How do you choose the first worthwhile use of AI in a manufacturing company?
Choose the first use of AI based on its value to the company, the reliability of the available inputs, the scope for human review and the limited impact of a potential error. A suitable pilot addresses one repeated task, has a clear owner and produces a measurable result. It should not begin with autonomous decisions about prices, capacity or customer commitments.
Four conditions for a good pilot
- A recurring problem - the task happens often enough for the saving to matter.
- Available inputs - the required data and knowledge exist, and their currency can be verified.
- Human review - a person can quickly spot an error and approve the result.
- Measurable value - time, error rate, response speed or another outcome can be compared before and after the pilot.
Suitable areas to start with
- finding information in verified technical documentation,
- classifying and extracting information from incoming documents,
- drafting a response to a standard enquiry,
- preparing the supporting information for a quotation,
- flagging missing or conflicting information.
In every case, AI should initially prepare a draft or alert. A responsible person then reviews the result and makes the decision.
What to avoid at the outset
Do not choose a process with unclear data, a large number of undocumented exceptions or a high potential impact if something goes wrong. An autonomous change to a production plan or a binding delivery date promised to a customer is not a suitable first test.
The pilot should give the company evidence that a particular task can be done better. It should not be a demonstration of as many features as possible.
Your next step
The Digital Business Review compares potential use cases and selects the one with the best balance of value, readiness and risk.
Book a free 30-minute call →