AI in Manufacturing ERP
How to find the AI opportunities in your shop that are actually worth acting on
August 18, 2026
Stop evaluating software. Start by working out where your business is actually losing time.
“We should probably look at AI” is where most manufacturers are right now. It is a reasonable place to be and a difficult one to act from, because it is not a plan and there is no obvious first step.
The instinct is to go looking at software. That is the wrong first move. Software demos tell you what a product does; they do not tell you where your business is losing time. Start there and you end up evaluating features against a problem you have not defined.
Better first move: work out where the time actually goes. That is a half-day exercise, not a procurement process.
Ask a different question
Most software evaluations open with “what features are you looking for?”
The more useful question is “where is your business losing time?”
The difference matters because the first question makes you describe a solution you have not designed yet, while the second one produces evidence. Once you have the evidence, the software conversation gets much shorter — you are matching capability to a ranked list of known problems instead of guessing.
Look for four things
In practice, AI opportunities in a job shop cluster around four signals. Walk your own operation and look for them.
Repetitive work that has quietly accumulated. Not the obviously automatable stuff — the things a skilled person does forty times a week that never made it onto anyone’s improvement list. Copying figures between systems. Re-typing the same email. Rebuilding the same report.
Information that exists but cannot be reached. The answer is in the system. Getting it out takes twenty minutes and three screens, so people either guess or do not ask. Every one of these is a decision being made on worse information than you actually have.
Expertise concentrated in too few heads. The person everyone interrupts. The estimator who knows what that customer’s drawings really mean. This is a business risk before it is an AI opportunity, and it is one of the few places AI genuinely helps rather than just accelerates.
Decisions made on instinct that could be made on evidence. Which supplier to use. Which job to expedite. Which quote to chase. Good instinct beats bad data, but it loses to good data.
Map three factories, not one
A useful framing when you go looking: your shop is running three factories at once.
- The production factory — the physical work. Where are the bottlenecks?
- The information factory — how data moves and gets found. Where does it stall?
- The knowledge factory — where expertise lives. Where is it concentrated?
Most shops have measured the first one and have never looked at the other two. That is usually where the unclaimed time is.
Then rank, honestly
Once you have a list, put every item on two axes: business value and ease of implementation.
The quadrant that matters is high value and easy — and it is almost always bigger than people expect, because the unglamorous items score well on both and get overlooked in favor of the ambitious ones that score badly on the second.
From there a roadmap falls out naturally:
- Quick wins (0–90 days) — high value, low effort, no dependencies
- Operational improvement (3–9 months) — real value, needs some groundwork
- Continuous improvement (9–24 months) — the compounding work, once the habit exists
Ship the quick wins first. Not because they are the biggest prize, but because they are how you build the internal confidence to attempt the rest. A team that has seen AI save real hours on a real job is a completely different audience for the nine-month project.
Sequence beats ambition. The roadmap that gets finished is worth more than the roadmap that is correct.
Who needs to be in the room
Five to eight people, and the mix matters more than the number. Operations, production, engineering, estimating, purchasing, customer service — plus whoever owns IT, and at least one person with the authority to say yes.
Leave out the people who do the work and you will produce a list of opportunities that are not real. Leave out the decision maker and you will produce a list that goes nowhere.
Run it yourself, or run it with us
Nothing above requires us. If you want to work through it internally, the framework is yours — map the four signals, look at all three factories, rank on value and effort, sequence into three tiers.
If you would rather have it facilitated, that is what our AI Opportunity Workshop is. Two to three hours, five to eight of your people, structured across six parts. You leave with an opportunity report, a ranked matrix, a phased roadmap, and an identified set of quick wins.
It is a working session, not a demo. We do not show you the product unless you ask.
The full outline — all six parts, the attendee list, and what you walk away with — is below. If it looks like the right shape for your shop, book a call and we will set one up.
→ Download the workshop outline → Book a call to arrange one
Not ready for a session yet?
Start with the reading. 25 Practical AI Use Cases for Job Shops is the list most shops recognize themselves in.
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.