AI Glossary / Data engineering and architecture / Beginner

Database

Definition

A database is an organized collection of data that can be stored, accessed and maintained as a coherent whole. It is usually managed through database software that provides ways to retrieve information and control changes.

What it means

A business database might hold products, orders and customers. A personal application might use one for notes or household inventory. The collection can be small or large; size alone does not define a database.

It is useful to distinguish the data from the database management system, or DBMS, which manages it. People sometimes use “database” loosely for both. In a technical discussion, separating them helps clarify whether the subject is the information, the software or the service running that software.

Databases use different organizing approaches. Relational systems represent information in tables and relationships; document and graph systems offer other structures. None of these forms guarantees that the stored claims are true. A database can preserve and retrieve incorrect information just as efficiently as correct information.

How it works

A design specifies what data represents and how records are identified and related. Applications use supported queries or APIs to read and write it. The management software applies the relevant access rules, constraints and storage operations.

Features vary, but many systems provide indexes for efficient retrieval, transactions for grouped changes and mechanisms for coordinating concurrent users. Their guarantees depend on configuration and operation, not merely on being called a database.

Deployment also varies. A database can be managed by a separate server, a cloud service or an embedded engine within an application. SQLite illustrates an embedded approach in which the engine does not require a separate database-server process. The logical collection should not be confused with a particular machine or a requirement to be online.

Three practical examples

1. Online shop: an application records available products and accepted orders, using defined identifiers so each order line refers to the intended product.

2. AI glossary website: a database stores terms, article versions, categories and review states, allowing the website to show the approved version and retain appropriate editorial history.

3. Local planning tool: an embedded database stores tasks and notes on a device, so the application can search and update them even without a network connection, subject to its own design.

Think of it this way

Think of an organized records office with labeled collections and established ways to find and update information. The database is the organized information, while the management system is comparable to the procedures and tools used to maintain it. The analogy does not capture every technical feature, such as concurrent transactions or automated recovery after failure.

Worked example

A fictional glossary contains three term records: T1, T2 and T3. Each has a stable identifier and a displayed name. Articles are stored separately with an article identifier, a term reference, a version number and an editorial status.

Term T1 has two article versions: version 1 is published and version 2 is draft. A website query should select the approved published version according to the project's rules, rather than simply choosing the most recently edited row.

If an editor renames T1, its stable identifier can preserve the relationship to both article versions. Using the displayed name as the only link would make renaming more fragile.

The example illustrates organization, identity and retrieval rules. The database does not decide what publication approval means; the application and editorial process must define and enforce that meaning.

Illustrative case study

Illustrative fictional case: Spruce Services stores customer records in a shared spreadsheet. As several staff members update it, they create conflicting copies and inconsistent identifiers. The team moves the workflow into an application backed by a database.

It defines customer identity, required fields and update permissions, then checks imported records for duplicates. It also tests backup restoration and clarifies who resolves conflicting source information.

The fictional lesson is that the migration needs both software capabilities and agreed operating rules. Copying the spreadsheet into a database would not, by itself, settle duplicate identities or make every existing address correct. No measured productivity or accuracy gain is claimed.

Common misconceptions and mistakes

Do not assume a database must be a giant server or use SQL. Do not treat a database, a DBMS and a running database instance as interchangeable in every context.

Do not equate stored data with verified knowledge. Do not assume a backup exists just because the service is available, or that replication protects against every accidental deletion. Avoid confusing application search indexes with authoritative source records unless that role is explicitly intended. Finally, a database's technical permissions do not replace clear business rules about appropriate data use.

Business use

Use a database when information needs organized retrieval, dependable updates, relationships or coordinated access beyond what simpler storage provides. Choose its structure and capabilities around actual operations rather than popularity alone.

For AI applications, decide which records are authoritative and which are derived, such as embeddings or extracted summaries. Preserve source references and update paths so corrections can reach the appropriate outputs. Review access, retention, recovery and data quality as part of the operational design. The aim is information that remains understandable and manageable as people and software change it.

Check your understanding

An AI assistant retrieves a customer's old address accurately from a database. Does accurate retrieval establish that the address is still current?

Answer: No. It establishes what the queried records contain. Currency depends on the source, update process and the meaning of the stored fields.

Related terms

Learn next

Learn the difference between a database and its management software, then practice retrieving and updating a small example. Next, study identifiers, constraints and transactions so you can explain how records stay related and how multi-step changes are controlled.

Sources

Oracle, database concepts and the distinction from management software: https://www.oracle.com/database/what-is-database/

Oracle, introduction to databases and DBMS responsibilities: https://docs.oracle.com/html/E10713_02/intro.htm

SQLite, embedded serverless operation: https://www.sqlite.org/serverless.html

PostgreSQL, transactions as grouped operations: https://www.postgresql.org/docs/16/tutorial-transactions.html

The glossary and customer-service examples are original fictional illustrations.

Previous: Data wrangling Next: Database-as-IPC