AI for E-commerce Marketing: A 90-Day Rollout Plan

Before Day 1: the baseline

Write these down. Screenshot the dashboards. If a number doesn’t exist, that’s a finding — not a reason to skip it. Without this page, Day 90 is a vibe.

This is also the moment to name the stack honestly. Owned store plus three marketplaces is four catalogues, four image sets, four CS inboxes. AI does not merge them for you. It makes it cheaper to keep them consistent — if you brief it that way.

Days 1–30: foundations

Do not start with ads. Do not start with “get us into ChatGPT.” Start with the three things every order already touches: the listing, the photos, the reply when someone asks a question.

Listings

AI drafts the long-form copy and the attributes; a human edits. Channel-specific formatting, not one blob pasted four ways — Shopee, Lazada, TikTok Shop and your Shopify or WooCommerce page each want a different shape. The review is the work. Unreviewed listings are how you publish confident nonsense at catalogue scale.

Images

Do not go fully synthetic. Use AI for cleanup and background on real photos — dust, colour-cast, a clean white or lifestyle crop — not to invent the product. Fully generated product shots are starting to hurt once shoppers notice. Minimum set: four to six images per SKU, the same set adapted per channel rather than a different shoot for each.

CS macros

Pull the top 20 questions from the last 90 days of tickets. Size, shipping, COD, returns, “where is my order,” the product-specific ones no template ever included. AI drafts the macros; a human signs every one before it can fire. Typical first-contact resolution on those intents is 40–60% on a new programme. Vendors quote 60%+ after six to twelve months of grounding — treat that as a ceiling, not a Day 30 target, and only after you’ve measured it on your own queue. It still needs an escalation path a person owns. The ticket-type split is in AI for customer support.

Day 30 checkpoint. A sample of listings has been rewritten and reviewed. The image set is real photos, cleaned, not generated. The top 20 macros are live with a named owner. Conversion rate and first-response time have a number next to them that you can compare to Day 1.

Days 31–60: ads, audience, email

Now you have something worth putting spend and a list against. Not before.

Ad variants

Draft more variants than a person would, from the listings you just rebuilt — headlines, primary text, short-form hooks. Minimum two or three live tests, not one “AI campaign.” TikTok Shop’s AI Video Maker is useful here if TikTok Shop is already a channel; it is not a reason to open TikTok Shop. Keep a real host on camera for organic social. Fully synthetic presenters read as ads even when they aren’t, and they train the audience to scroll.

Four email flows

If you only have a newsletter, you do not have email. Stand these four up in days 31–60:

AI is useful here for subject-line variants and first-draft flow copy. That is not a reason to skip the human edit on the first send of each flow. It is a reason to stop writing every email from a blank page.

Day 60 checkpoint. At least two ad tests have enough data to kill or keep. Four flows are live. Email’s share of revenue has a number, even if it’s still small. Social has a real person on camera on a cadence you can point at.

Days 61–90: AEO, analytics, automation

This is the phase people try to start with. It only pays if the first 60 days exist. Some vendor studies report AI-referred sessions converting harder than generic organic; treat those as directional, not a 90-day target. The prize is being citable when a shopper asks ChatGPT or Gemini what to buy. It is not a blog post titled “ChatGPT’s favourite [category] brands.”

We track this on the LOKAL Visibility Index — 22 ecommerce brands plus 34 hospitality properties, Oct 2025–Mar 2026, across ChatGPT, Perplexity, Gemini and Google AI Overviews. Citation is a different job from “doing SEO.” For the method, see AI SEO in the Philippines and the ecommerce-specific version on ecommerce SEO.

• Make the pages citable. Product and Organisation schema, FAQ on the SKUs and the category pages, a review-velocity plan so there is something recent to cite. This is markup and operations, not a prompt.
• One dashboard. Store, marketplaces, ads, email, CS — enough to see whether a listing change, a flow or an ad test moved the baseline you wrote on Day 0. Four platforms without one view is how 90 days vanish into channel reports.
• At least three cross-system automations. Examples that earn their keep: new SKU → draft listing + attribute set for human review; paid order → post-purchase flow; ticket tag → macro with escalation. If you want this built rather than sketched, that’s automation.

Re-run the same 10–15 AI category questions from Day 1. That’s the only AEO checkpoint that matters. “We published schema” is an activity. “We showed up in two answers we didn’t show up in before” is a result.

Day 90 checkpoint. Citations have moved, or you know why they haven’t. One dashboard is the default view. Three automations are in production, not in a slide. The baseline page from Day 0 has a second column.

What breaks this

The 90 days fail in the same five ways, repeatedly:

• Skipping the baseline. Then every lift is a story. You cannot tell a listing rewrite from seasonality if you never wrote Day 0 down.
• Fully synthetic images. Fast, cheap, and increasingly a conversion tax. Clean real photos. Don’t invent the product.
• Unreviewed macros. A wrong shipping time or return rule, sent at 70% of ticket volume, is not an efficiency. It’s a CS incident on a timer.
• AEO as a one-off. A single “best of” article does not get you cited. Schema, FAQ, reviews, and a brand that already ranks — then you ask the models again.
• Channels as separate decisions. Shopee, Lazada, TikTok Shop and the owned store will drift unless the listing, the image set and the offer are briefed as one catalogue. AI makes the drift faster if you let each channel prompt itself.

What this costs in the Philippines

Philippine e-commerce marketing typically runs ₱80,000–₱600,000 a month. Brands in the mid-funnel, doing ₱1M–₱5M a month, usually invest 15–22% of revenue. The 90-day plan is a sequencing of that spend — listings, creative, media, email, a bit of build — not a new line item on top of it.

If you want this mapped onto the stack you actually run — owned store, Shopee, Lazada, TikTok Shop, the lot — that’s Total Visibility: ecommerce in the Philippines. Store build and ops sit with our Shopify Partner work. The AI-search half is AI SEO and ecommerce SEO. The wiring is automation. And if the team needs to run the hour and the brief themselves, Claude for Marketing is the session we teach it in.

Keep going: the AI guides hub and the e-commerce guides sit alongside this, and the Insights archive carries everything else the team publishes.