4,235 fields to fill. 18 minutes to a first draft.

Then I took a dive to make sure everything was acceptable.

From my desk today

Yesterday, I had to send product listing information to a retailer.

77 products. Each one needs at least 55 fields to define. That's 4,235 entries.

Plus 3 to 8 images per product, formatted to the retailer's specs and uploaded to a public server.

Not so long ago, I would have said: "Here we go for a few days of copy-paste!"

Earlier this year, I evaluated several AI agents. For this type of work, I adopted Claude.

So here's what I did instead:

  • I entered the retailer's product spec and image requirements.
  • I described the type of customers who shop at this retailer.
  • I pointed the agent to our existing product specs and information.
  • I wrote a prompt explaining how to fill the listing sheets, pick the images that best fit those customers, and adjust titles and descriptions for them, with the right SEO information.

Then I launched the agent.

18 minutes later, I had a first draft of the listing sheet and a first set of formatted images.

And that's when the second step started.

The retailer reality

If you've ever loaded products into a retailer's system or your own online store, you know this work.

Every retailer has its own template. Its own required fields. Its own image sizes.

So the same product gets typed, resized and uploaded again and again, one channel at a time.

It's slow. It's boring. And when people get tired, mistakes slip in.

And your listings matter more every year. 30% of Canadian shoppers now use AI shopping tools, up from 11% a year earlier. And 51% are open to letting AI handle the whole purchase, checkout included, once they've set their preferences. Those AI tools read your listings.

Source: Adyen, 2026 Retail Report (Canadian consumers), as reported by Retail Insider, January 2026.

That's exactly the kind of work AI is good at. But only if you stay in charge of what goes in, and what comes out.

Behind the scenes at Randmar

Here's yesterday's job, by the numbers.

4,235 fields, 18 minutes to a first draft
77
products
4,235
fields to fill (55+ per product)
3 to 8
images per product, formatted to spec
18 min
to a first draft
9
mistakes caught in my review

Source: Philippe Jeanneau, product listing for a retailer, October 2026.

What it means for you: the first draft is no longer the bottleneck. The review is.

And we're not alone. AI use is growing fast across Canadian business. But our own industry, wholesale trade, is among the slowest to adopt it.

AI use has tripled in Canadian business. Wholesale is lagging.

Share of businesses using AI (%)

  • 6.1%
    All businesses
    Q2 2024
  • 12.2%
    All businesses
    Q2 2025
  • 19.2%
    All businesses
    Q2 2026
  • 7.9%
    Wholesale trade
    Q2 2026

Source: Statistics Canada, Analysis on artificial intelligence use by businesses in Canada, second quarter of 2026.

What it means for you: AI use in Canada has tripled in two years, but most wholesalers aren't there yet. At Randmar, we'd rather be early, so the retailers we serve get the benefit now.

5 jobs I give AI

  1. Filling listing sheets. Hundreds or thousands of fields, from specs we already have, in the retailer's own template.
  2. Picking and formatting images. Choosing the images that best fit a retailer's customers, then sizing them to spec.
  3. Writing for the right shopper. Adjusting titles and descriptions to the retailer's typical customer, with the right SEO information.
  4. Enriching thin product content. At Randmar, we enrich product information using AI tools and our own team, so listings answer the questions shoppers actually ask.
  5. Answering everyday questions. Every Randmar account includes a virtual assistant for business questions, orders and product search.

3 jobs I never give AI

  1. The final review. I've always been a fan of the saying "garbage in, garbage out." It still applies with AI. Yesterday, I reviewed every listing and caught 9 mistakes: images that didn't perfectly fit the product listing.
  2. Deciding what's true when the source is wrong. Most of those mistakes came from source information that was unclear or wrong. AI can't know that. A person who knows the product can.
  3. The relationships. Retail is still a relationship business. AI can help me prepare for a meeting, but it can't sit across the table from a buyer, a brand or a store owner. Whenever I can, I like to talk with partners face to face. That's where trust gets built.

AI did the first 18 minutes. The part that protects the retailer, and our reputation, is still human.

My real point

Retailers, buyers and category managers are busy running their own business.

So at Randmar, we let AI do the heavy lifting on product definitions and imagery, and our people make sure every detail is right. The products you choose to support are set up to be winners, for your shoppers and for the AI tools they now use.

We do the work, so you don't have to worry about it.

Your quick win this week

Shop your own store the way an AI shopper would. Ask an AI assistant to describe and compare three of your best sellers, using your website. If the answers are wrong or thin, your product listings need work, and that's where to start.

Your turn

What's one job you'd hand to AI tomorrow, and one you never would?