Innovation glossary
Last reviewed August 26, 2026Plain-English definitions for the technologies, operating models and metrics used across Actuneuriat.
AI agent
Software that can interpret a goal, select or sequence actions and use permitted tools or data to advance the task. An enterprise agent is defined as much by its permissions, evaluation and escalation path as by the underlying model.
AI governance
The decisions, controls, evidence and accountability used to manage AI across its lifecycle. Effective governance connects policy to inventories, evaluations, access controls, monitoring and incident response.
Digital twin
A digital representation connected closely enough to a physical asset or process to reflect meaningful state and inform a real decision. A static three-dimensional model is not automatically a digital twin.
Embodied AI
AI that perceives and acts through a physical system such as a robot, vehicle or industrial machine. Reliability depends on sensing, controls, environment and task design, not only model capability.
Inference
The computation used when a trained model produces an output. In production systems, inference cost and latency depend on model size, hardware, workload design, utilization and where computation occurs.
Innovation portfolio
A deliberately managed set of initiatives with different time horizons, uncertainty and strategic roles. Portfolio health is not the same as the success of a single project.
Post-quantum cryptography
Cryptographic algorithms designed to resist attacks from both classical and sufficiently capable quantum computers. Migration begins with an inventory of cryptography and dependencies, not with an immediate replacement of every system.
Retrieval-augmented generation
A pattern in which a system retrieves selected information and supplies it as context to a generative model. It can improve grounding but does not guarantee accuracy, authorization or source quality.
Sovereign cloud
A cloud arrangement designed to meet defined requirements for jurisdiction, control, operations, data access or portability. Sovereignty should be evaluated as verifiable controls rather than a geographic label.
Total cost of ownership
The full cost of acquiring, integrating, operating, maintaining and retiring a capability. For automation, it includes engineering, safety, downtime, supervision, consumables and change management.