Shopify Sidekick: what it can do for a merchant

Sidekick has gone from something that answers questions to something that does tasks in your admin. The most useful parts are reusable skills and app generation. The parts that are still oversold are anything that needs judgment about your customers.

Shopify's Winter '26 release, which it branded the RenAIssance Edition, included more than 150 updates, most of them AI. With that many, it's hard to tell what will change your working day. So here's the short version.

What's different now?

Sidekick now does things instead of just suggesting them. It reads the page you're on, can see your store's data, and can take action instead of telling you where to click.

In practice, you can type plain instructions for things that used to mean digging through settings, like editing themes, building automations or generating Flow workflows. What it saves you isn't knowledge. It's the twenty minutes spent finding where a setting lives.

Which parts are worth using?

Reusable skills. This is the most underrated addition. You save a prompt you use regularly as a repeatable action and run it again whenever you need it. That turns Sidekick from a novelty into something you rely on, because a prompt earns its keep on the fiftieth run, not the first. Anything you do every month is a good candidate, whether that's pulling a report, rebuilding a seasonal collection or tidying up before a sale.

Custom app generation. Sidekick can build a small app inside the admin, let you tweak it visually, test it against your store and install it. For problems that are specific to how you operate but too small to justify hiring a developer, this is a real change. It won't replace engineering, but the gap between "annoying manual process" and "worth building something" just got a lot smaller.

Sidekick Pulse. Suggestions it comes up with on its own after looking at your store. They're useful for pointing your attention somewhere. Treat them as a list of things to look into, not a to-do list.

What still needs a human?

Sidekick knows your store's data. It doesn't know your business.

It can't tell you whether you can afford a discount, because it doesn't know your contribution margin after fulfillment, returns and the shipping costs you absorb. It can't tell you which customers are worth keeping, because that depends on cohort behavior it isn't modeling. And it can't tell you what makes you different from a competitor, because that lives in conversations, not in your admin. Once you've worked that out, the Knowledge Base is where you tell AI agents.

The pattern is that it's strong at getting things done inside Shopify and weak at anything that needs context from outside it. That's the right way around, because the execution is the tedious bit.

Does it replace an agency or a developer?

It replaces some tasks, but not the work itself. We'd rather our clients used it, because what it does well is the stuff nobody enjoys paying an agency for.

What it doesn't do is decide what should change. Ask Sidekick to build a bundle and you'll get a bundle. Whether bundling makes sense for your margins, whether the bundle should be a discount or an added-value offer, and whether it eats into sales of a better-performing SKU are all still judgment calls. Getting those wrong faster isn't progress.

Where should I start?

Pick one boring task you did last month and will have to do again this month. Do it through Sidekick, save it as a skill, and run it again next month.

Doing that once will tell you more than reading through the feature list, because it tests what matters most: whether the output is good enough to use without editing the second time around. If it is, you've won back a bit of your week. If it isn't, you've learned where the limits are for your store.

Don't start with anything customer-facing. Copy, pricing and merchandising are where an answer that sounds right but isn't will cost you money, and they're also where you're least likely to spot the mistake.

We build Shopify automation for ecommerce teams. The first question we ask is always what to automate, not what to buy.