Notes · AI and data on the shop floor
The data rule that scales a small company
- Lead and inquiry
- Costing
- Sample and Pre-BOM
- Sourcing and materials
- Production and maintenance
- Management reports
A small company runs on what a few people know. That is a strength, until the business grows faster than anyone’s memory.
The problem that success creates
When we rebuilt our website to be found properly, the number of inbound requests rose by an order of magnitude. A good problem, and one that exposed how much we were typing the same information in several places: a name in an email, again in a spreadsheet, again in a quotation, again in a sample request. Each copy was a chance to disagree with the others.
One rule that governs the rest
We adopted a single rule: each piece of information is entered once, then reused and tracked. It sounds modest. It changes how you choose tools, because the question stops being “what does this software do?” and becomes “where does this fact live, and who reads it from there?”
What it looks like in practice
- Shared internal systems, so the data is not on one person’s laptop.
- An equipment register where each machine has a QR code and a maintenance record, scanned on the floor and read by the office.
- A lead workflow where AI helps capture and analyse the data, so people spend their time on judgement and not on re-typing.
- A clear path from a customer inquiry to an approved sample, with the same record following it all the way.
Two principles worth defending
- AI processes as much of the data as possible; a named human approves every decision step. The assistant does the groundwork. A person judges and is accountable.
- Structure should free the people who carry the know-how, not replace them. In a small factory, the know-how sits in people. A good system makes it easier to share and harder to lose.
Honest status
This is a pilot. The real test is full adoption by the team and measured benefits on complete projects, and we say so. A rule that is only followed by the person who wrote it is not yet a rule.
Questions this note answers
- What does 'enter it once' mean in practice?
- Each piece of information, such as a customer inquiry, a machine record or a sample request, has one place where it is entered. Everything else reads from it, so nobody retypes it and every change is tracked.
- Can AI make decisions in a manufacturing business?
- AI can process and analyse as much of the data as possible, but a named human approves every decision step. Structure should free the people who carry the know-how, not replace them.
Faced the same question, or see it differently? I like comparing notes. Get in touch →