AI on top of an ERP reads what is in the ERP. This checklist is a way to find out what that is before a project depends on it. None of it needs special tools: most checks are a query or a count that your own team can run.
Score each one honestly. A weak area is not a reason to stop; it tells you where the first piece of work lies.
01Materials
- How many material codes are active, and how many have had no movement in three years?
- For five common items, how many codes exist for each?
- What share of descriptions follow a standard pattern?
- What share have a class, a unit and a manufacturer part number filled in?
- Is a new code checked against existing ones before it is created?
02Vendors and customers
- How many vendors share a tax number or a bank account with another vendor record?
- How many records lack a tax number altogether?
- Who may create a vendor, and who approves it?
- When were inactive vendors last blocked?
- Are customer records duplicated across sales offices?
03Transactions
- Are invoices entered through more than one channel?
- What share of purchase orders are raised after the invoice arrives?
- How consistent are units and currencies across plants?
- Can you trace one item from request to payment without leaving the system?
- How many manual corrections are posted each month, and for what?
04Documents
- Where do contracts, drawings and certificates live, and in how many places?
- What share are scans with no searchable text?
- Are they linked to the vendor, material or asset they concern?
- Who is allowed to see what, and is that written down?
- Is there one current version of each, or several?
05Reading the result
Look for the area where the answers were worst and the money is largest. That is the first project. Fix the data a use depends on, then build the use: clean materials before stock and procurement work, clean vendors before payment checks, readable documents before search.
You do not need perfect data to begin. You need to know which data the first use relies on, and to make that part trustworthy.
In short
- AI on an ERP inherits the ERP's data, good and bad.
- Twenty simple counts across materials, vendors, transactions and documents show where you stand.
- The weakest area with the most money in it is the place to start.
- Fix the data the first use depends on, not everything at once.
Questions
Who should run this checklist?
Someone who can query the ERP, together with the people who own each area: stores, purchasing, finance and document control.
What score is good enough to start?
There is no pass mark. The checklist is for choosing the first piece of work and knowing the risks, not for granting permission.
Does this apply to ERPs other than SAP?
Yes. The questions are about the data and how it is created, which is the same in any ERP.
Can Quantum Beetle run this with us?
Yes. An AI Audit includes a readiness assessment of data, systems and process, and ends with a written plan of what to do first.
Sounds like your problem?
Tell us about it. We'll say honestly whether the swarm can help, and what it would take.
Read next
- How to find duplicate material codes in your SAP material masterMaterial master data
- What sovereign AI means, and when on-premise AI is the right callOn-premise AI
- Duplicate payments: how they happen and how to stop them before the money leavesFinance controls
- What master data management is, in plain words, and why it decides whether AI worksMaster data
- How to clean a material master: a plan that survives contact with a real plantMaterial master data
- On-premise language models or a cloud AI service: how to choose for company dataOn-premise AI