Best AI for Furniture Designers: Sketches, Cut Lists and Quotes

For a small furniture maker, use a general chat assistant (ChatGPT, Claude or Gemini, about $20 a month) to turn a brief into a structured part list and a draft quote, and a cut-list optimiser for sheet material. Do the arithmetic in a spreadsheet. AI helps with structure and wording and is unreliable at totals and geometry.

A maker who builds to order has a different problem from an interior designer who specifies pieces from a catalogue: every job needs its own parts list and its own price. One made-to-order oak dining table, taken from a client's brief to a cut list and a priced quote, shows where each tool earns its place and where it needs a check. The price figures and timber rates are illustrative; swap in yours.

Follow me on Instagram@sagnikteaches

The five jobs, and which tool does each

A custom job goes through the same chain every time: capture the brief, make a drawing you can build from, list the parts, buy the material, price the work. AI is useful at some links and a liability at others.

Connect on LinkedInSagnik Bhattacharya

The pattern is clear: AI is strongest at the first and last steps (language), middling at the middle steps (structure) and weakest wherever a wrong millimetre costs a board. The rest of this page follows the table.

Subscribe on YouTube@sagnikteaches

Worked example: one oak dining table, brief to $2,260 quote

A client wants a dining table for six: 1800 mm long, 900 mm wide, 740 mm high, solid oak, a plain square-edged top, four square legs and a simple apron frame, finished in hardwax oil. They want a price and a delivery time. Here is the route.

Step 1: the brief, turned into questions

Paste the client's message into an assistant and ask it what is missing rather than what to build. A prompt that works:

Here is a client's message about a dining table. List the questions I
must answer before I can price it, grouped as: size and use, wood and
finish, structure, delivery and install, timing. Do not suggest a
design. Keep each question to one line.
[paste message]

Back come questions on leg thickness, whether the top may move seasonally (and so how it will be fixed), whether the client wants a particular board width, the finish sheen, access for delivery, and the date they need it. Answer them in one call with the client and you have a specification. This is the cheapest AI step in the workflow and often the most valuable, because most underquoting starts with an unasked question.

Step 2: the parts, with the allowances an assistant forgets

Give the assistant the specification and your construction rules as plain statements: "Aprons are 90 by 22 mm, tenoned into the legs with 30 mm tenons; the top is fixed with figure-eight fasteners; the legs are 70 by 70 mm". Ask for a parts table with finished and rough sizes, using a rough allowance you state (for example plus 4 mm in thickness and 5 mm in width, plus 50 mm in length). The result, once checked and corrected against the drawing, looks like this (illustrative):

A typical first draft does not look like that. It lists the long aprons at 1360 mm, the distance between legs, with no tenons, so they would be 60 mm short. This is the typical error: an assistant calculates the visible length and forgets the hidden joint. Whenever you use AI for a part list, add a line to the prompt, "state the length of each tenon, dowel or housing and add it to the cut length", and check every rail and apron by hand against your drawing.

Step 3: material, with a sum you can check

Do this in a spreadsheet. Multiply each rough part by its quantity to get volume in cubic metres, add a waste allowance, and multiply by your timber price. Illustrative figures:

Add 25 per cent for defects, saw kerf and board shapes that do not match your parts, giving about 0.110 m³. At an assumed $2,800 a cubic metre for kiln-dried oak, timber costs about $307. Hardwax oil, glue, fixings and abrasives add an estimated $65. Whole-board buying will push the real order higher than the net volume, so check your supplier's stock lengths: for the aprons, two 2.4 m boards each yield one long and one short apron (1470 plus 670 plus kerf is about 2150 mm), which is better than you would guess from volume alone.

Step 4: labour and price

Labour is where small makers underquote, because the work is spread over several days. Estimate by operation rather than as a lump, with the assistant prompting you for the steps you forget:

At an illustrative workshop rate of $55 an hour that includes overheads, labour comes to $1,210. Add materials of $372 and the cost is $1,582. To earn a 30 per cent margin on the selling price, divide by 0.7: $1,582 divided by 0.7 is $2,260. A deposit of 40 per cent is $904. Note the difference between markup and margin: a 30 per cent markup on cost would give $2,057, which is a 23 per cent margin. Most underpricing comes from mixing the two up.

Type these figures into the spreadsheet, not the chat. If you want the assistant involved, have it write the formulas or read your sheet and list what could be wrong; the tutorial on why AI is bad at maths and what to check in quotes explains the pattern.

Step 5: the quote, in words the client can act on

Give the assistant the spreadsheet's final figures and ask for a one-page quote: what is included, the finish, the lead time, payment terms, what is not included (delivery, installation) and how long the price holds. Ask for a plain tone and no more than 200 words. Check each number against the sheet and read the exclusions twice. A quote that omits "price holds for 30 days" will be read as holding for ever.

