---
title: "Glossary of data and AI in business processes"
description: "The terms we use when we talk about data and AI in processes, from test bench to shadow AI, each with the page that explains it."
url: https://www.joinx.it/en/resources/glossary
---

# The words we use to talk about data and AI in companies

Here we bring together the terms that recur across our pages, explained as we use them in projects, each with a link to the page that covers it in depth.

## The terms, in alphabetical order

Some are our own names, such as the solutions and modules; others come from the trade or from the law. For AI Act terms we give the meaning we use in our articles, while how the law applies to your case is for your legal and compliance functions to establish.

Adoption

With Adoption, the organisation and its people come to use AI and to steer it, because we run adoption as a project, with a plan and someone accountable for it. We write the plan with the Organisation and HR department and with the heads of the functions involved, and it says which roles change and how we measure the use of AI at work.

[The Adoption solution](https://www.joinx.it/en/solutions/adoption)

Agent sheet

Before it goes into production an agent has a sheet, which those accountable for the process and those who carry out controls can consult at any time. It sets out the mandate, the owner, the systems the agent accesses, the rules, the approval points, the conditions under which the agent stops and the measures, and it is the first document an auditor asks to see.

[The sheet in the Agents module](https://www.joinx.it/en/technology/agents#la-scheda-di-un-agente)

AI agent

An AI agent does a job inside a process, in the systems the company already uses, applies the rules of the Company Brain and stops when a person's decision is needed. The agents we build have a mandate and an owner within the company, and for security purposes we treat them as a new type of user, with an identity and only the permissions the task requires.

[The Agents module](https://www.joinx.it/en/technology/agents)

AI literacy

Article 4 of the AI Act asks providers and deployers to attend to the AI literacy of the staff who use AI systems. As currently worded, it asks for measures to support the development of these skills, without requiring a set level to be guaranteed for each person, and in our projects it is handled through Adoption, with role-based paths for which we document attendance, content and results.

[The article on the AI Act](https://www.joinx.it/en/resources/ai-act-what-it-asks-of-companies)

Approval point

This is the step at which an agent's work stops for the decision of a person, identified by role in the agent sheet. It works if the approver has in front of them the sources and the reason the case has come to them, and so that approval does not become an automatic gesture we keep up sample checks even when results are good.

[The article on the approval point](https://www.joinx.it/en/resources/agents-owner-and-approval)

Company Brain

In the Company Brain we bring together the company's processes, rules and decisions, with the reasons behind them, in a knowledge base that people and agents consult within precise permissions. It starts with the first project and contains what that process needs, written so that later processes find it ready.

[The Company Brain module](https://www.joinx.it/en/technology/company-brain)

Compass

Compass is the questionnaire, with an interview led by an AI agent, through which in Mapping we gather the work as described by the people who do it. Participation is voluntary and each participant receives a personal compass, while management receives only aggregate readings by role and by process, which play no part in individual appraisals.

[Compass in Mapping](https://www.joinx.it/en/solutions/mapping#compass)

Compounding Intelligence

By Compounding Intelligence we mean the principle of compound interest applied to knowledge: an AI project leaves the company with data, rules and skills that make the next one faster and less costly. When a project closes we write down what it leaves that can be reused, and when the next one starts, what was actually reused.

[Compounding Intelligence](https://www.joinx.it/en/compounding-intelligence)

Data Foundation

Data Foundation is the module that prepares data for people and for agents. We connect the sources without replacing existing systems and give each data item a shared definition, an owner and a quality rule, so that the analyst, management control and an agent all read the same number.

[The Data Foundation module](https://www.joinx.it/en/technology/data-foundation)

Data Intelligence

In Data Intelligence we build, on the company's data, the analyses, predictive models and scores that decision-makers need, for example to rank by risk the accounts to chase for payment. The model provides estimates and priorities, the decision stays with people, and alongside it the model that contributed is recorded, with the version in use.

[The Data Intelligence module](https://www.joinx.it/en/technology/data-intelligence)

Data product

Data becomes a data product when it has an owner, a usage sheet and quality measured against a threshold, and is ready for those who have permission to use it. When we prepare the data for a process we deliver it in this form, documented in a place that whoever comes next can find without having to ask.

[What Data Foundation contains](https://www.joinx.it/en/technology/data-foundation#che-cosa-contiene)

Deployer

Under the AI Act, the deployer is whoever uses an AI system in the course of their own business, while the provider develops it and places it on the market. Most companies are deployers, but in the case of high-risk systems, whoever substantially modifies one, or puts it on the market under their own name, takes on the provider's obligations.

[The article on the AI Act](https://www.joinx.it/en/resources/ai-act-what-it-asks-of-companies)

Deployment

In Deployment we take a use case that has already been chosen into production, inside the company's systems, from defining the case and preparing the data through to testing the agents and release. We consider the work finished when the team that runs the process uses it every day, with a measure of what has changed.

[The Deployment solution](https://www.joinx.it/en/solutions/deployment)

Domain-specific vertical AI

This is our name for solutions that grow out of knowledge of an industry's processes and organisations, thanks to the business experts of the Join Business Group, who work in the same group as our engineers. A solution of this kind connects to the systems already in use, including the older ones that are harder to integrate, and works on the company's data and knowledge.

[About us](https://www.joinx.it/en/about-us)

Forward deployed engineer

These are our data and AI engineers who work on site at the company, alongside the people who know the process, because many rules only emerge by watching the work as it happens. They stay with the client's team from the analysis of needs through to release into production and full adoption of the solution, without stopping at a prototype.

