Learn more →
H
Holdout group
Lines, sites or queues deliberately left on the old way of deciding, so the difference in outcomes measures the real effect. Learn more →
Human-in-the-loop
A human approves each decision before it executes. Learn more →
Human-on-the-loop
The system executes within a time window or envelope while a human monitors and can override. Learn more →
I
Insight-to-action gap
The delay and loss of reasoning between noticing a problem in data and changing something in an operational system. Learn more →
K
Knowledge graph
A connected representation of entities and relationships used to assemble decision context across systems. Learn more →
M
Mission-capable rate
The share of time an asset such as an aircraft can perform at least one of its missions. Learn more →
O
OEE (overall equipment effectiveness)
A manufacturing metric combining availability, performance and quality. Learn more →
Operational intelligence (OI)
Real-time analytics on live operations — what is happening right now across lines, fleets or sites — usually with alerts. Learn more →
P
P10 / P50 / P90
Percentile outcomes of a probabilistic forecast, used to show uncertainty ranges instead of single numbers. Learn more →
Physical AI
AI systems that perceive, reason and act in the physical world through robots, vehicles, drones and machines. Learn more →
Physical intelligence
A term used for AI capabilities that let machines understand and act in the physical world; often used interchangeably with physical AI. Learn more →
Predictive maintenance
Using condition data to predict equipment failure. Decision intelligence adds the decision about what to do given plans, spares and cost. Learn more →
Prescriptive analytics
Analytics that recommend actions, typically via optimization or simulation; decision intelligence extends it with authority, execution and feedback. Learn more →
Prior authorization
A health-plan requirement to approve certain services before they are provided; a high-volume decision well suited to AI-assisted triage with clinical review. Learn more →
R
Retrieval-augmented generation (RAG)
A technique that retrieves relevant documents or data and gives them to a language model as context for its answer. Learn more →
S
Safety envelope
Hard physical or regulatory limits that no recommendation or automated action may cross. Learn more →
Scenario intelligence
Generating feasible options for a decision and simulating their outcomes, uncertainty and trade-offs before commitment. Learn more →
Shadow mode
Running a new decision model alongside people without acting on it, to compare what it would have done with what actually happened. Learn more →
W
Write-back
Committing a decision’s action into an operational system such as ERP, MES, WMS or a fleet manager. Learn more →