At the Spring '26 Edition, Shopify introduced Campaign Autopilot, a tool that plans and launches marketing campaigns for you from the admin itself. Shopify describes it plainly as a virtual marketing agency, aimed at the brands that cannot afford an agency or a freelancer. So the question asks itself, and several clients have been putting it to us these past weeks. If Shopify already sets up my campaigns, why do I need an agency?
We are going to answer it with the only authority that makes sense here, that of someone who lives this every day. We are not going to pretend that a tool automating part of our work is useless, nor sell you fear to protect ourselves. We are going to tell you what Campaign Autopilot does well, where it falls short and, above all, what the underlying risk is of handing your growth to an algorithm whose objective is not exactly yours.
What Campaign Autopilot actually is
Campaign Autopilot is an AI agent that recommends, launches and optimises paid and organic campaigns from the Growth tab of the Shopify admin (the old Marketing tab). It works on your store's data (catalogue, customers, orders) and operates for now across three channels: Meta Ads (Advantage+ campaigns on Facebook and Instagram), Shop Campaigns (ads in the Shop app, available in the US and Canada) and Shopify Messaging (email automations such as abandoned cart or welcome series). Shopify has announced more channels on the way, such as Microsoft Advertising, ChatGPT Ads and Snapchat.
The way it works is deliberately simple. You set a budget and some guardrails (including a ROAS target), the tool proposes "tactics", you approve or reject each one, and from there it moves the budget towards what performs. A couple of important and honest nuances, because they are true. For the creative it uses your catalogue images, it does not generate images with AI, so what comes out looks like your store. And nothing is published without your approval. It is also worth knowing that it is in early access, not general availability, so it is a new tool with no proven track record at scale yet.
What it does well
Let us be fair, because it has real strengths. For a brand just starting out, that has never run an ad and has no budget for an agency, Campaign Autopilot lowers the barrier to entry considerably. It handles the technical setup, which for many people is the first wall, and coordinates several channels from a single place instead of forcing you to jump between tabs. That it uses your real catalogue and that you approve each tactic gives it a welcome degree of control.
As a starting point for someone with no marketing and a very tight budget, it is a reasonable option. Denying that would be dishonest. The problem is not what it does, it is what it cannot do and what it takes for granted without telling you.
The underlying problem: its objective is not your profit
Here is the heart of the matter. Campaign Autopilot optimises towards orders and revenue, with ROAS as the guardrail. And revenue is not profit. A campaign can have a great ROAS and be losing money, because the algorithm does not know your contribution margin, nor your product cost, nor which SKUs are loss leaders and which give you oxygen. It only sees the attributed sales figure.
We measure POAS, the profit on ad spend, not the platform's gross ROAS. It is a difference that changes everything, and we have explained it in full in POAS vs ROAS. A system that chases revenue without seeing margin tends to do exactly what generates the most conversions in the short term: push discounts and promotions (in Shop Campaigns it can include Shop Cash offers), scale the best-selling SKUs even if they are the lowest-margin ones, and buy orders that look good on paper and subtract from your P&L. The algorithm is not deceiving you, it is simply optimising the metric it has been given, which is not yours.
What happens when it does not work
The part almost nobody tells you. Shopify itself warns that results take time while the system learns and that it cannot guarantee them. It is an honest warning. The problem comes the day POAS collapses, because then there is no one to ask why.
A campaign stops working for many reasons, and almost none of them are inside the tool. Creative fatigue, audience saturation, a price or margin problem, a landing page that converts badly, a competitor that has come in strong, seasonality, a stock failure. Campaign Autopilot only touches the mechanics of the ad, so faced with a drop it will keep doing the only thing it knows, redistribute budget across channels, without diagnosing the real cause. An agency looks at the whole funnel, from the creative and the landing to price, email and repeat purchase, finds the bottleneck and acts where the problem actually is. The automation does not have that view, nor anyone who answers for the result.
"It learns from millions of stores", and why that does not always suit you
Shopify presents it as an advantage that the tool learns from patterns across millions of stores and that this is why it recommends better than a generalist agency. It is a strong argument, but it has a flip side. Learning from the sector average means optimising you towards what works on average for your category. And your advantage as a brand is not to resemble the average, it is to differentiate yourself from it.
