How do you prepare a manufacturing company for artificial intelligence?

A six-step process

1. Choose a specific task. Not ‘implement AI’, but, for example, preparing draft quotations or finding technical information.
2. Describe the current process. Who initiates the task, what information they need and who checks the result.
3. Verify the data sources. Define which system or document is authoritative and how often it is updated.
4. Add rules and exceptions. Record situations that an experienced person handles differently from a standard case.
5. Set permissions and approvals. AI should receive only the data it needs, and an important output must have a responsible person.
6. Run a limited pilot. Measure accuracy, time saved and the number of cases a person had to correct.

What not to skip

The greatest risk is not that AI will be unable to formulate an answer. The risk is that it will use an outdated price, an incomplete stock position or a rule that does not apply to the particular customer.

For every input, the source, currency and owner must therefore be known. For every output, it must be clear whether it is a proposal or an authorised automatic action.

How large should the first project be?

Small enough to be checked safely, but significant enough for the company to recognise the value. The first pilot is not intended to prove that AI can do everything. It should verify that one specific process works better than before.