In most job shops and contract manufacturers, the bottleneck is not the machines. It is the office. RFQs wait in the estimator's inbox for a week, purchase orders get typed into JobBOSS or Epicor by hand, and the production schedule lives in a whiteboard and a spreadsheet the plant manager updates on Monday. Every one of those handoffs costs lead time and margin.
InTheory acts as your fractional AI and technology lead. We would start with one expensive recurring workflow, often quoting or order entry, measure how it runs today and improve it inside the ERP you already own where possible. Estimators, engineers and the plant manager still approve every price, routing and ship date that goes to a customer.
The estimator who knows which jobs run on the Haas and which need the five-axis is also the person reading drawings, chasing outside processing prices and answering production questions. RFQs queue up, turnaround stretches past what buyers will wait, and quotes that do go out are rarely followed up.
Customer POs arrive as PDFs and get keyed into the ERP. Material certs, packing slips and inspection reports get filed by hand. Shipping data is re-entered for the carrier and again for the invoice. Each re-key is a chance for a wrong revision, quantity or due date to reach the floor.
A typical small plant runs an ERP, QuickBooks or another accounting package, a CAD system, a quality spreadsheet and a shared drive of work instructions. None of them share data cleanly, so the plant manager rebuilds the same picture of WIP, late orders and capacity every week.
Setup tricks, customer-specific packaging rules and why a part needs a second deburr live in the heads of people who have been there 20 years. As they retire, new operators and inside sales staff ask the same questions repeatedly and mistakes reach customers as quality escapes.
We fix one at a time, starting with the one that costs you most. Each page shows the steps, the tools, what stays human and what we would measure.
Follow up on every open quote at sensible intervals, record why jobs were won or lost and give the estimator a clear list of quotes worth a phone call this week.
See the workflow →Pull part data from RFQs and drawings into a structured estimate with material, routing and outside processing pre-filled, so the estimator reviews and prices rather than starts from a blank sheet.
See the workflow →Extract line items, revisions, quantities and due dates from customer POs, and file certs and packing slips against the right job, so office staff stop retyping PDFs into the ERP.
See the workflow →Connect the ERP, accounting, CRM, shipping and quality records so a job's data moves once, with clear rules about which system is the source of truth.
See the workflow →Sort customer emails about order status, expedites, quality concerns and change requests, draft replies from ERP data and route quality issues to the right person quickly.
See the workflow →See on-time delivery, quote hit rate, backlog, scrap and machine utilization in one place, refreshed from the ERP, instead of a weekly spreadsheet built by hand.
See the workflow →Let operators, schedulers and inside sales find work instructions, customer requirements and past job notes quickly, with answers that cite the controlled document and respect who can see what.
See the workflow →Choose an ERP, MES, quoting tool or quality system with written requirements, scripted demos using your own parts and reference checks, instead of picking whichever vendor called last.
See the workflow →How we deliver this: AI automation for small business or fractional CTO leadership, scoped in writing and run over 90 days. See pricing or the sample scope.
We start with what you already pay for. New tools have to earn their place.
Manufacturers handle customer drawings, specifications and pricing that are often covered by NDAs, and defense work may involve ITAR-controlled technical data or Controlled Unclassified Information subject to NIST SP 800-171 and CMMC requirements. Customer drawings should never be uploaded to public AI tools without checking contracts. Quality records under ISO 9001 or AS9100 also need document control. We would review where data goes for each workflow with your quality manager, but this is not legal or compliance advice.
General information, not legal advice.
Defaults are illustrative for small manufacturers (new production orders won per month). Change them to your numbers.
a month in staff time, before any lost sales
Illustration only, not a forecast or guarantee. Recovered staff time becomes a cash saving only if spending actually declines or the time goes into paid work. Tool costs are an assumption; we replace every number here with yours before recommending a scope.
Usually not. Most small manufacturers use a fraction of their ERP's features, and the quickest gains often come from cleaning up item masters, routings and customer records so the data you already have is trustworthy. A new ERP is a large project and sometimes the right call, but it should be decided on clear requirements, not on hope.
Only in tools whose terms and settings you have checked, and only where your customer contracts allow it. Defense and aerospace work may involve ITAR or CUI that cannot go to most cloud AI services at all. We would review this with your quality and contracts lead before any drawing or spec touches a new tool.
It varies. For some shops it is RFQ turnaround, because faster quotes win more work. For others it is PO entry errors that cause wrong revisions or late shipments. We would look at your quote log, order history and late shipment reasons in the first weeks before recommending a pilot.
Typically the owner or general manager, the estimator or engineering lead, the person who enters orders and the plant manager or scheduler. For a 90-day pilot we would ask for a few hours a week from the people closest to the workflow, mainly to review drafts and test with real jobs.
No, but it fits established shops with steady RFQ volume, an ERP or job tracking system in place and a measurable bottleneck. A very small shop running on QuickBooks and spreadsheets may get more from a few process changes than from a paid engagement, and we would say so on the first call.
In 30 minutes we look at one recurring workflow, what it costs you today and whether there is a credible case to improve it. If there is not, we will say so.
30 minutes with Phil or Martin. No technical brief, no pitch deck. Phil or Martin replies personally, usually within one business day.