Optimising towards the average pattern pushes every store in the same vertical towards the same type of campaign, the same offer and the same angle. It works for not falling behind, not for standing out. The strategy that builds a brand starts from exactly the opposite: its specific positioning, its value proposition and its audience, not the average of its category.
Control and brand image
It is true that the creative uses your images, so you are not going to see strange AI-invented pictures. But the real control is more limited than it seems. You cannot change the audience, the system chooses the segment and the objective for you. The creative is assembled automatically from the catalogue, with no art direction, no campaign concept and no considered message. And the offers it proposes can clash with your positioning, because a discount or a Shop Cash-style incentive that improves today's conversion may be teaching your customer to always wait for the markdown and eroding your brand in the medium term.
Put another way, the images are yours, but the judgement to choose them, combine them and decide which product to push belongs to the algorithm. You can end up promoting SKUs you do not want associated with your brand, or with a tone that is not yours, simply because they convert a little better in the short term. Brand consistency is not a metric the system knows how to optimise.
Where an agency comes in and the tool does not
The whole conversation is clearer if you separate two planes. The execution of the advertising mechanics, which is what Campaign Autopilot automates, and the strategy, which is where it is decided whether those mechanics work for your business or against it.
We are not an agency that executes what an algorithm decides, we design the strategy the algorithm cannot see. The building of the positioning and the value proposition, the creative concept and the angle of each campaign, the real profitability measured in POAS and not in ROAS, the optimisation of the landing and the checkout, the pricing and margin strategy, the retention and repeat-purchase system that raises LTV, and the full-funnel coordination across channels the tool does not even touch. That is the plane where money is genuinely made or lost, and it is a plane of human judgement, not automatic optimisation.
So, do I turn it on or not?
Our honest recommendation, with no defensive posture. If you have a store just starting out, zero marketing structure and a budget that does not stretch to an agency, try it with tight guardrails and a small budget. It is a valid starting point for beginning to move data, and better that than doing nothing.
If your objective is to scale profitably, protect your positioning and you have any complexity of margins or catalogue, do not let it run on its own. Use it, at most, as a supervised complement, never as the sole pilot of your growth. Handing authority over your budget to an autonomous agent with no one watching the real profitability is the quietest way to watch an ad investment evaporate. The tool is interesting. What it is not, yet, is a substitute for the judgement that turns advertising into profit.
Frequently asked questions
Does Campaign Autopilot replace an agency? It replaces one part, the execution of the advertising mechanics. It does not replace the strategy, the judgement on real profitability, the brand direction or the diagnosis when something stops working. For a store starting out with no budget it may be enough; for a brand that wants to scale with margin, no.
Does it measure the real profitability of my campaigns? Not the way you need it to. It optimises towards orders and revenue with ROAS as the reference, it does not know your margin or your product cost, so it does not measure POAS, which is the profit on the investment. It can have a good ROAS and be losing money.
Do I lose control of my brand? In part. It uses your catalogue images, but you do not choose the audience or the product it pushes, the creative is assembled on its own with no art direction, and it may propose discounts or offers that do not fit your positioning. Control over brand consistency is limited.
What happens if it does not work? Shopify itself warns that it does not guarantee results. If performance drops, the tool only redistributes budget, it does not diagnose the cause, which is usually outside the ad (creative, landing, price, stock, competition). That is where an agency looks at the whole funnel and acts.
What kind of store does it make sense for? For brands that are starting out, with no marketing structure and a very tight budget, as a starting point and with tight guardrails. For brands looking to scale profitably and look after their positioning, it falls short as a sole solution.
Can I use it alongside an agency? Yes, and it is often the most sensible thing. The tool can cover part of the execution while the strategy, the POAS measurement and the brand direction are handled by whoever answers for the result. What is risky is letting it run on its own, unsupervised.
If you are weighing up turning on Campaign Autopilot, or you already have it running and are not sure whether it is paying off, we will review it with you. At STRAT we look at your account, your margins and your campaign structure to decide what is worth leaving on autopilot and what is better controlled by hand, because how we run paid media for ecommerce is measured in POAS, not in the platform's gross ROAS.
