AI for a Laser Cutting Business: Price by Material, Time and Waste

A laser cutting shop can use AI to organise a quote from sheet cost, parts per sheet, measured cutting time, setup and finishing labour, and a defined allowance for unusable offcuts. Give it your own rate card and nesting result, ask it to show the arithmetic, then compare its estimate with the machine log before sending a price.

The machine is not the pricing source by itself. A part can have a short cut path yet consume most of an expensive sheet, while a small nested part can require many pierces and long handling. Keep material, time and waste separate so a low quote can be traced to the assumption that caused it.

Follow me on Instagram@sagnikteaches

Make a material card for each sheet you actually buy

Record material type, supplier specification, thickness, sheet width and length, paid price per sheet, freight allocation, usable margins, grain or face direction, minimum permitted gap, storage condition and last purchase date. Use the invoice and the physical sheet, not an AI estimate of “typical” prices. If two suppliers sell nominally similar plywood with different cores, their cut quality and time may differ; keep separate records until shop tests show they can share a rate.

Connect on LinkedInSagnik Bhattacharya

The material card also states what the laser may safely process under the machine maker's instructions and your shop procedure. Do not accept an unidentified plastic because a customer says it “looks like acrylic”. Start from the machine maker's materials list: Trotec's cutting basics, for instance, list acrylic, polycarbonate and several other plastics, but the only foam they list is marked “PVC free”. Material identification, extraction and safe settings are operational checks. A quote can be provisional pending confirmation, but a model should never decide from a photograph that a material is suitable for cutting.

Subscribe on YouTube@sagnikteaches

Suppose an illustrative 600 × 400 mm birch plywood sheet costs $18 delivered to the shop. A design uses 90 × 60 mm rectangular blanks. Pure area division gives 240,000 ÷ 5,400, or about 44 blanks, but ignores edges, kerf, gaps and orientation. A tested nesting file might yield 35 good blanks. At 35, the raw sheet allocation is $18 ÷ 35 = about $0.51 per blank before cut time and handling. Use the actual nest and useful output, not the area-only maximum.

Keep a separate card for offcuts. Record material, thickness, usable rectangle, date, condition and location. A remnant has value only when the shop can identify and retrieve it for a later job. If a box of scraps cannot be matched to a material or is too warped to use, it is waste for pricing purposes. Trotec's laser cutting basics describe a kerf barely larger than the beam itself, and its project guides suggest cutting a small test shape with your intended settings and measuring it to find the kerf. Even so, your own gap, hold-down and quality requirements determine the real yield.

Use an intake rule for customer-supplied material. The shop may not know its exact composition, flatness or behaviour under the approved machine settings. The customer may expect the quote to exclude material, but the shop still incurs testing, setup and the risk of a failed piece. State what evidence is required, whether a test coupon is available, and who bears the cost of a damaged supplied blank under the agreed terms. If the material cannot be identified safely, decline the cut rather than asking AI for a plausible setting.

Check whether the quoted thickness is the nominal catalogue value or a measured sheet. Warped or variable stock can change focus, fit and reject rate. In an illustrative quote, a client describes “3 mm plywood” while the delivered sheet measures unevenly and requires another test. The material card should link to the specific stock lot used for the estimate. A one-line note such as “price subject to approved material sample” is more honest than pretending the drawing fixes every production variable.

Price the nest, not just the outline

Ask for the customer's final geometry or a clear drawing with units, quantities, material and finish. Check whether the file has duplicate paths, open contours, hairline marks, tiny internal details or text that is actually engraving rather than cutting. AI can help prepare an intake checklist and identify inconsistencies in a file summary, but machine software and an operator should inspect the real paths. A cut length without a valid file is a provisional number.

For each nest, capture sheet count, good parts per sheet, cutting path length, number of pierces or starts, estimated machine time, test pieces, expected rejected parts and remnant status. Different geometries with equal total path length may run differently because of pierces, acceleration or multiple passes. Keep “machine estimate” and “actual logged time” in distinct columns. When they diverge, update the quoting factor by material and job type instead of forcing the invoice to match the original guess.

A small sign business requests 70 identical plywood tags. The tested nest yields 35 per sheet, so two sheets are needed. The shop reserves one extra sheet only if its own reject history or the customer's deadline warrants it; do not blindly add a full sheet to every quote. If two test tags can be cut in a spare corner before production, the test cost might be lower. The quote sheet should make that choice visible. A hidden “miscellaneous” line cannot tell the owner whether material yield or risk drove the price.

For a different job, a client asks for five large panels that each occupy 70% of a sheet. Area-based allocation suggests 3.5 sheets, but the geometry permits only one panel per sheet. The material commitment is five sheets, subject to any reusable remnants. This is the main reason to inspect nesting output rather than divide total cut area by sheet area. AI can explain the result to the client, but it should not invent a denser nest to win the order.

Compare at least two approved nest orientations when the job is material-heavy. If rotating a part by 90 degrees increases output from 30 to 35 per sheet, check whether grain, printed face, bend direction or visible pattern allows it. For 70 tags, both nests still need at least two sheets, but the 35-per-sheet layout leaves no spare parts while the 30-per-sheet layout would require a third sheet. The correct choice therefore depends on quality and reject allowance as well as raw yield. Record why the chosen orientation was accepted.

