---
title: "Do we need clean data before starting AI?"
description: "No. You need a few hundred representative real examples for the target workflow and permission to use them. The first build often includes the data plumbin"
canonical: https://sudosapient.si/answers/do-we-need-clean-data-before-starting-ai
updated: 2026-10-27
publisher: Sudo Sapient (https://sudosapient.com/)
---
# Do we need clean data before starting AI?

**No. You need a few hundred representative real examples for the target workflow and permission to use them. The first build often includes the data plumbing for that workflow.**

Language models are built to handle messy text, so perfect data is rarely the blocker. What matters is that you can export a few hundred real examples of inputs and the decisions or outputs that followed, and that your policies allow you to use them with an AI provider. A prototype on a random sample of real data reveals the edge cases early, which is far cheaper than a data-cleaning project that runs first.

## Go deeper

- [AI Readiness Checklist: 25 Yes/No Checks for One Workflow](https://sudosapient.si/readiness-checklist)

## More on transformation and strategy

- [What is AI transformation?](https://sudosapient.si/answers/what-is-ai-transformation)
- [How is AI transformation different from digital transformation?](https://sudosapient.si/answers/how-is-ai-transformation-different-from-digital-transformation)
- [How long does AI transformation take?](https://sudosapient.si/answers/how-long-does-ai-transformation-take)