Glossary

Evidence of AI intervention

The record of which model intervened in a process, with what input, what output it produced and what action followed.

The record of which model intervened in a process, with what input, what output it produced and what action followed. It extends audit and process mining to the behaviour of artificial intelligence, not only that of people.

It is what makes a process with AI inside it auditable. With no record of which model intervened, on what input and with what output, the answer to "why was this decided" is that the system said so — which, in front of an auditor, is not an answer.

On this site: Governed AI →

Related terms

Agentic process automation

Automation where LLM-based agents execute activities with bounded autonomy inside a process — the emerging frontier of the orchestration market.

AI applications in ITSM

The use of AI — classification, prediction, grounded answers, agents — inside IT service management practices, and the market that evaluates it.

AI management system

A framework requiring artificial intelligence systems to be managed with objectives, roles, impact assessment and continual improvement.

AI subordinate to the objective

The principle governing the incorporation of artificial intelligence into a process: every AI intervention must be justified by its contribution to the objective, a constraint or an…

AI that assists versus AI that decides

An operational design distinction: AI may decide autonomously when the decision is structured, reversible and auditable; it assists but does not decide where expert judgement, ambiguity or…

Audit

The ability to demonstrate who did what, when, from where and on which object. It rests on per-instance trails recording action, date, time, user and device, for both active and closed instances.

Evidence

The set of data, documents, logs, states, decisions, signatures, timings and integrations that demonstrate what happened in a process instance.