It is tempting to start an AI project by hiring an engineer. But if your data is scattered, incomplete or inaccessible, that engineer will spend most of their time cleaning it. Checking data readiness first saves time and money.
1. Do you know where the data lives?
- List every system involved: CRM, ERP, store, spreadsheets, email, shared drives.
- Note who owns each one and how it can be accessed (export, API, database).
2. Is it accessible?
AI tools need programmatic access. Check whether your systems offer APIs or exports, and whether licences allow it. Data trapped in PDFs or paper may need extraction first.
3. Is it consistent?
- Are product names, customer IDs and dates formatted the same way everywhere?
- Are there duplicates or conflicting records?
- Do fields mean the same thing in every system?
4. Is there enough of it?
For AI that uses existing models (like a chatbot over your documents), you need complete, up-to-date documents. For custom predictive models, you need enough historical examples — including the outcomes you want to predict.
5. Is it labelled?
If you want AI to classify or predict, you need examples with the right answer attached: "this email was a complaint", "this order was returned". Missing labels are a common hidden cost.
6. Is it allowed to be used?
- Check privacy rules for personal data (for example GDPR)
- Review contracts with customers and suppliers
- Decide what data may be sent to external AI services
Quick scoring
| Status | What it means |
|---|---|
| Mostly "yes" | You are ready to start an AI pilot. |
| Mixed | Start with a short data-integration project first. |
| Mostly "no" | Focus on connecting and cleaning systems before AI. |
Data work is valuable on its own
Even before AI, connecting systems and cleaning data reduces manual work and improves reporting. Many of our API integration projects deliver value immediately — and make later AI projects much faster.
Need your systems connected first? See our API integration services or hire an API integration developer.