AI readiness for ecommerce brands and what to fix first

You bought the AI. You switched it on. And the emails still go out to everyone at once, same words, same hour.

You're not doing it wrong. In Salesforce's 2026 State of Marketing report, which surveyed 4,450 marketers across 26 countries, 51% said their campaigns still feel generic even with AI running. And 98% of the teams already using AI hit at least one data problem before personalisation could work. So the tool isn't what's holding you back. The ground it's standing on is.

It's the most common feeling in ecommerce right now. The AI is on, and nothing feels different.

Around 80% of retail and consumer brands are now using or piloting generative AI, by NVIDIA's count. Almost every one of them is standing on the same soft ground, wondering why the clever new thing sends dumb old emails. The gap between adopting AI and being ready for it is the whole game, and almost nobody names it.

That ground has a name. AI readiness. It isn't how clever your model is. It's whether your data, your lifecycle and your measurement are in a state an agentic marketing platform can actually act on. Get it right and the same AI returns roughly three times what it did. Get it wrong and you've automated a blast.

What does AI readiness actually mean for an ecommerce brand?

AI readiness is the state of your foundations, and the biggest one is data. Salesforce found that agentic marketing has to pull from seven separate data sources on average to work, and 98% of the teams already using AI hit a data barrier before they get there. Your readiness is simply how few of those barriers you've got left.

I think of it as five foundations, stacked like a ladder:

• Data. Can the AI see the order, the browse, the email open and the support ticket in one place?
• Lifecycle coverage. How much of the customer journey do you actually send to, and how much is dead air?
• Personalisation depth. Is it a first name in a subject line, or the right product at the right moment?
• Decision cadence. How often do you decide what to send, once a campaign or once a customer?
• Measurement discipline. Can you prove a send made money, or are you guessing?

Most brands are strong on one rung and blank on the next. A lovely welcome flow sitting on data the AI can barely read. Or rich data feeding a tool that only ever fires two campaigns. The rung you're weakest on is the one capping your return, so you don't fix all five at once. You climb. And where you sit on the ladder decides how much the AI hands back, long before the model does.

Why does your data decide the result before the AI does?

Because the AI can only act on what it can see. In the same Salesforce data, 46% of marketers said they don't have the customer preference data they need to make anything relevant, and teams that had unified their data were 1.4 times more likely to reach people at the right moment. The signal was there. It was just scattered across tools that don't talk.

Picture it. Your Shopify order history sits in one place. Browse behaviour in another. Email engagement in Klaviyo. Support in a fourth tool. Ask an AI to choose the next message for one customer and it's reading a quarter of the story, so it plays safe and sends the generic thing. You're back where you started, now with a subscription fee.

The first fix is boring, and it's the one that pays. Get the obvious signals into a single view: what someone bought, what they looked at, when they last opened, how far into the subscription they are. If you're on Shopify and Klaviyo, that link is where readiness begins, which is exactly where a connected Shopify and Klaviyo setup earns its keep before you automate anything on top of it.

How do you know if your lifecycle coverage is ready?

Map the journey and look for the dead air. Personalisation lifts revenue by 5 to 15% when it works, going by McKinsey's research, but only across the moments you actually cover. Most brands cover two: welcome and abandoned cart. The long middle is empty.

That middle is where the money leaks. The stretch after a first order, when someone quietly decides whether they'll come back. The second order that never gets asked for, because the flow stopped at thanks for buying. The replenishment point, when the bottle is about to run out and a nudge would land. The slow lapse, the customer drifting off with no goodbye. Every empty stretch is a moment the AI could have used and didn't, because nothing was ever set to cover it.

The fastest way to see your own gaps is to replay one real customer's journey and mark every place you went silent. That's most of what our free readiness audit does. It finds the single biggest gap in your lifecycle before you change a thing.

How do you prove the AI is working and not just busy?

You hold back a control group. McKinsey puts the marketing efficiency gain from good personalisation at 10 to 30%, but you only get to claim it if you can see it, and you can only see it against a slice of customers the AI left alone. Same window, same conditions, one group worked and one not. The difference is your answer. Everything else is a story you're telling yourself.

This is the foundation most brands skip, and it's the one that turns AI from a guess into a number you can take to a board. A control group is the cheapest insurance you'll ever run, and it costs you nothing but the discipline to keep a group untouched. If a tool can't show you lift against one, it's asking you to trust it. Don't.

A control group is the cheapest insurance you'll ever run. If a tool can't show you lift against one, don't trust it.

What does readiness actually return?

One caveat I'd rather say out loud than bury. When we talk about what PilotX returns, it's modelled against a control group you set, a slice of your customers the agents never touch, so the lift is the difference and nothing else. It's modelled, not measured yet, and we have one live customer, so I won't wave a case study at you. Numbers dressed as certainty are the fastest way to lose a room that's seen a few of these tools already.

What the model says is that the return climbs with your readiness. Steering a few channels yourself, the floor sits near 19%. As the agents work alongside you and you approve each move, roughly 30%. When the confident moves run on rules you've set, up to 50% more revenue, against that same control group. That's about three times where legacy personalisation tops out, and the reason is readiness, not the model. You choose how much to steer, and the ladder rises as your foundations do.

So start with the gap, not the platform. We'll find the biggest hole in your lifecycle and build the fix on your own products, free, before you connect a thing. No deck, no six week onboarding, no rip and replace. Then a ten minute replay on your real customers, and if it earns it, a 14 day pilot for a few cents per decision, measured against a control group. If you want the shape of that first, the offer lays it out. Either way, the honest first move is finding out where you actually sit on the ladder.