Is Your Data Ready for AI? Free Checklist

Most AI projects in small and mid-size businesses stall for the same reason: the data is not ready. This free checklist helps you test your supply chain data before you spend money on AI tools, and shows what to fix first.

Why data readiness comes first

AI tools learn patterns from your records. If part numbers are duplicated, lead times are out of date or inventory balances are wrong, the tool will learn the wrong patterns and present them with confidence. Fixing data is rarely exciting, but it improves today’s decisions even if you never buy an AI product.

The NIST AI Risk Management Framework treats data quality as part of managing AI risk, and its companion NIST Generative AI Profile (NIST AI 600-1) flags information integrity as a core concern. You do not need to adopt either framework formally to benefit from their questions.

The checklist

Score each item as yes, partly or no. Be honest; the point is to find gaps.

Item and supplier records

Every item has a single part number, description and unit of measure.
Obsolete and duplicate items are marked or merged.
Each purchased item has a current preferred supplier and a realistic lead time.
Supplier records include contact details and status.

Transactions

Sales or usage is recorded at the item level for at least one to two years.
Large one-off orders, returns and corrections are identifiable.
Purchase orders show promised and actual receipt dates.
Production or work orders record actual quantities and completion dates.

Inventory accuracy

On-hand balances match physical counts closely enough that planners trust them.
Cycle counting happens on a regular schedule, with differences investigated.
Locations and units of measure are consistent across systems.

Ownership and access

Someone owns each data set and is responsible for its accuracy.
Data can be exported from your ERP or spreadsheets without manual retyping.
You know which data is confidential and who may share it with outside tools.

How to read your score

Mostly yes: you are ready for a focused AI pilot. Pick one decision, such as reorder suggestions, and trial a tool on your own data.
Mixed: run a short cleanup on the weakest section first, usually lead times or inventory accuracy, then pilot.
Mostly no: start with the basics. A cleaner item master and a regular cycle count will pay off before any AI tool does.

Our supply chain data readiness checklist goes deeper on each area, and our purchased parts review turns lead-time cleanup into a monthly habit.

Quick fixes that make a difference

Merge duplicate part numbers and mark obsolete items inactive.
Compare promised and actual supplier dates for your top items and update lead times. Our reorder point calculator shows how lead time drives stock levels.
Use an ABC XYZ classification to focus counting and cleanup on the items that matter most.
Agree on one owner for the item master and a simple rule for creating new items.

Protect the data while you prepare it

Readiness also means knowing what must stay private. Before sharing data with any AI vendor or public tool, mask customer names and prices, and confirm how the tool stores and uses your data. The NIST Cybersecurity Framework and CISA’s AI resources offer practical security guidance for small organizations.

Next steps

If your checklist shows gaps and you want help closing them, see our AI support services. We start with a free call, and scope is agreed before we start. You can also read our guide to AI in supply chain for small business for use cases once your data is ready.

A 30-day readiness plan

If your score shows gaps, a short, focused plan works better than a large cleanup project that never finishes:

Week 1: export your item master, open purchase orders and 12 to 24 months of usage history. Run the checklist and list the three biggest gaps.
Week 2: fix duplicate and obsolete items for your highest-value A items first.
Week 3: update lead times for your top suppliers by comparing promised and actual receipt dates.
Week 4: cycle count your A items, investigate differences and rerun the checklist.

At the end of the month you will have cleaner data, a clearer picture of what remains and a better basis for deciding whether to trial an AI tool.

Frequently asked questions

How much history do we need?

More is generally better, but quality matters more than length. One to two years of clean item-level history is a reasonable starting point for most forecasting and inventory tools, because it shows at least one full seasonal cycle.

Can AI clean our data for us?

It can help. AI tools are good at suggesting likely duplicates or grouping messy descriptions. A person who knows the parts should still approve every change, because a wrong merge can cause real ordering errors.

Is a spreadsheet good enough?

Often, yes, for a first pilot. What matters is that the data is consistent and owned. Our guide to outgrowing Excel explains when it stops being enough.

Is Your Data Ready for AI? A Free Supply Chain Checklist