Watch for tiny internal cutouts. A design with 20 small holes may have less total cut length than a plain outline but more pierces, smoke marks and manual cleanup. Quote those operations explicitly when your log shows they matter. Ask the customer whether the tiny features are functional or decorative before redesigning them. AI can draft the clarification: “The file has 20 internal cutouts under the agreed minimum feature size; would you accept a revised pattern for test cutting?” It should not silently alter the geometry.

Set a shop rate from real logged capacity

A machine hourly rate is a business decision. List recurring machine-related costs, maintenance, extraction, consumables, utilities and the capital recovery approach your accounts support. Divide by realistic billable machine hours, not every hour the workshop is open. Keep setup and finishing labour visible if the operator is occupied while the machine runs. A shop that uses one blended rate can still quote accurately if its historical data supports that choice; the model should not choose a rate from a generic web answer.

Suppose the illustrative shop uses $48 per logged machine hour, $24 per operator hour and a $12 minimum setup allocation. The two-sheet tag job takes 18 machine minutes per sheet and 20 minutes of operator setup plus 15 minutes of finishing. Machine cost is 36 ÷ 60 × $48 = $28.80. Operator time is 35 ÷ 60 × $24 = $14.00. Add $36 for two sheets and $12 for the setup allocation: $90.80 before a reject allowance, selling fees and the desired profit. These are assumed internal figures, not vendor prices.

If the shop's approved waste and reject allowance for this tested pattern is $4.20, the estimated job cost becomes $95.00. At 70 good tags, cost is about $1.36 each. A selling quote can add the shop's chosen margin and any delivery or design work, itemised so the customer understands what is included. If the customer requests 50 tags instead, still inspect the nest: two sheets may remain necessary even though the quantity fell. Cost per tag could therefore rise.

A second time example: machine software predicts 30 minutes but the actual run logs 42 because the material needs a slower verified setting. At $48 per hour, the difference is 12 ÷ 60 × $48 = $9.60 for that run. Do not blame the model's prose. Correct the time factor for that material and thickness, then review open quotes that used the old assumption. The next quote should cite the measured log range and a current test if the geometry differs.

Keep queue time out of the machine-cost calculation unless the customer is paying for a reserved capacity or rush service under a stated rule. A job can take 40 minutes of cutting but wait two days for a slot; multiplying two days by the hourly machine rate would be absurd. Conversely, an urgent order can displace a planned run and create a genuine scheduling cost. Price that using a consistent shop policy and explain the deadline commitment separately from the physical cutting time.

If the operator must stay beside the machine for this material, include their time during the run. If a tested process allows them to do other productive work, avoid double-counting the same hour in two jobs. In a small shop, setup, monitoring, unloading and cleaning often overlap; record each stage once. An illustrative time sheet might show 10 minutes setup, 36 minutes machine run with 12 minutes attended, and 25 minutes sorting and packing. The operator total is 47 minutes, not 71, if the attended time is already part of the run.

Give offcuts a rule the team can apply

Choose a minimum reusable size for each material and record a remnant only after it is labelled and stored. A 200 × 300 mm clear acrylic remnant may be useful for small signs; a narrow strip of the same area may not fit any planned product. Value a remnant conservatively based on jobs that actually use such sizes. Do not subtract its full retail sheet price from today's job. Otherwise the current quote may be too low and a later quote may also claim the material was free.

Illustratively, a $30 sheet produces one large panel and an offcut that the shop can reliably use for small nameplates. The pricing rule may allocate $24 to the panel job and carry $6 into labelled remnant stock. If the offcut is damaged or remains unused beyond the shop's review period, write off the $6 in the business records and revise the rule. The arithmetic is internal; the customer sees the agreed product price, not an unsupported promise that offcuts always have resale value.

Run a monthly remnant audit. Count pieces labelled as reusable, pieces actually drawn from storage and pieces discarded. If the shop credits $6 per large-panel job but only one in four remnants gets used, the credit is optimistic. Revise the allocation using observed reuse, or set a narrower criterion for what enters remnant stock. A digital list without a shelf location is not useful; the operator should be able to retrieve the exact piece without searching through anonymous scraps.

A third offcut example is grain direction. A plywood remnant is physically large enough for a later part, but its face grain runs the wrong way for that customer's design. Its recorded area overstates usable stock. Add orientation to the remnant record. A fourth is protective film removed from acrylic, leaving scratches that exclude it from a premium finish. A fifth is a sheet with a clamp zone that the nesting preview counts as free area. These are concrete reasons to compare a digital nest with real material.

Let AI draft the quote explanation, then recalculate it

Provide the model a row of approved numbers, the formula and any uncertain assumptions. Ask it for an itemised internal calculation and a short customer explanation. Keep exact arithmetic in a spreadsheet and reconcile each number. Do not paste confidential customer artwork into an external AI tool without the required permission and your own data controls. The guide to AI arithmetic checks is useful when a fluent answer makes a wrong sum hard to notice.

