· Dark Oak · 7 min read
A factory MD in the Midlands said it to us on a call last month: “Our problem isn’t capacity, it’s paperwork.” He runs a 60-person engineering shop, books are full into Q3, and he’s spending Saturday mornings tidying up quotes and invoice queries because his office team can’t get to them during the week.
We hear a version of that sentence almost every week. We’re a small Sligo-based engineering studio, and most of our first calls with Irish manufacturers — typically owner-managed shops between 10 and 150 staff — start the same way. Capacity on the shop floor is fine. The constraint is in the office, in the in-tray, in the spreadsheet that someone updates on a Friday afternoon and nobody else fully trusts.
This is a survey of the five hours-eaters we keep meeting. None of them are exotic. All of them quietly cost a small manufacturer one or two full-time-equivalent staff a year if left alone. For each, we’ll describe what it actually looks like, sketch the options ranked roughly by cost, and say which one we typically recommend.
Quote generation from RFQs
An RFQ (Request For Quote) lands in a generic inbox, usually with a PDF drawing pack and a covering email. Someone — often the owner, sometimes an estimator — opens it, reads the spec, pulls material prices from a supplier portal or an old email, looks up a similar job from last year for a sanity check, applies margin rules from memory, and types the result into a Word template. Maybe an hour per quote, more for anything with fabricated parts. Multiply by ten or twenty quotes a week.
The bottleneck is rarely the pricing logic itself. It’s the lookups, the formatting, and the fact that the estimator is the only person who knows where everything lives.
- Cheap configuration: a shared quoting template in Google Sheets or Excel with linked material price tables and a margin calculator. A weekend of work. Solves the formatting problem but not the lookups.
- Low-code workflow: a tool like Make or n8n to pull RFQs out of the inbox, extract drawing references, and pre-populate a quote draft with current material prices from a supplier feed. Two to four weeks. Works well when your suppliers expose any kind of API or scheduled price file.
- Custom agent build: a small AI agent that reads the drawing pack, identifies the parts and materials, looks up live prices, and drafts a quote for the estimator to review. Four to eight weeks. Worth it when quote volume is high and drawings are mostly in one or two formats.
We usually recommend the low-code workflow first. The agent build is genuinely useful, but most shops we meet haven’t yet structured their supplier prices anywhere a machine can read them — and that’s the cheaper problem to fix.
BOM compilation and revision control
The BOM (Bill of Materials) is the master list of what goes into a product. In most of the shops we visit, the engineering team maintains it in a spreadsheet because the spreadsheet is fast and flexible. The ERP (Enterprise Resource Planning) system holds a different version of the same BOM because purchasing and finance need it there. When the engineer revises the spreadsheet, somebody — often the same engineer, sometimes a purchasing clerk — re-keys the changes into the ERP. Sometimes a week later. Sometimes never.
The time cost is real, but the bigger cost is the rework when the two BOMs drift. We’ve seen shops order the wrong revision of a part because purchasing pulled from an ERP record that hadn’t been updated since January.
- Cheap configuration: tighten the change process — a single shared spreadsheet, dated revisions, a Monday morning review where purchasing confirms the latest version. Free, and surprisingly effective if the team will actually do it.
- Low-code workflow: a sync that pushes BOM changes from the engineering spreadsheet into the ERP on a schedule, with a review queue for anything ambiguous. Two to three weeks if your ERP has a usable API.
- Custom build: a proper BOM management layer with revision history, approvals, and ERP sync. Six to twelve weeks. Heavy.
For a 30-person shop, we almost always recommend the low-code sync. The custom build is the right answer for shops with regulated products or multiple sites — and there the cost is justified by audit requirements, not just time saved.
Stock reconciliation across physical, ERP, and supplier portal
This is the one nobody enjoys. Components arrive on different POs (Purchase Orders) at different times. The warehouse marks them received on paper. Someone updates the ERP. The supplier portal shows yet another view because it’s tracking what’s been shipped, not what’s been booked in. By Friday afternoon, somebody is in a spreadsheet trying to make all three views agree before the weekly stock report goes to the MD.