Which assistant for which part of the work

The three main assistants overlap heavily, so choose by habit and by one practical difference: how well each reads a photographed sketch with handwritten dimensions. Test this yourself with one of your own drawings; results vary and change with model updates. Current list prices from vendor pages: ChatGPT Plus $20 a month, Claude Pro $20 a month ($17 on an annual plan) and Google AI Pro $19.99 a month. If you only do this a few times a month, a free tier may do; if you would rather keep client briefs out of training, use a plan whose settings you have checked.

Create a saved project or workspace for your workshop, loaded with your construction rules (standard tenon sizes, rough allowances, finish times, hourly rate, margin target). It saves retyping and it keeps answers consistent from job to job. ChatGPT and Claude both offer Projects. ChatGPT lets you share a project with collaborators even on Plus, while sharing a Claude project needs a Team or Enterprise plan. Gemini has Gems, which Google is turning into skills for personal accounts from 17 November 2026.

For image work, the assistants can produce concept pictures of a table in a room, which clients like. They are not drawings: leg proportions, apron heights and overhangs come out as whatever looks plausible. Use them to talk about mood, not about dimensions. The comparison in ChatGPT versus Gemini for furniture room scenes goes into the picture side in detail.

Cut-list software: when an optimiser beats a spreadsheet

Solid timber is bought by the board, and a spreadsheet handles it well. Sheet material is different: the order in which you place parts on a 2440 by 1220 mm sheet decides whether a job takes one sheet or two. Optimiser software arranges parts to minimise waste, and it respects grain direction and blade width.

Cutlist Optimizer is a browser tool whose free tier allows 5 calculations a day and 5 saved projects; Silver and Gold subscriptions lift those limits and add features such as optimisation priority, and a $4.90 payment gives unlimited calculations for three days. Its options include grain direction and blade thickness, and it exports to PDF, image and CSV. CutList Plus fx is desktop software bought once: Silver $89 (up to 50 parts a project), Gold $249 (unlimited parts, edge banding, and customer pricing and proposals) and Platinum $499 (adds part importing and DXF export). Check both vendors' pages for current terms.

A quick sum shows the value. A bookcase 800 wide, 1800 high and 300 deep from 18 mm board needs two sides (1800 by 300), two top and bottom pieces (764 by 300) and three shelves (764 by 280). The net area is about 2.18 m². A full sheet is 2.98 m², so perfect nesting would use 73 per cent of one sheet, but real jobs rarely nest perfectly. Without an optimiser you would order two sheets to be safe; with a plan that proves everything fits on one, you save a sheet on each such job. At a hypothetical $90 a sheet, ten jobs a year saves $900, which covers Gold several times over.

Ask the assistant to reformat your cut list as a CSV with columns for part name, length, width, quantity and grain, ready to paste into the optimiser. This is a good use of AI: it is reshaping text rather than doing arithmetic.

Reading a photographed sketch: what the assistant sees and misses

Most makers start from paper. A photograph of a sketch with dimensions in pencil goes into any of the three assistants, and the usual ask is "list the parts". Do it, but split it into two prompts, because a single instruction invites the assistant to guess what it cannot read.

Look at this photo of my sketch. First, transcribe every dimension and
label you can read, exactly as written, and mark any you are unsure of
with a question mark. Do not infer sizes that are not written down.
Then stop and wait.

The transcription is the check. In a typical run on a sketch with pencil figures, the assistant reads 1800, 900 and 740 correctly, reads a 70 as 40 where the figure was cramped beside a leg, and skips a dimension that ran off the edge of the page. Correct those, then send the second prompt: "Using these dimensions and my construction rules, list the parts." Splitting the job puts the likely errors in a place where you can see them. It also gives you a record of what the assistant was told, which helps when a number looks wrong later.

Before and after makes the point. The client's message says "a sturdy table about the size of our old one, in oak, maybe with a drawer". After the questions step, the specification reads: 1800 x 900 x 740 mm, solid oak, 30 mm top, 70 mm legs, no drawer, square edge with a 2 mm arris, hardwax oil in matt, delivery to a ground-floor room with a 900 mm door, needed in eight weeks. One is a hope; the other is a job you can price.

Stock-length planning for the legs and the top

The apron plan above works because two parts share a board. Do the same for the rest and you spot where waste hides. The four legs are 730 mm rough; a 2.4 m board of 75 by 75 yields three after kerf (3 x 730 = 2190 plus 2 kerfs), so you need two boards, with the extra leg-length left for a stretcher or a test piece. The six top boards are 1850 mm each: a 2.4 m board gives one and leaves 550 mm, so six whole boards cost you more timber than the net volume says. If your supplier sells random lengths, ask for boards of 2.0 to 2.1 m to cut waste, and note that you want similar colour from one tree, which a yard will often arrange if you ask before the boards are pulled.

Put this into the prompt for the assistant when you ask for a purchase list: "Plan cuts from boards 2400 mm long, 4 mm kerf, and list how many boards of each section I need and what is left over." Check the result by sketching one board. The assistant is often right on the count and wrong on the leftover, because it forgets the second and later kerfs. A purchase list you have verified saves a second trip to the yard.

Mistakes that turn up in real jobs

A few failure modes recur and are worth rehearsing before your first live job.

