An AI glossary for business, explained clearly

Short definitions of the terms that come up when you bring artificial intelligence into a company, and how each one takes shape in rekiam.

Updated

A

AI adoption

The extent to which the workforce actually uses AI in its work. It's measured through active people, features used and perceived quality, and it's what makes spending on AI pay off.

Adoption and spending

AI agent

An AI that doesn't just answer but plans and carries out tasks using tools: searching, reading documents, drafting emails or querying a system. In rekiam the agents are Reki Bots, which work with scheduled tasks and triggers and leave anything sensitive awaiting approval.

Discover Reki Bot

AI governance

The set of rules, permissions and controls by which an organisation decides who uses AI, what for, with which data and under what oversight. It includes usage policies, sensitive data protection, compliance review and traceability.

AI governance in rekiam

AI operating system

A platform that acts as the shared layer on which a whole company uses AI, rather than a standalone assistant. It brings context, permissions, integrations, bots, procedures and spending control together in one place.

See features

B

Batch processing

Applying the same instruction to many documents at once, for example extracting the supplier, date and amount from hundreds of invoices, and getting the result in a table.

Document batches

C

Canvas

A panel next to the chat in which a document is written and edited side by side with the AI: it drafts, you correct by hand, you select a paragraph to ask for a change, and versions are kept that you can restore.

The rekiam canvas

Context

The information the AI needs to answer the way the company would: who it is, what it sells, how it speaks and which documents are the reference. In rekiam it's written by level (organisation, department, section and person), and each item is a context, a rule, a prohibition or a procedure.

Contexts in rekiam

Corporate AI

Artificial intelligence configured for a specific company: it knows its culture, tone, policies and documentation, it's organised by departments and teams, it applies rules nobody can bypass and it works with the company's tools. Unlike a generic AI chat, it doesn't start from scratch in every conversation.

How it works in rekiam

H

Human approval

The principle whereby the AI prepares an action but a person confirms it before it happens. In rekiam, sending, sharing, inviting or deleting stays as a draft with Send, Edit or Cancel, and company rules can also require a lead's sign-off.

Lead approval

I

Instruction inheritance

The mechanism by which a higher level's instructions apply automatically to the levels below. In rekiam there are six layers, from the platform down to the person, and a lower level can never override a rule or prohibition from above.

The inheritance engine

Isolation between companies (multi-tenant)

An architecture in which several organisations use the same platform with their data kept completely separate: every piece of data belongs to an organisation and is only queried within it.

How rekiam protects data

M

MCP (Model Context Protocol)

An open standard for connecting an AI to external systems, such as a CRM, an ERP or a database, through “MCP servers” that expose queries and actions. In rekiam, an administrator connects the company's systems from a catalogue or with a custom connector.

Company connectors

P

Prompt injection

An attempt to give orders to an AI through content it reads, such as an email or a web page crafted to deceive it. rekiam mitigates it by treating whatever comes from outside as third-party content, not as orders: after reading it, anything the AI wants to write or send waits for the person's approval.

Security in rekiam

R

RAG (retrieval-augmented generation)

A technique in which the AI first searches a document base and answers from what it finds, citing the source. In rekiam, when the company's documentation doesn't fit into the context in full, the AI searches and reads what it needs at each moment, respecting what each person is allowed to see.

Knowledge at scale

Runnable procedure

A company process turned into a form and a list of steps that the AI follows in order alongside the person, with checkpoints, evidence and verification criteria, and with every run tracked.

Procedures in rekiam

S

Sensitive data

Information that mustn't leave the company unchecked: Spanish DNI, NIE and CIF numbers, IBANs, cards, Social Security numbers, passwords or confidential terms. A corporate AI platform must be able to warn about, mask or block them before they reach the AI or go out in an action.

Sensitive data protection

SSO and SCIM

SSO is single sign-on with the company's identity provider (Microsoft Entra ID, Okta, Google Workspace). SCIM is the standard by which that directory automatically adds, updates and removes people.

Controls for IT

T

Trigger

An event that sets a bot to work without anyone writing to it: a new email, a Slack mention, a meeting starting, a new file in a folder or an alert from another system via webhook.

Reki Bot triggers

Your company already has its own standards. Make the AI follow them.

Tell us how your team works and we'll show you rekiam with examples from your day-to-day: your departments, your documents and your policies. The first call is free.

Already a customer? Log in to rekiam

La R de rekiam, tejida