Draft an internal laser cutting cost breakdown and a plain customer
quote note. Use only these approved values: sheet price [ ], sheets [ ],
machine minutes [ ], machine hourly rate [ ], operator minutes [ ],
operator rate [ ], setup [ ], waste allowance [ ], delivery [ ].
Show every multiplication and addition. Flag missing inputs.
Do not invent a selling margin or claim a finish not approved below.

An illustrative bad output says “70 tags at $1.10 each; total $77” despite the supplied $95 cost before profit. The correction is not a nicer customer sentence. Recalculate from the sheet and identify the missing machine or material line. Another output may state that all offcuts will be reused. Replace that with the actual remnant rule and avoid promising a waste-free process. A human approves the final quote, particularly when it contains an unusual minimum charge or a client's supplied material.

Use a quote version number. If the customer changes thickness from 3 mm to 6 mm, the old cut time, material card and nest may no longer apply. If they add engraving, capture it as a separate operation rather than burying it in cut length. If quantity rises enough to change sheets per part, rerun the nest. The engraving quote and proof workflow covers a different kind of change: customer text and position approval.

Close the loop with the machine log

After each job, record planned and actual sheets, good and rejected parts, estimated and actual machine time, setup and finishing time, offcuts stored, and the reason for rework. Review the difference by material, thickness and geometry. A single slow job may be an exception; repeated underestimates indicate the rate card needs a new factor. The log should include machine downtime separately, so a breakdown does not quietly become the normal cutting time for the part.

For the illustrative tag job, suppose the actual run uses two sheets, 43 machine minutes, 38 operator minutes and yields 68 good tags because two are scorched. The planned quote promised 70. The first action is to make the missing two from available material and honour the agreed order terms; the costing review follows. At $48 per machine hour, 43 minutes costs $34.40, $5.60 above the 36-minute estimate. Operator time costs $15.20, $1.20 above estimate. The yield change is another cost, not a number to conceal.

Ask what caused the difference. Were the cuts slower because the plywood batch differed? Did the file contain duplicate paths? Was the operator's 15-minute finishing estimate too short? Did the extra parts require another setup? A model can group notes and suggest fields to investigate, but the operator and log establish the cause. Update the material card or file intake accordingly. For a general workshop quoting process, see custom job quote controls.

Before the next batch of quotes, test one known job using the revised sheet and compare the result with its historical log. If the model's explanation and the spreadsheet agree, the nest fits the actual sheet, and the operator accepts the time assumption, the draft is ready for approval. A price is credible because its inputs can be traced and changed, not because an AI answer looks precise to two decimal places.

Use a short quote sign-off that names the material card version, nest file version, machine estimate, log comparison, allowance for rejects, and reviewer. For repeat orders, confirm that the customer has not changed the artwork and that the stock and machine settings still match the original. A repeat button is safe only when those conditions hold. If a file changes by one hole or a sheet comes from a new supplier, rerun the relevant check. That small discipline makes later price revisions explainable to both the owner and the customer.

Questions about laser job estimates

Can AI calculate cutting time from a drawing alone?

It can help organise the geometry, but the quote should use your machine software's estimate and measured logs for the same material, thickness and settings. Travel, pierces, acceleration, focus, multiple passes and handling can change real time. Treat any AI-only time as an unverified assumption and test the job or a representative sample before promising a price.

Should every offcut be charged as waste?

No. Define a reuse rule based on material, minimum useful dimensions, identification and storage. Charge the current job for unusable area and its fair share of the purchased sheet. A reusable remnant may carry a recorded value into a later job, but do not credit it twice. If offcuts rarely get reused, the historical evidence should change the pricing assumption.

Is the machine hourly rate the same as the operator's pay?

No. A shop rate may need to recover machine ownership, maintenance, consumables, power, extraction, space and realistic utilisation as well as labour where appropriate. Keep operator setup and finishing time separate from unattended machine time when those costs differ. Use your own accounts and logs to set rates; a generic AI figure is not a defensible shop rate.

Further reads

• Joinery AI: Build a Cutting List and Material Order From a Drawing — Compare material yield planning in another trade.
• Can AI Price a Job Accurately? What Tradespeople Should Check — Understand where general AI pricing estimates fail.
• How to Compare Supplier Quotes Side by Side With AI — Check material purchase prices before updating a rate card.
• Predictive Maintenance With AI: Realistic for a Small Workshop? — Consider downtime when setting machine capacity.
• How to Set Up an Approval Step for AI-Written Quotes — Make a human quote approval gate explicit.
• Countertop Fabricator Software: Templating, Slabs and Where AI Helps — A vendor question list for templating, slab layout, quotes and cut files, with a scorecard, a yield sum and where ChatGPT helps with quote wording.
• How to Use AI in a Machine Shop: Five Jobs Beyond Quoting — RFQ triage, work instructions and inspection paperwork costed in hours and dollars, plus scheduling notes and customer updates for a 12-person shop.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Trotec Laser, laser cutting basics FAQ (kerf and materials table) and a project guide on measuring kerf with a test cut (read 5 November 2026). Cost figures are illustrative shop inputs, not vendor prices.