Automation for manufacturers
Job cards, quality records, maintenance logs, customer reporting.
The paperwork that follows every part through the shop.
Find out where you stand · about 5 minutes · no sales call attached
The reality on the ground
You haven't got a system. You've got fifteen years of decisions.
An MRP that was right when you bought it, customised so heavily it can't be upgraded, from a supplier who's since been acquired twice. Accounts in something separate. Quality in something else again. A machine interface running on a PC nobody's allowed to touch.
None of them talk to each other. The integration layer is a person with a keyboard.
The workarounds became the process.
Somewhere there's a spreadsheet that was a temporary fix in 2017 and is now business-critical. It lives on one desktop. It has macros in it written by someone who left in 2021.
There are steps in your process that exist because of a customer complaint nobody remembers. There's double entry happening because two departments don't quite trust each other's numbers, and both are a bit right.
The real communication channel isn't on the org chart.
It's a WhatsApp group, a whiteboard in the planning office, and a verbal handover between shifts.
Which works, until the person who was on the call is off sick. Sales promises a date because sales can't see the schedule. Production changes a run and finds out it mattered when the customer rings.
Decisions live in email threads. Nobody can find them.
And the reporting describes what happened, not why.
Month-end takes a fortnight. Three people have three versions of the numbers and all three are defensible. The MD's own spreadsheet is the actual dashboard, because it's the only thing that answers the question being asked.
The data exists. Getting it out is the problem.
None of this is a failure of management.
It's what happens when a business survives fifteen years by solving whatever was most urgent that month. Every one of those decisions was reasonable at the time.
And it doesn't get fixed for the same reason it built up: everyone is flat out running it. The people who understand it well enough to change it are the same people holding this week together.
Nothing here is broken enough to stop the job going out.
That's exactly why it never gets fixed.
Where the hours go
Production scheduling
Planned in a spreadsheet, communicated verbally, adjusted on a whiteboard. Nobody outside the room knows the current state without asking.
Job cards and works orders
Printed, marked up by hand on the floor, typed back into the system afterwards. Sometimes days afterwards.
Quality records and NCRs
Captured on paper at inspection, filed, entered into the quality system by an administrator. Trend analysis is theoretically possible and practically never happens.
Maintenance and calibration
A schedule someone maintains manually. Breakdowns recorded in a book. No link between the two, so nothing predicts anything.
ISO documentation
Version control by filename. Procedure_v4_FINAL_updated_March.docx
Goods-in and stock
Delivery notes checked against orders by hand. Stock figures that are accurate on the day of the count and drifting from the day after.
Customer-specific reporting
Each significant customer wants their own format. The same underlying data, reformatted repeatedly by someone who could be doing something else.
What we'd look at first
Capture at the point of work.
The single biggest win in most manufacturing businesses is removing the paper-then-retype cycle. Data entered once, on the floor, on a tablet or scanner, flowing straight into the system.
Connect scheduling to reality.
If the schedule and the shop floor disagree, everything downstream is guesswork. Getting live job status visible is often a small build with a disproportionate effect.
Turn quality records into quality data.
Once NCRs are structured rather than filed, patterns become visible. This is where AI genuinely helps — spotting the correlation across two years of records that no human was ever going to see.
Get the knowledge out of heads.
Where a process only exists because one person knows it, that's a business risk with a retirement date attached. Documenting it is unglamorous and it's usually the most valuable thing we do.
The regional picture
Manufacturing accounts for 15% of employee jobs in Hull and East Yorkshire — more than double the national share, and HEY LSIP 2.0 names it as one of four Growth Driving sectors, alongside its status as a UK Industrial Strategy priority.
The plan reports hard-to-fill vacancies in technical roles, limited local supply of engineering and manufacturing skills, an ageing workforce, and cost barriers to specialist training.
It also notes that while manufacturing is projected to contract in net job numbers, replacement demand will continue to generate vacancies — because people are leaving the workforce faster than they're being replaced.
Read it at heylsip.com
The honest reading: you will not solve this by recruiting. The people aren't there. The only remaining lever is getting more out of the hours you already have.
We're not going to tell you to rip out your ERP.
Most of what we do connects what you've already got, rather than replacing it. Replacement is expensive, disruptive and usually unnecessary.
If a replacement genuinely is the right answer, we'll say so — and we don't sell software, so we've no reason to say it unless it's true.
Start with five minutes
The Zeus Readiness Map asks 16 questions about how your operation actually runs, and gives you a written plan at the end — whether or not you ever work with us.
hello@zeus-ai.co.uk