A customer moved an order up by three days. It clashes with two others. Planner to decide.
Early North, for manufacturers
Every factory has people who spend their day chasing.
Changed orders, late materials, missing papers, price rises nobody passed on. Early North builds AI that reads what your team chases, flags what needs a decision and drafts the next step. Your people approve everything.
The steel surcharge rose 4%. Price revisions drafted for three customers. Commercial head to approve.
Thursday’s resin delivery isn’t confirmed. The supplier was asked at 14:06. Buyer to follow up if no reply.
02 / A NORMAL WEEK
Six moments from a normal week in a factory that makes to order.
Each moment costs you money, and none of it shows up until the month’s numbers come in. Turn a card to see what would catch it in time, and what your team still decides.
Customer planner
"We’re pulling next week’s order in by three days. Please confirm today."
The email sits unread until Wednesday. You pay overtime, and another customer’s order goes out late.
Customer purchasing
"This shipment has the old revision. Our order said revision C."
The new drawing came on WhatsApp and never reached the shop floor. The whole batch comes back.
Shift supervisor
"The material hasn’t come. Line two stops after lunch."
The supplier slipped two days ago and nobody noticed. People and machines stand idle.
Customer quality
"We need the inspection report for today’s shipment, tonight."
Your best engineer is typing reports at eight at night.
Customer accounts
"Payment released, less deductions as per our debit note."
The payment is short and nobody can say why for weeks. A debit note is the customer’s list of what they’ve cut from your bill.
Your finance head
"Raw material’s up again. Did we ever claim last quarter’s increase?"
Your costs went up. Your prices didn’t. That margin rarely comes back.
Most factories have three or four of these. The factory review shows which ones cost you most, so you fix those first.
03 / BETWEEN THE DESKS
Every department owns its part. Nobody owns the gaps between them.
Sales owns the quote and accounts owns the invoice. But the change that arrives on WhatsApp, the late delivery and the short payment fall between desks. Your ERP (the system that runs orders, stock and invoices) records what happened. It doesn’t chase what’s about to go wrong.
The relay line connects Sales, Planning, Purchase, Stores, Production, Quality, Dispatch and Accounts in that order. Four gaps can lose information: the schedule change between Sales and Planning, the late delivery between Purchase and Stores, the paperwork between Production and Quality, and the short payment between Quality and Dispatch.
04 / FAILURE MODES
How AI goes wrong in a factory, and how we stop it.
Quality engineers list failure modes before a part goes into production. We do the same for every workflow.
| Failure mode | What it looks like | How we stop it |
|---|---|---|
| Automating a broken process | The old mess, only faster | We map how the work really runs first, and fix the process where needed |
| A wrong reading | A drawing number or revision is misread | Anything uncertain goes to a person, and nothing leaves the factory without approval |
| Knowledge only in someone’s head | The workflow misses what your planner just knows | We sit with that person and build their rules in |
| The floor stops trusting it | People quietly work around it | We design with the people who’ll use it, and show them what changes before launch |
| A pilot that never goes live | An impressive demo, then nothing | Every build has a go live date and is measured against the problem it was meant to fix |
| Nobody owns it afterwards | It drifts and breaks | We run it and report monthly on what it caught |
We won’t build anything whose purpose is cutting jobs. AI should take the chasing off your people’s desks, so they have more time for the work only they can do.
05 / WHY NOW
Why this is practical now, and wasn’t three years ago.
Software can now read messy paperwork
Orders, drawings and emails reach you in every format. Until recently, software couldn’t read them reliably, so a person had to.
The cost of running capable AI fell about 280 times between late 2022 and late 2024. A check that once needed a person can now run on every order.
Source: Stanford HAI, AI Index Report 2025
The people who catch these slips are hard to replace
In most factories, one experienced planner or buyer spots these problems from memory.
In the US alone, manufacturers may need 3.8 million more people by 2033, and about half of those jobs could go unfilled. When that person is away or leaves, the checks go with them.
Source: Deloitte and The Manufacturing Institute, April 2024
Neither is a reason to rush. Both are reasons to find out what your gaps cost, so you can decide with numbers.
06 / FIT CHECKLIST
It’s worth talking if you can tick five or more of these.
You've ticked 0 of 10.
Fewer than five usually means the gain would be small. That's worth knowing too.
Book a free call07 / ONE FACTORY
Illustrative factory, not a client
A quarter in one factory.
In mid April, the factory’s main raw material went up 12%.
Its biggest customer kept paying the old price for months.
Nobody noticed until the owner asked one question.
Read what happened next
08 / WHERE TO STARTStart with a conversation.
A free 30 minute call about how your factory runs and where things get stuck. If there's a fit, we'll suggest a next step. If not, we'll say so.
The call is free, with no obligation.