• Finished sizes ordered as rough. The assistant lists a 30 mm top thickness; you buy 30 mm boards; planing leaves 26 mm. Always label both sizes and order from the rough column.
• Quantities not scaled. Ask for a chair and a table in one go and the assistant may give four legs once and forget the chairs' legs. Ask for each piece separately and join the lists yourself.
• Silent assumptions. Hardwood for the client but softwood in the assistant's example. Put your materials in the project instructions.
• Totals that do not add. Ask the assistant to list line items only, and let a spreadsheet total them. If you do ask it to add, request a second pass that recomputes each line, and still check.
• Leaving out hidden labour. Drawing, ordering, client calls, delivery and finishing time are costs. List them as lines, even short ones, so they reach the price.

When a quote goes quiet, a short reminder email helps, and the playbook in following up silent quotes adapts easily to furniture.

A second job: a made-to-order desk, priced in fifteen minutes

Once the template exists, a smaller job goes quickly. Suppose a client asks for a walnut desk 1400 by 700 mm with one drawer. Paste the brief and your project rules into the assistant, ask for questions, answer them, request the parts table, and move the rough sizes into the spreadsheet you built for the table. Change quantities and sizes; the formulas do the rest. Labour is usually scaled from a similar job, so use the table's 22 hours as a baseline and adjust: the drawer adds about 5 hours, a smaller top saves about 2. The assistant helps by asking what you may have missed (drawer runners, a handle, a cable hole). Fifteen minutes later you have a draft quote. The first run takes an afternoon; the tenth takes a quarter of an hour.

Keep the sheet as a template, one tab per job, so you can compare actual hours against estimated hours when the piece is done. After five jobs, tell the assistant your estimated and actual hours by operation and ask what the gaps show. If sanding always runs over by an hour, the estimate is wrong, not the maker.

What to buy first

Start with no new spend: a free or existing paid assistant, a spreadsheet and the free tier of an optimiser if you cut sheet goods. Spend the first fortnight building the project instructions and the quote template. If you use a 3D modelling tool such as SketchUp, check what its current plans include before assuming AI features come with the one you own; I could not confirm which plan carries them, and plans change. Add paid cut-list software only when the free limits bite, and buy a business plan of a chat assistant only if several people share the workshop's knowledge base.

The biggest gain is rarely a tool. It is a written, repeatable way to price and to specify a job, so the first quote leaves the workshop in an hour rather than a day. For the wider flow from quote to invoice to payment, see the quote-to-cash tutorial.

Furniture maker questions about AI

Can AI turn a hand sketch into a cut list?

Partly. A chat assistant can read a photo of a sketch with dimensions written on it and propose parts, but it often misses joint allowances and thickness changes. Treat its first list as a draft, and check every part against a dimensioned drawing before you cut.

Is cut list software worth paying for?

For solid timber and a handful of jobs a month, a spreadsheet and an assistant are enough. For sheet materials, an optimiser pays back fast because it saves whole sheets. Free tiers exist, so test on a real job before you buy.

Which AI is best at the maths?

None of the chat assistants is dependable at long arithmetic. They are good at structure and poor at totals. Put the numbers in a spreadsheet and use the assistant to write formulas and check the logic, never to add up the final price.

Can I send AI-written quotes straight to clients?

Read every figure first. The assistant is useful for wording, terms and a clear breakdown, but it can drop a line item or repeat a price. A quote is an offer you may have to honour, so the final check is yours.

Further reads

• Best AI Tools for a Small Ecommerce Store: 7 Worth Paying For — Selling finished pieces online: tools worth paying for.
• Turning On-Site Voice Notes Into Job Sheets and Quotes With AI — Turn workshop or site voice notes into job sheets.
• Jewellery Pricing Calculator With AI: Metal, Labour and Markup — The same material, labour and markup logic for another craft.
• Writing Client Updates and Snag Lists With AI for Builders — Client updates and snag lists written with AI.
• AI Estimating Software for Small Builders: Six Options Compared — Estimating tools compared for small building firms.
• AI Quality Control for Small Manufacturers: Where to Start — Quality checks for a small production workshop.
• Cabinet Design Software: What AI Adds for a Small Cabinet Shop — Compare cabinet design software by price band, CNC output and cut-list accuracy, and see what AI renders and drafting really add for a small shop.
• AI Design Software for Clothing: Mood Board to Sample Brief — Move from inspiration to an annotated sample brief, with a practical check on generated flats, colours and ownership.
• AI Furniture Visualisation for Makers: Show a Piece in the Room — A maker-side workflow for a room-photo preview that supports decisions without pretending to be a measured installation drawing.
• Furniture Lead Times: Set Realistic Made-to-Order Dates With AI — How a small furniture workshop works out an honest delivery range from its capacity and queue, and tells customers at the right moments.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Cutlist Optimizer and CutList Plus fx vendor pages (plans and prices checked November 2026); ChatGPT, Claude and Gemini vendor pages for plan prices. Timber prices and labour rates in the worked example are illustrative.