Glasgow has one of the densest concentrations of working craftspeople in the UK, spread across studio blocks in Trongate, Kelvinbridge, and the old industrial units out toward Dennistoun. Over the past year, the conversation in these spaces has shifted from whether to use AI tools to how, if at all, they fit into a day that still runs on kiln schedules, loom tensions, and client deadlines. We spent several weeks visiting workshops to ask a simple question: has artificial intelligence actually changed how you work, or is it just noise?

What we found was neither blanket adoption nor blanket rejection. Five makers, working in ceramics, textiles, furniture, jewellery, and print, gave us five different answers. None of them are using AI to make the actual object. All of them have thought harder about it than most office workers who've never touched clay or wood.

Why We Went Looking in Glasgow Workshops

Glasgow's craft scene sits at an interesting junction. The city has strong art school lineage through Glasgow School of Art, a tight network of maker collectives such as Many Studios and WASPS, and a customer base that includes both tourists wanting something "authentically Scottish" and locals who commission bespoke pieces. That mix means craftspeople here are unusually exposed to both ends of the AI conversation: buyers asking "did a computer make this?" and suppliers pushing AI-powered design software as a time saver.

We chose five workshops based on recommendations from two maker collectives, aiming for a spread of trades and business sizes, from a one-person jewellery bench to a furniture workshop employing four people. We asked each maker the same set of questions: what AI tools, if any, they'd tried, what stuck, what they dropped, and what they'd tell another craftsperson considering the same path.

The answers below are not meant as a verdict on AI in craft generally. They're a snapshot of five specific decisions, made by five specific people, in one city, during one particular stretch of 2024 and early 2025.

Case One: A Ceramicist's Research Shortcut

Fiona, who runs a small ceramics studio near Shawlands, makes functional stoneware: mugs, bowls, serving dishes, sold through a weekend market stall and a handful of local shops. Her AI use is narrow and specific: she uses a chatbot to speed up glaze chemistry research.

Before she had a baby in 2023 and had to cut studio hours, she'd spend evenings reading glaze calculation forums to work out oxide ratios for a particular celadon effect. Now she describes the result she wants and the clay body she's using, and asks the tool to suggest a starting recipe and explain the chemistry behind why certain oxides behave as they do at cone 6 versus cone 10. She still tests everything in test tiles, usually six to eight firings before she trusts a glaze on actual stock, because the suggested ratios are frequently off and sometimes unsafe if taken at face value.

"It's replaced about two hours of forum-reading a week, not the testing," she told us. She was clear that she wouldn't trust a glaze recipe from any source, AI or human, without firing test tiles first. Anyone mixing glazes should know that some oxide combinations produce toxic fumes or unstable glazes, and a professional ceramics tutor or supplier should be consulted before working with unfamiliar materials.

Case Two: A Textile Maker's Client Mockups

Lauren makes cushions, wall hangings, and small runs of printed fabric from a shared studio in Trongate. Her business depends heavily on custom commissions, which means a lot of back-and-forth with clients before a single stitch happens. That back-and-forth is where she's brought AI in, specifically image generation tools, to produce quick visual mockups of pattern ideas.

Previously, she'd sketch three or four pattern directions by hand, scan them, and email them over, a process that could take half a day per client. Now she generates eight to ten rough visual directions in under an hour, using them purely as a conversation starter: "does the client want busier or calmer, warmer or cooler." None of the generated images are used as final designs. Once a direction is agreed, she redraws and develops the actual repeat pattern herself, by hand or in Illustrator, because the AI-generated images don't translate into usable textile repeats and often contain small structural errors a trained eye catches immediately.

She estimated this has cut her mockup-to-approval time from around five days to two for a typical client. The risk she flagged: some early AI-generated mockups included motifs that resembled copyrighted characters or recognisable brand patterns without her prompting for them, which she now checks for carefully before showing anything to a client.

Case Three: A Furniture Maker Who Opted Out Entirely

Not every workshop we visited has adopted anything. Callum runs a furniture workshop out of a converted warehouse near Dennistoun, building commissioned tables, chairs, and storage pieces in oak and ash. He and his three employees have made a deliberate decision not to use AI tools anywhere in the business, including for admin or marketing copy.

His reasoning wasn't anti-technology in a general sense. He uses CAD software daily and a CNC router for some joinery work. His objection was specific to generative AI: he's concerned about the environmental cost of large AI models at a point when his own business markets itself partly on sustainable, low-impact sourcing, and he didn't want a mismatch between his marketing claims and his actual practices.

He also pointed to a practical reason. Furniture commissions depend on trust built through conversation, and he worried that AI-written emails or AI-generated renderings would start to feel generic to the kind of client who commissions a £3,000 dining table precisely because they want something that doesn't feel mass-produced. His workshop still writes every client email by hand and photographs pieces with a camera rather than generating marketing images.

