Estimates out while the drawing is still open.
AI helps a metal fabrication shop most in the paperwork around the metal, not the metal itself. Quoting from drawings and RFQs, submittals, job status emails, chasing POs, and reporting are the work we automate, at a fixed $5,000 to $15,000 per system, built on accounts you own. We are a one-person firm in Charleston, SC, run by a licensed P.E. who reads the same drawings you quote from.
Where the hours actually go
Nobody opens a fab shop to write emails, but that is where the week goes. An RFQ lands with a drawing package and someone senior has to read it, do the takeoff, build the quote, and then chase it. A customer wants to know where their job is, and the answer means walking the floor and then writing it up. POs sit unconfirmed. Month end means an evening of pulling numbers into a report nobody enjoys building.
That work is real and it has to happen, but almost none of it needs the person currently doing it. Most of it follows rules you could dictate in an afternoon. Work that follows rules is work software can do, with a person checking the parts that matter.
What we automate in a fab shop
Every shop is different, so we start by mapping yours. But the same jobs come up again and again:
- RFQ intake. Read the incoming package, pull out scope, quantities, and due dates, and flag what is missing before your estimator opens it.
- Quote assembly and follow-up. Draft the quote from your pricing logic for a human to approve, then follow up until it gets a yes, a no, or a real reason.
- Submittals. Assemble the package in your format, track what has been sent, and nag about what has not come back.
- Job status. Answer the where-is-my-order emails from your job data instead of pulling someone off the floor.
- PO chasing and back-office paperwork. Confirmations, acknowledgments, and the filing that piles up.
- Reporting. Open quotes, jobs in progress, and aging paperwork in one plain report that shows up on schedule.
A P.E. is a different conversation
O'Donnell AI is one person. Jake O'Donnell is a licensed professional engineer who built and operated industrial gas systems, including a commercial CO2 recovery plant. A marketing agency selling AI is selling software it has never lived with. An engineer who has run a plant is selling something he trusts his own operation to.
I read drawings for a living before I automated anything. I know the difference between a quote you can draft from a rule and a quote that needs an estimator's judgment, and we will tell you which is which instead of pretending the machine can do both.
The discipline is not theoretical. Starting in February 2026, I built my own automation fleet in about 135 days. It runs 70 scheduled jobs with more than 10,000 logged runs, and the failure rate came down from 23.7% in May 2026 to 2.0% in June 2026. The telemetry is published publicly, so you can check the claim instead of taking my word for it.
Pricing and ownership
The AI Ops Audit is $2,500 flat and takes about two weeks. You get a written map of the automatable work in your shop with time savings put in dollars, and we implement one quick win before the report lands, so you get value even if we never speak again. The first three clients pay $1,000, openly because we are building our public case-study bench.
Automation builds run $5,000 to $15,000 per system, fixed scope and fixed price. Managed operations, if you want us to keep running things, is $1,500 to $2,000 per month, month to month, with no long-term contract. An estimate with a wrong takeoff gets caught before it goes out. Every run is logged and checked against rules written for your shop, and the monthly report says in plain English what got done.
You own everything. The accounts, the keys, the data, and the systems we build. AI usage costs pass through at cost with no markup. If you leave, it all stays yours and keeps working.
Questions we hear
Can AI actually quote from our drawings?
Parts of it, and we will be straight about which parts. It can read an RFQ package, pull scope and quantities, flag missing information, and assemble a draft quote from pricing rules you define. The judgment calls stay with your estimator, who approves everything before it goes out. The audit maps exactly where that line sits in your shop.
What does this cost?
The audit is $2,500 flat, and the first three clients pay $1,000 because we are building our public case-study bench. Builds are $5,000 to $15,000 per system at a fixed price. Managed operations runs $1,500 to $2,000 per month with no long-term contract. Plant and industrial engagements are quoted individually.
Do you actually know manufacturing, or is this generic AI consulting?
The founder is a licensed professional engineer who built and operated industrial gas systems, including a commercial CO2 recovery plant. That does not make us fabricators, and we will not pretend it does. It does mean we read drawings, understand submittals, and know what it feels like when paperwork stands between a shop and getting paid.
What happens if we stop working with you?
Everything keeps working and everything stays yours. We build on accounts you own, with your keys and your data, and AI usage costs pass through at cost. Managed operations is month to month, so leaving is as simple as not renewing.
How do I know the automations will not quietly break?
Quietly breaking is the failure mode we built the whole practice around. Every managed run is logged and validated against checks specific to your business, and you get a plain-English monthly report of what the AI did. Our own fleet of 70 scheduled jobs has more than 10,000 logged runs, with the failure rate brought from 23.7% in May 2026 to 2.0% in June 2026, and that telemetry is published publicly.
Start with one bad quote
Email us the last RFQ that ate an afternoon and we will tell you which parts of it a machine could have handled, no charge. Questions are welcome too, even if you are nowhere near ready to buy anything.