[How Deployment works](https://www.joinx.it/en/solutions/deployment#come-si-svolge)

Fractional Chief AI Officer

This is the Presidio option in which one of our senior professionals works alongside management as a part-time head of AI. They attend the management committee at regular intervals and keep a single view of data and AI initiatives, each with an owner and a measure, while objectives and decisions remain with management.

[The option in Presidio](https://www.joinx.it/en/solutions/presidio#fractional-chief-ai-officer)

Governance

Governance is the module that sets the rules for using agents and models and keeps the evidence. We record agents, models and actions as the work happens and collect the evidence the regulations require during the project, so when an auditor asks who approved a decision, the answer is already written and dated.

[The Governance module](https://www.joinx.it/en/technology/governance)

Harness

This is the set of instructions, tools, rules, access rights and memory that keeps an agent at work inside a process. It is usually built into each agent, which carries its own copy, whereas in X Lab we build it once, in the company's world, and every agent inherits it.

[The world of X Lab](https://www.joinx.it/en/x-lab#il-mondo)

High-risk AI system

The AI Act treats as high-risk the uses listed in its Annex III, and in the industries we work in the most frequent are the assessment of the creditworthiness of natural persons, risk assessment and pricing in life and health insurance, and the recruitment or evaluation of staff. Anyone using a system of this kind must assign its oversight to people with the necessary competence and authority, keep its logs and inform workers before using it in the workplace.

[The article on the AI Act](https://www.joinx.it/en/resources/ai-act-what-it-asks-of-companies)

Levels of autonomy

Agents work at three levels of autonomy, Observe, Prepare and Act within limits, which run from gathering information, to preparing a complete piece of work, to defined actions within agreed thresholds. We usually start at the first or second level; the move to the next is decided by whoever is accountable for the process, looking at the results on the test bench; and irreversible decisions always stay with a person.

[The levels in the Agents module](https://www.joinx.it/en/technology/agents#tre-gradi-di-autonomia)

Lineage

Lineage traces where each data item comes from and how it has been transformed, so that whoever uses it can verify it. In Data Foundation it is part of the work on quality and traceability, together with measuring how complete and up to date the data is.

[What Data Foundation contains](https://www.joinx.it/en/technology/data-foundation#che-cosa-contiene)

Mapping

With Mapping we examine the processes, data, systems and organisation of one area or of the whole company, and give management a portfolio of initiatives ranked by value and feasibility, each with its business case. The result also includes the list of initiatives not to take on for now, and the documents belong to the company, which can use them with any supplier.

[The Mapping solution](https://www.joinx.it/en/solutions/mapping)

Model agnostic

This describes our approach, which is independent of individual models and technology vendors: models, runtime environment and cloud can be replaced while rules and evidence stay intact. When we change an agent's model we do not rebuild the agent, and we test it again on the test bench before putting it back to work.

[The platform architecture](https://www.joinx.it/en/technology#architettura)

Ontology

The ontology is the part of the Company Brain that describes the company's concepts and the relationships between them, such as customers, contracts, guarantees and policies. Like the data, the definitions and the rules, it belongs to the client, in its own systems and in open, exportable formats.

[What the Company Brain contains](https://www.joinx.it/en/technology/company-brain#che-cosa-contiene)

Presidio

Presidio is the ongoing service through which we monitor, update and improve agents and models in production as models, data, regulations and procedures change. On the shared dashboard the company reads the same measures we do, and the knowledge transfer written into the contract allows it to leave Presidio when it is ready.

[The Presidio solution](https://www.joinx.it/en/solutions/presidio)

Run log

In the run log, Governance notes, for each action by an agent, when it took place, on which data, under which rules and who approved it. An auditor can consult and verify it, and the log remains the company's even when it decides to run the agents in-house or with another supplier.

[What Governance contains](https://www.joinx.it/en/technology/governance#che-cosa-contiene)

Security

Security is the module that protects data, access and agents within the company's security rules, and enforces on the systems the rules that Governance sets. We give each agent an identity and only the permissions of its mandate, and we check against the mandate any action that a piece of content suggests to it.

[The Security module](https://www.joinx.it/en/technology/security)

Shadow AI

This is the name for AI tools already in use in the company but outside any control, such as an email assistant or a service that summarises meeting minutes. With Mapping we find them in the interviews and by cross-checking what emerges against the list of systems and licences, then we enter them in the register, each with an owner and a risk level.

[The inventory in Governance](https://www.joinx.it/en/technology/governance#the-inventory-of-ai-systems-already-in-use)

Sovereign AI

For us, sovereign AI means that the company keeps control of its own data and knowledge, decides how to govern the solutions and remains free in its technology choices. That is why we work in the client's cloud, in a European cloud or in an isolated environment, and connect systems and agents using open standards such as MCP and A2A.

[The platform architecture](https://www.joinx.it/en/technology#architettura)

Test bench

The test bench brings together the real cases, with known outcomes, on which we test an agent before release and at every change of model. The thresholds are agreed before building and, if the agent does not meet them, we do not release; looking at these results, whoever is accountable for the process also decides when the agent moves up to the next level of autonomy.

[What the Agents module contains](https://www.joinx.it/en/technology/agents#che-cosa-contiene)

Workspace

Workspace brings the agents' work to the people who have to use or approve it, in their everyday tools, such as email, the company chat or a process page. Next to the draft the person finds the sources and the reason for the check, and sees only the data their role already allows.

[The Workspace module](https://www.joinx.it/en/technology/workspace)

## Apply these terms to one of your processes

If a term describes an issue you have on your desk, we would rather look at it together on your own case.

[Let's talk](https://www.joinx.it/en/lets-talk)
