The vocabulary of custom software and AI, in plain words.

The terms of business software, artificial intelligence and automation, defined simply, for company owners, with no needless jargon.

Agentic AI: An approach to AI where the system does more than answer: it runs multi-step actions inside your tools (read, decide, act) to reach a defined goal.
AI agent: An artificial-intelligence program built to carry a task from start to finish (prospecting, support, reporting) with minimal human supervision. Where a chatbot answers, an agent acts.
API: An interface that lets two pieces of software exchange data automatically. In 2026, how well a tool integrates by API has become the first criterion when choosing business software.
Automation: Running repetitive tasks (invoicing, follow-ups, reporting) without anyone touching them. According to McKinsey, around 30 % of working time can be automated.
Cloud Act: A 2018 US law allowing American authorities to reach data held by an American company, even when the servers sit in Europe. Hence the question of sovereignty.
Code ownership: The fact that a piece of software’s source code is delivered to you and belongs to you, 100 %. It is what separates an asset (custom) from a rental (SaaS).
CRM: Customer Relationship Management: software for managing the client relationship (contacts, opportunities, sales follow-up). Often at the heart of a small company’s operations.
Custom software: Business software designed specifically for one company, as opposed to a standard product. It follows your processes and, done properly, belongs to you (code delivered).
Dashboard: A screen showing the key indicators of the business in real time, so you can decide fast without hunting for the information across several tools.
Data sovereignty: Keeping your data under European jurisdiction, out of reach of the Cloud Act. It assumes EU hosting and a supplier with no US ownership.
Database: The system that stores and organises a company’s data in a structured way. One central database is the foundation any working AI needs.
ERP: Enterprise resource planning: a suite that centralises the company’s functions (finance, stock, HR, sales) in one database. Powerful, but standardised and slow to deploy.
EU hosting: Storing data on servers located in the European Union, under GDPR. A compliance argument that weighs more every year, particularly in the public sector.
GDPR: General Data Protection Regulation: the European framework governing the processing of personal data. With AI, GDPR and the AI Act apply at the same time.
Generative AI: Artificial intelligence that produces content (text, image, code) from an instruction. It answers and generates on demand, where an agent carries out a whole task.
KPI: Key Performance Indicator: a measure of progress towards a goal (revenue, processing time, conversion rate).
LLM (Large Language Model): A large language model trained on enormous volumes of text (GPT, Claude, Mistral and others). It is the engine behind most generative AI tools and agents.
Lock-in (supplier dependency): A situation where leaving a piece of software or its publisher becomes very costly or impossible. Custom software with the code delivered and reversibility removes that risk.
MVP: Minimum Viable Product: the first version of a piece of software, cut back to the essentials and put in service fast so it can be tested and improved with real users.
No-code: Tools that let you build automations or applications without programming (Make, Zapier, Airtable). Ideal for prototyping, but they hit a ceiling once the need becomes critical.
OCR: Optical character recognition: reading text automatically out of a document (invoice, contract). Paired with AI, it removes the data entry.
POC (Proof of Concept): A working prototype built quickly (often 3 to 6 weeks) to prove an idea works and to price its return before investing fully.
PostgreSQL: An open-source relational database known for its robustness. It is the technical foundation we recommend for building business software that can grow.
Prompt engineering: The craft of writing the instructions you give an AI so it returns the right result. It is often 60 to 70 % of what makes an AI project work, more than the choice of model.
RAG (Retrieval-Augmented Generation): A technique that connects an LLM to your own knowledge base: the AI looks the information up in your documents before answering, which makes it reliable on YOUR data.
Reversibility: The ability to take back your software, your data and the documentation and hand them to another supplier. Without it you are captive. A key criterion when choosing who builds for you.
ROI: Return on investment: what a project brings in, measured against what it costs. In automation and AI, a real return is counted in hours actually replaced.
SaaS: Software as a Service: software rented by subscription and hosted by its publisher. Quick to start with, but you own nothing and the cost runs for life.
SaaS spaghetti: A pile of SaaS tools that do not talk to each other, where you spend more time bridging them than working. The number one efficiency problem in small companies.
Supabase: An open-source platform built on PostgreSQL (database, authentication, API, real time). Self-hostable in the EU, and a solid base for a custom application.
Webhook: A mechanism that notifies one piece of software in real time when something happens in another. The basic building block of automation between tools.

A term you need cleared up for your project?

We speak your language, not jargon. A call to turn what you need into a concrete plan. Free.

Book a call · free

Let's work together