Guide

What is a data ontology?

In short

A data ontology is a map of your operation: the objects you work with, such as customers, orders or case files, the links between them and the actions people take, all mapped onto the systems you already run. It gives your data, your applications and AI one shared language.

The pieces, without jargon

  • Objects

    The things your organisation works with and people call by name: a customer, an order, a case file.

  • Links

    How they relate: a customer places orders, a case file concerns a property.

  • Actions

    What people do with them: confirm an order, request documents, issue an invoice. With permissions and a record.

  • On top of your systems

    The ontology does not replace the ERP or the case management system. It sits on top and says where each piece of data lives.

Two examples

Generic examples, to illustrate the idea. Every organisation has its own.

Objects

A town hall
Case file, citizen, property, local tax
A business
Customer, order, invoice, warehouse

Links

A town hall
A citizen opens a case file that concerns a property, and the property generates a local tax
A business
A customer places an order, the order ships from a warehouse and generates an invoice

Actions

A town hall
Register an application, request documents, settle a tax
A business
Confirm an order, reserve stock, issue and collect an invoice

Where the data lives today

A town hall
Case management, the municipal register, tax collection, spreadsheets
A business
ERP, CRM, warehouse software, spreadsheets

Why it matters

Without a shared language, every system names things its own way and every report needs its own translation. With an ontology, a dashboard, an internal tool and an AI assistant all use the same definitions.

It is especially useful for AI: an assistant that knows what a case file is and which actions it may take works within clear limits and leaves a record of what it does.

How to build one step by step, without buying a platform

  • Start with one process

    One that hurts and can be measured. There is no need to model the whole organisation on day one.

  • Name the objects with the people who do the work

    The definitions come from the operation, not from a systems meeting.

  • Link, do not copy

    Each object points to the system where its data lives, without duplicating it.

  • Add actions with permissions

    Every action has someone allowed to take it, and it is recorded.

  • Grow in stages

    Each new process reuses the objects that already exist.

When a platform’s built-in ontology is the right call

Some data platforms come with an ontology built in; the term was popularised in enterprise software by platforms such as Palantir Foundry. If you have many data sources at scale, a team to run the platform and budget for licences, it can be the right choice, and we can work alongside it.

If not, a lightweight ontology on top of your systems, in code you own, covers the essentials and can grow later.

Do you need a platform? Read this

Frequently asked questions

Is an ontology the same as a database?
No. The database stores the data. The ontology describes what it means, how it relates and what can be done with it, even when it is spread across several systems.
What is the Palantir ontology?
Palantir calls Ontology a layer of its Foundry platform that models objects, links and actions. The general idea, a data ontology, does not depend on any product and can be built on the systems you already have.
Do I need to buy a platform to have an ontology?
No. It can be built step by step on your ERP, your CRM or your case management system, in code you own. A platform makes sense when the scale of the data and the team justify it.
What does it have to do with artificial intelligence?
It gives AI context and limits. The AI knows which objects exist and which actions it may propose or take, with permissions, a record and human review where needed.
Where do we start?
With a fixed-price Recon: we pick one process, name its objects with the people who do the work and see which systems hold its data.

Keep reading

  • Do you need Palantir?

    A data platform, or integrating what you already have: an honest table for deciding before you spend on licences.

  • Custom software: when it pays off

    An honest table for choosing between custom and off-the-shelf, and how it is built on what you already run.

  • Subcontracting for integrators

    What we take on inside an integrator’s project, how it is contracted and how we protect their client relationship.

  • Public sector

    Legacy systems, unified data and citizen-facing apps, delivered through integrators and framework agreements.

See all services

Start with a Recon.

We map the process, the data and the systems. You leave with a plan and a fixed price, and if we carry on, it is credited against the Launch.

Book a Recon

Palantir and Foundry are trademarks of Palantir Technologies Inc. SME Analytica is not affiliated with, endorsed by or a partner of Palantir.