The Real Knowledge Base · from our own projects
Which business data needs to be ready before you use AI?
Not everything at once. Choose one task and see which data it uses. For each data item, set down which version takes precedence and who keeps it up to date. Also agree who decides when information is missing or contradictory, and what AI may do with the data. The rest can wait.
Why you start with the task
Tidying everything up first takes a lot of time before anything works. Every company has an old list or a duplicate file somewhere. Giving everything to AI causes the opposite problem. Contradictory versions then make it unclear which information should apply. That is why you start with the task. It determines which data needs to be correct now, and you set down in advance which version the system uses.
Which data a quotation uses
Take a quotation. Choose which request, product data, prices and terms take precedence. Agree who assesses missing or contradictory information. A request without a size or with an old price then goes to a person first.
At an SME with two locations, a request now produces a quotation in the house style. It comes with an accompanying email, in two languages. Order confirmations and invoices there come from one brand source, so one change works through everywhere.
Three agreements per task
- Version to use. For each data item, everyone knows which version counts. Remove outdated prices and discontinued products.
- An owner. Someone keeps it up to date and knows when it changes.
- Use. Which customer data may AI use, and may it train on it? Set that down within the technical and legal conditions of your company.
The data does not all need to move to one new system for this. A system can draw on what is already running.
Who the data belongs to
Agree what a supplier does with your data, and what it learns from it. The software we build ourselves as TID runs in your own environment. What we learn from your designs and target audience data, we use only for you. The trade-off between renting and having software built is covered in Rent AI software or have it built?
Where it goes wrong
- Wanting to tidy everything up first. Then it takes a long time before anything works.
- Giving everything to AI. Then the system may work with old and duplicate versions.
- Data without an owner. Then it is correct at the start, and not later.
- No agreement on cases where the information is unclear. Then the system decides for itself which version counts.
What this means for you
You can only tell whether your data is usable by assessing it for a real task. Which task comes first depends on your processes and on what is already running.
Frequently asked questions
Does our data first need to move to one new system?
No. A system can draw on what is already running. Revenue, margin, outstanding items and stock then come together in one overview, from your existing systems.
How do you know whether your data is ready for AI?
A version that takes precedence, an owner and agreements on cases where the information is unclear are preconditions. Whether your data is usable also depends on the task.
May AI work with customer data?
That depends on the technical and legal conditions, and on what you agree with customers and suppliers. Set down what AI uses and whether it may train on it. How we handle your data is agreed per client.
Sources and accountability
- This article describes our own way of working, as set out on our TID and CID pages.
- The quotation example comes from a TID project at an SME whose name we do not disclose.
- It shows what the system delivers.
- That source does not say which data was checked beforehand.
- Our advice on these trade-offs is based on this way of working.
- These sources give no measured result for how much time organising data takes. That is why there is no figure here.
Are you unsure whether your data is usable for AI?
In the future-proof scan we look at what is ready and what is holding you back.
Request the future-proof scan