These short definitions help you read the guide. Open the linked chapter for qualifications and worked examples. Product labels can differ, so establish what the actual application does.
| Term | Meaning in this guide | Read more |
|---|---|---|
| Agent | A system that uses a model to choose some next steps within a task and application limits. | Chapter 34 |
| API | An interface through which software requests data or operations from another system. | Chapter 34 |
| Automation | Software that carries out a defined task or sequence under specified rules. | Chapter 32 |
| Authoritative source | The record or approved document a business uses to establish a particular fact or state. | Chapter 8 |
| Baseline | The defined comparison condition, with its measured results distinguished from assumptions. | Chapter 5 |
| Cash flow | Cash received and paid during a defined period. | Chapter 27 |
| Confabulation or hallucination | Generated content that is false or unsupported; fluent wording does not establish evidence. | Chapter 9 |
| Context | Information available to the model for the current task. | Chapter 35 |
| Contribution | Revenue less the explicitly included variable costs; the cost scope must be stated. | Chapter 23 |
| EBITDA | Earnings before interest, taxes, depreciation, and amortization; it is not cash flow. | Chapter 26 |
| Evaluation | A documented assessment against criteria for the intended task. | Chapter 9 |
| Fine-tuning | Additional training that changes model parameters; adding a source document need not do this. | Chapter 8 |
| Guardrail | A constraint or check on behavior or outcomes; its actual enforcement must be understood. | Chapter 18 |
| Human review | An accountable inspection of a specific result against the necessary sources and criteria. | Chapter 9 |
| Idempotency | Defined handling that lets the same intended request be repeated without an additional effect. | Chapter 32 |
| MCP | Model Context Protocol: a protocol for connecting AI applications with context and capabilities. | Chapter 34 |
| Prompt | Instructions and supplied context for a model task. | Chapter 7 |
| Prompt injection | Instructions embedded in an input or retrieved source that attempt to redirect the assistant. | Chapter 33 |
| RAG | Retrieval-augmented generation: supplying retrieved information to support generated answers. | Chapter 8 |
| ROI | Specified benefit minus specified cost, divided by that cost; the period and benefit view matter. | Chapter 26 |
| RPO | Recovery point objective: the point in time to which data must be recoverable. | Chapter 39 |
| RTO | Recovery time objective: the target recovery duration for a defined capability. | Chapter 39 |
| Staff capacity | Time available for other work; it does not automatically reduce cash expenditure. | Chapter 26 |
| Workflow | A defined arrangement of steps, dependencies, review, actions, and exception handling. | Chapter 32 |