It's worth noting his business is larger and more established than the others we spoke to, with an eight-month waiting list, which gives him more room to make a choice other makers might not be able to afford.

Case Four: A Jeweller Using AI for Pattern Ideas Only

Priya works from a bench in a shared jewellery studio near Hillhead, making fine silver and gold pieces, often with engraved or textured surface patterns. Her use of AI sits in a tightly bounded space: generating rough visual references for engraving and texturing patterns, nothing else.

She described her process: she'll ask an image generation tool for variations on a motif, say, "botanical line pattern inspired by thistle seed heads, suitable for engraving on a narrow band," and look through a batch of 20 or so results for one or two interesting structural ideas. She never uses the output directly. Instead she redraws the chosen idea by hand at the scale and material thickness she's actually working with, because metal engraving requires line weights and depths that generated images don't account for at all.

Her reason for drawing such a firm line: engraving and hand-piercing require a physical understanding of how a line will behave when cut into metal, something no image generator factors in. She also raised a client-facing concern. Two clients had specifically asked, unprompted, whether any part of their commissioned piece involved AI, and she wanted an honest answer ready. "I tell them straight: the idea seed might have come from a tool, the design and every cut is mine," she said.

Case Five: A Printmaker's Mixed Approach

The fifth workshop is actually a small co-operative of three printmakers sharing a studio and a Vandercook proof press near Partick, producing limited-edition screen prints and linocuts. Their use of AI is the most varied of the five, split across different tasks by different members, which made it a useful case for seeing how one workshop can hold several different attitudes at once.

One member uses an AI tool to generate quick colour palette variations before committing to a run, testing maybe 15 combinations digitally rather than mixing 15 batches of ink, which she estimated saves a half-day of ink mixing and test-pulling per edition. Another uses it occasionally for layout composition, arranging where text sits relative to an image, treating it like a faster version of moving tracing paper around. The third member doesn't use it at all and described feeling "mildly pressured" by the other two before they agreed, as a group, not to make it a studio-wide requirement.

What the print studio has not done: none of them use AI to generate the actual imagery that ends up on a plate or screen. All three were firm that the hand-cut linocut or hand-prepared screen is the entire point of the work, both for their own satisfaction and because their collectors specifically buy limited editions knowing each print involved physical labour at every stage. One member mentioned a buyer who asked to see the original lino block before purchasing, specifically to confirm it had been hand-cut.

What the Five Cases Have in Common

Across workshops with almost nothing else in common, trade, scale, customer base, three patterns kept showing up.

• AI touches research and planning, not the object itself. Every maker who used AI at all drew a hard line at the point where material meets tool, whether that's kiln, needle, chisel, or burin.
• Speed gains were modest and specific, not transformative. Hours saved ranged from roughly two to five hours a week, concentrated in research or mockup stages, not in the core making time.
• Client transparency came up unprompted in three of five conversations. Makers who use AI anywhere in their process have started preparing honest answers for clients who ask directly, rather than waiting to be caught out.

The furniture maker's opt-out and the printmaking co-op's internal disagreement both point to something less tidy than a trend: there's no consensus among Glasgow craftspeople, and several told us they expect their own position to keep shifting as tools change.

Common Mistakes

A few missteps came up repeatedly, either from the makers' own early experiments or from stories they'd heard from peers.

• Trusting an AI-generated glaze, dye, or chemical recipe without testing it physically first, which risks unstable or unsafe results.
• Using AI-generated imagery as a final design rather than a starting reference, which often produces pattern or motif results too close to existing copyrighted work.
• Not having a straight answer ready when a client asks whether AI was involved, which several makers said damaged trust more than the AI use itself.
• Assuming a tool that works for one material or trade will transfer cleanly to another, when glaze chemistry, textile repeats, and metal engraving each have different technical demands AI tools don't understand.

Practical Next Steps for Makers Considering This

If you run a craft business and you're weighing whether any of this applies to you, a few concrete steps came up across our conversations.

1. Pick one narrow task, research, mockups, or palette testing, rather than trying AI across your whole workflow at once.
2. Set a personal rule before you start about what stays fully hand-made, and write it down somewhere you'll actually stick to it.
3. Test anything material-related, glazes, dyes, finishes, physically before trusting it, and consult a materials specialist or trade tutor if you're unsure about safety.
4. Decide in advance how you'll answer a client who asks directly about AI involvement, since several makers found this question came up sooner than expected.
5. Talk to other makers in your trade before committing to a tool. The printmaking co-op's internal disagreement suggests this isn't something to decide alone if you share a studio.