Specialty and engineer-to-order (ETO) manufacturers run on thin project margins, disconnected data, and estimating processes that live outside the ERP. In a recent webinar, Appficiency’s Paul McDonagh and Eric Ribeiro teamed up with Shayan Rastgou, VP of Engineering at AskCipher, to break down where AI actually fits in a specialty manufacturing tech stack, and where it doesn’t.
Here are the five biggest takeaways. Watch the full session on demand:
Start With What You Already Own
Before buying another tool, ask what your current ERP already does that nobody switched on. Most manufacturers run their stack at half capacity. The panel’s framework breaks every AI decision into four paths: leverage (use what you have), integrate (connect a focused tool to a specific gap), buy (only if nothing you own solves it), or build (only for genuinely differentiated use cases). Technology decisions come after the business case, never before.
Data Readiness Decides What’s Even Possible
AI doesn’t fix messy or disconnected data; it just makes the consequences more visible, faster. The panel also flagged a data privacy risk most leadership teams underestimate: employees may already be pasting company data into free consumer chatbots without any policy in place. The choice between a third-party AI tool and a private, controlled deployment should be made before any use case is scoped.
Three Use Cases, Real-Time Savings
Estimating
AI reads the engineering takeoff, quantifies it, and generates the job cost budget and customer-facing estimate from the same source, instead of building each one separately. Manual process ran 4-6 hours; estimators now spend that time on pricing judgment instead of data entry.
Order and project automation
Once a customer signs the estimate, the project needs a set of structured records built out based on its phases and structure. AI identifies what needs to be created, validates for duplicates and inconsistencies, and builds it directly in NetSuite. A manual process that took up to 4 hours now takes about 15 minutes.
Controlled change order distribution
When a change order is approved, AI finds every record it impacts, prepares the required updates, and holds them until a human approves execution. What took roughly 3.5 hours per change now takes about 10 minutes.
In every case, a person still reviews and approves. AI removes the repetitive data entry, not the judgment calls.
AI Adoption is a Change Management Problem, Not a Tech Problem
“Every department can have its own AI use case, but no department should have its own AI strategy.” Estimating feeds job costing, job costing feeds WIP reporting, WIP reporting touches finance and IT. Siloed tools break at every handoff. Scaling AI requires shared governance, a defined pilot with a baseline and an end date, and buy-in from the people actually doing the work, not just a top-down mandate.
Use Four Questions Before Funding Anything
Where does AI fit in what you already have? What business problem are you solving? How will you implement and govern it? What’s the next decision to validate? If you can’t answer “what” and “next,” the use case probably isn’t ready for funding yet, and that’s a useful, cheap answer to get to early.
Frequently Asked Questions
Do we need a full system overhaul before using AI?
No. Assess how much of your current system you're actually using and how clean your data is first. Small, focused AI initiatives can create real value before any system replacement is on the table.
Will AI replace our team?
No. AI handles repetitive, time-consuming tasks; people still own judgment calls, reviews, and decisions. It's a tool your team gets trained on, the same way new software or equipment requires onboarding.
Where does AI plug into estimating, order setup, and change orders without disrupting job costing?
Before the budget or project records exist, not after. For estimating, AI reads the takeoff once and generates the budget and customer estimate from the same source, removing the double entry that causes reconciliation errors. For order setup and change orders, it works the same way: AI builds or updates the records automatically once a human approves the trigger, instead of someone keying them in by hand.
What's included in an AI Blitz Assessment?
A focused review of your current systems, where AI capability already exists and isn't being used, your top operational and finance pain points, and a data readiness check, leading to a practical plan for where to start.
Ready to see where AI fits in your operation?
Watch the full webinar recording here