What makes this hard isn’t the maths. It’s the exceptions — the part that arrived short, the substitution the supplier made without telling anyone, the GRN (Goods Received Note) that was filled in for the wrong PO.
- Cheap configuration: a single weekly reconciliation template that pulls from the ERP and a CSV export from the supplier portal, with a clear column for exceptions. A few days to set up. The work still happens, but it’s faster and the trail is clearer.
- Low-code workflow: scheduled pulls from the ERP and supplier portals into a single reconciliation dashboard, with automatic flagging of mismatches. Three to four weeks.
- Custom agent build: an agent that reconciles continuously, raises exceptions in a queue, and drafts the queries to suppliers. Six to ten weeks.
We usually recommend the low-code workflow here, with one caveat: it only works if your supplier portals actually expose data in a usable form. About half do. For the rest, you’re stuck with the configuration option until the supplier modernises.
Shift handover and operator notes
Operators write handover notes, scrap reports, and downtime causes into a notebook on the line. The notebooks get walked to the office on Monday morning, where someone — often the production manager — types the highlights into a spreadsheet or a report for the weekly meeting. Bits get lost. Handwriting gets misread. The operator’s actual insight (“the new batch of steel is harder than the spec sheet says”) gets reduced to a number in a downtime column.
This is the one where we’re most cautious about AI. The operator’s judgement is the valuable thing, not the paperwork around it.
- Cheap configuration: a tablet on each line with a simple structured form. Removes the transcription step. Two to three weeks to roll out, mostly change management rather than technical work.
- Low-code workflow: the same tablet form, plus a weekly summary that aggregates downtime causes and flags anything unusual. Adds a week or two.
- Custom agent build: an agent that reads free-text operator notes and produces a structured weekly summary, surfacing patterns the production manager might miss. Useful, but only after you’ve got the digital capture in place.
Recommendation: do the tablet form first, run it for three months, then decide if the summarisation layer is worth adding. Most shops find the tablet alone gives them 80% of the benefit.
Inbound supplier invoice processing
The classic 3-way match: the PO says what was ordered, the GRN says what arrived, the invoice says what’s being charged. When all three agree, the invoice gets paid. When they don’t — and they often don’t, because of part substitutions, short deliveries, or price changes — somebody has to investigate. That somebody is usually in the office on a Friday afternoon, and they’re never finished.
- Cheap configuration: a clearer exception log and a standing 30 minutes on Tuesday and Thursday to clear mismatches. Removes the Friday backlog but doesn’t reduce the work.
- Low-code workflow: OCR (Optical Character Recognition) on incoming invoices, automatic matching to PO and GRN, exception queue for the finance person to review. Three to five weeks. This is one of the highest-ROI low-code builds we do.
- Custom agent build: an agent that handles the match, drafts the supplier query for exceptions, and posts approved invoices to the accounts package. Six to eight weeks.
We almost always recommend the low-code workflow. Invoice matching is a well-trodden problem and the tooling is mature. The custom agent is worth it once volume is over a few hundred invoices a month.
What we usually recommend first
Nobody we meet needs a factory of the future platform. They need one specific Friday afternoon back.
Our consistent advice: pick the single hours-eater on this list that costs the most labour per week in your shop, and start there. Not the most interesting one, not the one with the best demo. The most expensive one. For most shops we talk to it’s either the invoicing or the quoting, but it varies.
We run a fixed-price Readiness Sprint at €1,500 to scope exactly that. Two weeks, we walk the process with your team, measure where the hours actually go, and come back with a costed proposal for a build — usually in the €4k to €8k range for the kind of low-code workflows described above. EU data residency, no surprise billing, no platform lock-in.
If you’ve recognised your own Friday afternoon in any of the above, get in touch via /#contact and we’ll set up a short call to see if there’s a fit.