Anyone who has ordered a banner or a vehicle wrap knows the problem: the only copy of the logo is a small image saved from a website, and it needs to be printed metres wide. Galt Signs & Graphics, a sign shop in Cambridge, Ont., didn’t buy a software subscription to deal with it. It built its own tools and put them online for anyone to use, free.

The shop now runs two web tools: Upscaler.ca, which uses AI to enlarge small images for print, and Vectorizer.ca, which turns pixel images into editable vector files for printers, vinyl cutters and lasers. Neither needs an account, and both process images on your own computer rather than uploading them, according to the sites.

They’re small examples of a bigger shift. Canadian businesses are adopting AI quickly, mostly by subscribing to tools someone else built. A few are taking the next step and building the tools themselves, because the pieces that used to require a software company are now free, open and small enough to run in a web browser.

Adoption has tripled. The small-business gap hasn’t closed

In the second quarter of 2026, 19.2% of Canadian businesses said they had used AI to produce goods or deliver services in the previous 12 months, Statistics Canada reported in June. That’s triple the 6.1% of two years earlier.

The averages hide a split. Businesses in information and cultural industries (42.3%) and finance and insurance (40.4%) lead. Construction (9.2%), wholesale trade (7.9%) and agriculture (4.5%) trail. Among the smallest firms, with one to four employees, 19.9% used AI, but 41.4% said it isn’t relevant to their business at all. Rural businesses (9.9%) are less than half as likely to use it as urban ones (21.0%).

The Business Development Bank of Canada sees the same pattern from a different angle. Nearly all small and medium-sized businesses invest in digital technology (96% in 2025), but only 30% use generative AI, its 2026 digital transformation study found. The ones that do are 24% more productive, BDC says. “Canada is not behind on adopting technology—but we are behind on extracting its full value,” said BDC chief economist Pierre Cléroux.

What a sign shop built

Galt Signs & Graphics is an independent shop in downtown Galt, Cambridge, that designs, prints and installs vehicle wraps, fleet graphics and office, window and wall graphics, according to its website. Its two tools target problems print shops deal with constantly.

Upscaler.ca enlarges logos, artwork and photos up to 16 times, to as much as 30,000 pixels a side. Instead of choosing a pixel size, you enter the print size and DPI you need, and the tool works out the rest and rates how sharp the result will be. It offers different AI models for logos and text, flat clip art and photos, plus a face-restoration option for portraits, and saves PNG, JPG or TIFF files.

Vectorizer.ca traces images into vector paths, keeping every colour or reducing them to as few as two, and lets you click any area to delete or recolour it. It exports to the formats that sign and maker equipment actually uses: layered SVG, CMYK PDF and EPS for print, DXF and HP-GL for vinyl cutters, and G-code for lasers and pen plotters. Small or blurry images are sharpened with AI before tracing, the site says, so the traced shapes come out cleaner.

The shop also publishes a free library of 481 Ontario regulatory road sign drawings with their MTO codes on its resources page.

Why this is suddenly possible

Three things make a project like this realistic for a business that isn’t in the software industry.

The AI models are open. Both tools list their models on the page, and they are open-source projects anyone can use under permissive licences, including Real-ESRGAN (BSD-3-Clause) and GFPGAN (Apache-2.0), plus upscaling models by Philip Hofmann. A small business doesn’t have to train anything; it has to choose, combine and package models that already exist.

Your computer does the work. Upscaler.ca runs the AI on the visitor’s own graphics card, in the browser, and the models work offline once downloaded. That means no servers to pay for as usage grows, and no customer files to store or protect. It also helps with the barrier businesses cite most: in the same Statistics Canada survey, cybersecurity or privacy concerns (13.4%) topped the list of things limiting AI use, ahead of cost (10.6%).

The skills are closer to home than they used to be. Galt’s owner, Anmar, writes on the company’s site that he has “had a passion for technology since I was a kid,” and left the sign industry for a while to study information technology, then worked as a signage broker before opening the shop. Plenty of small businesses have someone like that: an owner or employee with a technical background and a clear view of the problem.

What’s in it for the business

Every page of both tools says who made it, and each one’s FAQ answers the obvious follow-up: can the shop print or install the result for you? The shop also lists both tools on its own resources page.

Our analysis: A free tool is marketing people actually use. For a local business, being the one that solved someone’s problem, before they ever asked for a quote, may be worth more than another ad. It can also fix the business’s own bottlenecks: artwork that arrives too small or in the wrong format costs print shops time, and a tool that solves it helps the shop and its customers alike.

The limits

  • Your hardware matters. Because the AI runs locally, results on an older laptop can be slow, and the tool’s most detailed photo mode is meant for computers with a graphics card.
  • AI can’t invent what isn’t there. Upscaling adds plausible detail, not the original detail. For brand logos, a proper vector file from whoever designed the logo is still the best source, and AI-traced vectors still need a human check before they go to print.
  • Licences still apply. Open-source models come with terms. Builders need to read them; Galt’s tools credit each model and its licence on the page.

What other small businesses can take from it

The lesson isn’t that every bakery should write software. It’s that the cost of building a small, specific tool has dropped far enough that the question is worth asking. If customers keep hitting the same problem before they can buy from you, and open models or free platforms can solve it, a tool you build can solve it for them and put your name in front of new customers at the same time.

For the 41% of the smallest businesses that told Statistics Canada AI isn’t relevant to them, that may be the more useful way to think about it.