If you run a Canadian e-commerce business, you've been pitched at least a dozen "AI marketing automation" tools. Most of them produce beautiful dashboards and modest, sometimes invisible, lift in actual revenue.

The ones that work — the small set of automation flows that actually move money — are mostly the same across every e-commerce category I've worked with. Here's what they are, what stack to build them on, and what to ignore.

The four flows that produce returns

Across the e-commerce clients I've worked with, four automation flows consistently produce meaningful revenue lift. Everything else is optimization theater.

1. Cart abandonment with progressive incentive. A customer adds to cart, doesn't check out, the system sends a sequence: reminder at 1 hour (no incentive), gentle nudge at 24 hours ("still thinking about it?"), final attempt at 72 hours with a 10% discount or free shipping incentive. AI tunes the message to the specific product they abandoned and their browsing history. Typical lift: 8-15% recovery of abandoned carts.

2. Post-purchase follow-up with cross-sell timing. After the order ships, a sequence: delivery confirmation when tracking shows delivered, review request 7 days later, cross-sell recommendation 21 days later (timed to when the product would naturally be replenished or complemented). AI personalizes the cross-sell based on what they actually bought. Typical lift: 12-25% increase in second-purchase rate.

3. Win-back for lapsed customers. A customer who purchased 6-12 months ago and hasn't returned gets a tailored win-back: AI selects the product offer most aligned with their previous purchases, the email subject line is personalized, and the discount is calibrated based on their prior LTV. Typical lift: 4-8% reactivation rate on lapsed customers — small percentage of a large pool, real money.

4. Restock and back-in-stock notifications. When a product comes back in stock or restocks, customers who previously viewed or tried to buy it get notified within minutes. AI picks the right message tone based on how high-intent the previous interaction was. Typical lift: 30-50% sell-through on the first day of restock.

That's it. Those four flows account for the vast majority of revenue lift from marketing automation in any e-commerce business under $20M ARR. Everything else — sentiment analysis on social mentions, predictive churn scoring at the household level, AI-generated weekly newsletter content — is mostly noise.

The stack I deploy

For Canadian e-commerce clients, the standard build:

Shopify for the storefront (or BigCommerce, WooCommerce — the principles port). Events from Shopify are the trigger source for everything downstream.
Klaviyo for email and SMS — best-in-class for e-commerce, native Shopify integration, strong segmentation.
n8n as the orchestration layer for anything Klaviyo can't do natively (cross-system data enrichment, custom logic, multi-step decisioning).
OpenAI for the AI personalization layer — product recommendation, message tuning, subject line generation.
PostgreSQL or the existing CRM as the source of truth for customer state.

This stack is boring on purpose. Shopify + Klaviyo handles 80% of what a typical e-commerce business needs. n8n + LLM handles the 20% where the off-the-shelf flows aren't smart enough.

Where AI actually helps vs where it's marketing speak

Genuine AI value-add in e-commerce automation:

Subject line generation. A/B testing with LLM-generated alternates measurably beats human-only subject lines for most products.
Product recommendation in cross-sell. "Customers also bought" recommendations from a pre-trained model outperform manually curated lists, especially in the long tail of SKUs.
Message tone tuning. An angry post-purchase complaint email gets a different tone of reply than a confused one. AI classifies and adjusts.
Pricing personalization within compliance. Showing different shipping thresholds or discount offers to different customer segments based on price sensitivity. This is real, but stay inside Canadian consumer protection law on this — opaque dynamic pricing draws scrutiny.

What's mostly noise:

AI-generated email body content. A small lift on some metrics, but the risk of off-brand or hallucinated content usually isn't worth it for transactional flows.
Predictive churn scoring at the individual level. Most useful at aggregate trend level, less useful at "this specific customer is 73% likely to churn next month."
Sentiment analysis on social mentions. Interesting, rarely actionable for small e-commerce.

What to measure

Three numbers matter for the four flows above:

Recovered cart revenue as a percentage of abandoned cart value.
Second-purchase rate for customers who completed first purchase.
Reactivation rate for the lapsed-customer segment.

Track these in a single weekly digest. If any of them isn't moving, the corresponding flow is broken or needs tuning. If all three are moving, you're winning.

What not to put in your weekly report: total revenue from automated flows. It's misleading — most of it would have happened anyway, and the metric encourages you to claim credit for organic traffic.

What it costs

A standard build for a Shopify e-commerce client doing $500K-$5M ARR:

Build cost: $4,000-$8,000 for the four flows + LLM personalization layer + dashboards
Klaviyo subscription: based on list size, typically $200-$1,500/month at this scale
n8n hosting + LLM tokens: $50-$200/month variable

For most clients in this range, the incremental revenue from the four flows clears the cost within 60-90 days.

Where I push back

If your e-commerce business is sub-$200K ARR, much of this is premature. Use Klaviyo's out-of-the-box flows, run the basic cart abandonment, and skip the rest until you have volume to optimize against.

If your e-commerce business is over $20M ARR, you've outgrown my advice — you need a dedicated marketing-ops team and a more sophisticated stack than what I'm describing.

The sweet spot is mid-market Canadian e-commerce: enough volume that automation has real leverage, small enough team that you can't afford a full in-house marketing-ops function. Book a discovery call if this is you, or read more about how we approach workflow automation including the e-commerce stacks.