AI in Manufacturing ERP
How Job Shops Can Use AI Without Starting a Big Project
October 5, 2026
Most of the value sits in the work around the machines: finding old quotes, chasing job status, building the report for Monday’s meeting. Here is where to start.
Start with the work around production, not production itself. Job shops get the quickest return from AI when it finds information, drafts routine paperwork and flags problems early, so estimators, planners and buyers spend their day on decisions instead of searching. You do not need a big project or perfect data to begin. You need one repetitive task that annoys people, and the will to fix it.
This post pulls the main ideas from our updated playbook, The AI-Powered Job Shop. The full version, with all seven chapters and a 90-day roadmap, is a free download.
Every job shop runs three factories
Walk the floor and you see the first one: machines, material, operators, jobs moving from one work center to the next. This is where improvement money has gone for decades, into setup times, throughput and machine utilization. Most shops are good at it.
The second is the administration factory. Looking up an old quote. Chasing job status. Answering “where’s my order?” Re-keying data from one system into another. Building the report for the production meeting. None of it makes a part, and it takes up a large share of your best people’s week.
The third is the knowledge factory, and it lives in people’s heads. The estimator who knows which jobs made money. The machinist who solved this exact problem five years ago. The buyer who knows which supplier always ships late. That knowledge leaves the business one retirement at a time.
Where AI fits: Your machines and your people already run the first factory well. AI does its most useful work in the other two.
Where AI helps first: follow one order
Take a single order from inquiry to delivery and there is time to win back at almost every stage. The questions below are the kind a team would ask in plain English. They are illustrative examples, not a list of specific product features.
1. Quoting “Have we made anything similar before?” The estimator starts from past jobs, prices, drawings and actual costs instead of a blank sheet. Quotes go out sooner and pricing gets more consistent.
2. Releasing the job “Show me every engineering change on this part.” Revisions, bills of material and routings come back in one place, so the job reaches the floor with the right information attached.
3. Scheduling “What needs my attention today?” A morning summary lists the jobs at risk, the material shortages and the late operations. The planner’s first hour goes on fixing problems, not finding them.
4. Purchasing “Which supplier deliveries could hold up production this week?” The buyer works through the exceptions instead of checking every open line by hand.
5. Customer calls “Where is job 48217?” The current operation, expected finish and any risk come back while the customer is still on the line. Nobody has to walk the floor to find out.
6. Running the business “Why did on-time delivery drop last month?” The owner gets the bottlenecks, supplier delays and capacity issues behind the number, and the Monday meeting is about what to do next.
None of this asks anyone to become an AI expert. People keep doing the jobs they know. The searching and re-keying around those jobs is what shrinks.
Three myths that keep shops waiting
“We’re too small.” Smaller shops often have the most to gain. When the estimator also looks after customers and the production manager also does the purchasing, every hour of admin you remove goes straight back into the business.
“Our data isn’t good enough.” Nobody’s data is perfect. You already hold years of quotes, job costs, purchase orders and drawings, so start with what you have and clean it up as you go. One honest caveat: if actual hours were never booked against jobs, AI cannot invent them. Some shops find that gap in the first month, and fixing it pays off with or without AI.
“AI will replace our people.” It takes over searching, summarizing and report building. Judgment, experience and customer relationships stay with your team.
The playbook covers two more: the fear that it will be another disruptive ERP project, and the belief that you need a full AI strategy before you start.
A 90-day way to start
Days 1 to 30: find the friction
Skip “where can we use AI?” Ask where people search for information, which reports take hours to build, and what knowledge sits with only one or two people. Rank what you find by business impact, not technical difficulty.
Days 31 to 60: take the quick wins
Daily production summaries, order status lookups, quote history and document search. None of these change how people work. They just remove effort. A good sign at this stage is someone asking, “Can it help with this too?”
Days 61 to 90: make it a habit
Expand what worked, capture more of the know-how, automate the routine workflows, and start measuring. Quote turnaround, hours of admin removed and how fast customer questions get answered are better measures than how many features you switched on.
One rule runs through all three phases: the business owns this, not IT. Operations, engineering, purchasing and customer service know where the friction is. Technology follows them.
Where Tangle fits
Tangle is the world’s first self-customizing ERP for manufacturers, with Milo wired into every customer’s instance. Milo is Tangle’s built-in AI engineer. You describe a change in plain English, Milo builds it, and you preview it before it goes live.
Illustrative example: A production manager asks for a workflow that alerts production, customer service and sales whenever a job runs more than two days late. Milo builds it, the manager reviews and approves it, and there is no IT ticket in the queue.
Running a legacy ERP you are not ready to replace? Tangle Xray puts AI on top of the system you already have. It is a limited release for shops selected on fit, and you can join the waiting list.
Common questions
Is my job shop too small to use AI?
No. Small teams where people wear several hats often see the biggest difference, because every hour taken off admin goes back into quoting, production or customers.
Do I need clean data before I start with AI?
No. Start with the quotes, job history and purchase records you already have, and improve data quality as you go. Where key data was never captured, such as actual hours on jobs, fix that first.
Can AI work with my existing ERP?
It can. Tangle Xray adds AI on top of an existing ERP and is available as a limited release to shops selected on fit. If you plan to replace your ERP, AI is part of the core Tangle product, through Milo and the built-in AI assistant.
Where should a job shop start with AI?
With one repetitive task that frustrates people. Quote lookups, job status questions and the weekly production report are common first choices.
Get the full playbook
All seven chapters of The AI-Powered Job Shop, including the three hidden factories, a day with Milo and the 90-day roadmap.
- Download the playbook (PDF)
- Take the AI Readiness Assessment
- Want ideas you can use this month? Read 25 Practical AI Use Cases for Job Shops.
By Tangle Software Inc. Tangle is the world’s first self-customizing ERP for manufacturers, with Milo wired into every customer’s instance. tangle.io.