Is AI worthwhile in a manufacturing company, or is it still just hype?

AI makes sense in a manufacturing company when the company has its data in order and its processes documented. Without those foundations, applying AI to a fragmented environment will accelerate the chaos. I use a simple rule: if the company does not currently have one reliable source for its key indicators, it is not ready for AI, however good the proposition may sound.

Why so many manufacturing AI projects disappoint

AI learns from its inputs. When those inputs are inconsistent, the output will be confidently wrong. That is worse than having no output at all, because people will trust an incorrect figure in an attractive interface for a while.

A typical sequence looks like this. A company deploys a demand forecasting or predictive maintenance tool. The model receives data that is produced manually and several days late. The forecasts are inaccurate, people stop trusting them and the project is quietly switched off within a year.

AI did not fail. The data did.

Where AI already works in manufacturing today

Where machine-generated data is reliable:

Visual quality control. A camera plus a model: the input is an image that is created automatically. It does not depend on whether someone has recorded something.
Predictive maintenance based on sensors. Data from the machine, not from a form.
Document processing. Orders, invoices and delivery notes. The input is clearly defined.
Searching company documents. This replaces searching through shared drives.

The same applies in every case: the data is created automatically and is consistent.

Where AI will not help until something else is fixed

Production scheduling when completion reporting is delayed
Demand forecasting when historical data is incomplete
Reporting when departments do not share the same definitions of indicators
Anything intended to replace a decision that no one can describe today

A quick readiness test

Answer four questions:

Is there one source for every key indicator that everyone agrees on?
Is the data created automatically, or does someone enter it manually?
How old is it when the decision is made?
Can you describe the process that AI is meant to support clearly enough for a new employee to understand it?

Four clear ‘yes’ answers mean you can begin. Two or fewer mean that, for now, investing in getting your data in order will deliver a better return than investing in AI.

What I think

I do not consider AI to be hype or a cure-all. I see it as an amplifier: it amplifies a well-designed system and a poorly designed one alike, just in the direction you do not want.