I can help you find the right room now. Choose a fast path or type what you are trying to solve.
Glossary
Plain language for serious AI work.
Folium Systems uses technical depth, but customers should not need jargon to understand the first step. This glossary explains common terms in operational language.
Glossary finder
Governance
AI Systems Audit
Plain English
A structured review of workflows, data, tools, risks, staff readiness, and first proof opportunities.
Technical meaning
Discovery and risk mapping across workflows, data, systems, AI usage, owners, and proof candidates.
Buyer question
Where should we start without buying the wrong AI tool?
Proof
Proof Portal
Plain English
A working sandbox experience that lets stakeholders inspect a future workflow before production wiring.
Technical meaning
A static or sandboxed application surface used to test workflow behavior before production integration.
Buyer question
Can stakeholders inspect the idea before production wiring?
Knowledge
RAG
Plain English
Retrieval-augmented generation: a pattern where AI answers from selected sources instead of relying only on general model memory.
Technical meaning
A retrieval layer that selects approved source context and passes it into a model response path.
Buyer question
Can the AI answer from our approved knowledge instead of guessing?
Governance
Human Review Gate
Plain English
A required point where a person reviews, approves, rejects, or escalates AI-assisted work.
Technical meaning
A control point that requires a person to inspect, approve, reject, or escalate AI-assisted work.
Buyer question
Where does a person stay in control?
Architecture
Local AI
Plain English
AI that runs on hardware or infrastructure closer to the business, often for control, privacy, cost, or resilience reasons.
Technical meaning
Model execution on controlled hardware, private infrastructure, or customer-owned environments.
Buyer question
Can sensitive or costly work run closer to us?
Architecture
Hybrid AI
Plain English
An architecture that uses more than one runtime path, such as local models for private work and cloud APIs for broader tasks.
Technical meaning
A routed architecture that uses multiple model/runtime paths under one control layer.
Buyer question
Which tasks should be local, private, cloud, or routed?
Knowledge
AI Estate
Plain English
The collection of models, prompts, tools, agents, data sources, runtimes, logs, approvals, and operating records around AI.
Technical meaning
The managed portfolio of models, prompts, agents, tools, data sources, logs, releases, and controls.
Buyer question
Who owns all the AI pieces after launch?
Governance
Evidence Binder
Plain English
A review packet containing scope, tests, screenshots, known limits, owners, launch gates, and production requirements.
Technical meaning
A review artifact that packages scope, tests, screenshots, owners, risks, limits, and next-stage gates.
Buyer question
What proof do reviewers need before the next stage?
Governance
Launch Gate
Plain English
A required review point where owners compare evidence, risks, approvals, support readiness, and rollback paths before advancing an AI workflow.
Technical meaning
A formal advancement decision tied to evidence, support readiness, rollback, approval, and monitoring.
Buyer question
What must be true before this workflow advances?
Proof
Precondition Ladder
Plain English
A list of what must become true before a workflow, model, agent, or integration is allowed to move to the next stage.
Technical meaning
Operational AI term translated into a practical review or implementation decision.
Buyer question
What business decision does this term help us make?
Governance
Service Boundary Contract
Plain English
A plain-language operating agreement that names what a delegated service may do, what it may not own, what evidence it returns, and when it must stop.
Technical meaning
Operational AI term translated into a practical review or implementation decision.
Buyer question
What business decision does this term help us make?
Proof
Rollback Trigger
Plain English
A specific signal that tells the team to pause, revert, or return to a safer mode, even if the software is still technically online.
Technical meaning
Operational AI term translated into a practical review or implementation decision.
Buyer question
What business decision does this term help us make?
Knowledge
No-Drift Migration
Plain English
A migration pattern that moves workload carefully while preserving source truth, ownership, review, recovery, and customer-facing honesty.
Technical meaning
Operational AI term translated into a practical review or implementation decision.
Buyer question
What business decision does this term help us make?
Governance
Binding Governance
Plain English
Governance that is enforced by the system through approvals, access limits, logs, fail-closed behavior, and human gates instead of only being written in policy.
Technical meaning
Operational AI term translated into a practical review or implementation decision.
Buyer question
What business decision does this term help us make?
Knowledge
Evaluation Scorecard
Plain English
A structured record of how an AI workflow performed across answer quality, source grounding, safe actions, reliability, latency, and critical failures.
Technical meaning
A scoring model for answer quality, workflow completion, groundedness, safe actions, latency, and failure modes.
Buyer question
How do we know the workflow is reliable enough?
Architecture
Data Boundary
Plain English
The rules that decide what information AI can see, where it can travel, how it is masked or retained, and what actions require human approval.
Technical meaning
The policy and technical control surface for what data AI can access, retain, transform, or route.
Buyer question
What data can AI see, where can it go, and who approves exceptions?
Start here
The terms matter less than the first workflow.
Folium can translate the technology into a path your business can inspect, operate, and improve.