How is this different from an automation tool like Klaviyo flows or Zapier?

A flow runs the branch you drew for it. If a buyer replies with something you did not anticipate, the flow has nothing to say. An AI employee reads the situation, decides whether to act, chooses the channel and offer within your policy, and escalates when the right answer is outside its authority.

What stops an AI employee from discounting away my margin?

Guardrails are set before anyone starts work — discount ceilings, tone, which products can be offered, when to escalate. Every decision is tested against those rules before it reaches a customer. Anything outside them stops and waits for your approval.

How long does setup take?

Connect Shopify and your channels, set your policy, and the first employee starts the same day. There is no middleware to configure and no separate login for your team.

Which channels do they work across?

WhatsApp and email today, both arriving in one inbox with full customer context attached. SMS and storefront chat are on the roadmap, not shipped.

Can I start with one employee instead of the whole team?

Yes — most merchants start with Emma on cart recovery, because it is the fastest outcome to measure. You add the others once you can see the numbers.

What happens to conversations already handled by my support team?

They continue as they are. WhatsApp coexistence means your existing support stack keeps running while the AI employees take the volume you hand them.

How do I know they are actually working?

Every action is logged against the employee that took it, with the outcome attached — recovered revenue, resolved tickets, accepted upsells. You review their work the way you would review anyone else’s.

What does the free 60-day beta include?

The full team of five, every channel, and the complete work log. No credit card, and nothing starts billing automatically at the end of it.

Why Most Abandoned Cart Emails Never Get Read, And What the Data Says Actually Works

The average is hiding the real story

Klaviyo's most recent abandoned-cart benchmark, pulled from over 143,000 live flows, puts the average abandoned-cart email open rate at 50.5%, with a 6.25% click rate and a 3.33% conversion rate. Top-performing flows, the top 10%, hit 65.3% opens, 13.3% clicks, and 7.7% conversion, generating nearly eight times the revenue per recipient of the average sender ($28.89 vs. $3.65).

That gap between "average" and "top 10%" is the whole story. Abandoned-cart email isn't broken as a channel, it's broken as most people run it. The senders clearing 65%+ open rates aren't using a fundamentally different tool than the senders stuck at 40%. They're doing something different with it.

Before "will they open it" comes "will they ever see it"

Open rate assumes the email lands in the inbox in the first place. That assumption is getting shakier every quarter.

Validity's 2025 email deliverability benchmark report, based on real inbox placement testing across major mailbox providers, found that global inbox placement rates fell from roughly 87% in Q1 2024 to 82.3% by Q4 2024, while spam placement nearly doubled over the same stretch, from 4.5% to 8.6%. Put plainly: one in six legitimate marketing emails now fails to reach the inbox, and the trend line is getting worse, not better.

It's not evenly distributed either. Gmail, which now holds an estimated 48.5% share of consumer mailboxes, saw its own inbox placement rate decline by nearly five percentage points over 2024, landing around 87.2%. Microsoft inboxes were tougher still, averaging 75.6% placement. Apple Mail came in at 76.3%. If a meaningful share of a merchant's customer list is on Gmail, Outlook, or iCloud, a chunk of every "sent" abandoned-cart campaign is quietly never arriving, no bounce, no error, just gone to spam or nowhere.

Why filters are getting harder on exactly this kind of email

Spam filters don't primarily look at content anymore, they look at patterns. And the default abandoned-cart sequence most Shopify stores run is a pattern filters are specifically built to catch: the same subject line, the same static template, the same 10% discount, sent identically to every recipient on a fixed delay. That structure, one message, mass-personalized only by first name, blasted on a schedule, is close to the textbook definition of what inbox providers increasingly treat as bulk marketing rather than relevant correspondence, especially once engagement (opens, clicks, replies) on that pattern starts to decline across a sender's list.

Validity's report specifically flags rising sending volumes and static, low-engagement campaigns as a growing drag on deliverability, and separately notes that AI-generated bulk email is now common enough that it's making subscribers "less trustful of legitimate emails" as a category. The senders holding onto strong inbox placement are, disproportionately, the ones whose engagement signals stay high because the content is actually relevant to each recipient, not just addressed to them.

What the market is actually selling instead

Look at what's currently being built and sold in this category and the mismatch becomes obvious. TxtCart, one of the more visible "AI" cart-recovery tools on Shopify, is built primarily around SMS templates and copywriting guides, useful, but it's sidestepping the email deliverability problem by changing channel, not solving it. Much of the broader "AI cart recovery" content ecosystem (comparison posts, template libraries, swipe files) is optimized for writing a slightly better static message, not for generating a genuinely different message per customer. A better subject line on the same template sent to everyone is still the same template.

That's a structurally different approach than reasoning about each abandoned cart individually, what's actually in it, what the customer has bought before, how much time has passed, and generating a message from that context rather than filling in a mail-merge field. GoPrimy's cart recovery automation works this way: each recovery message is generated per customer from their actual cart contents and order history rather than assembled from one fixed template, which is the direction the deliverability data points toward, not because any platform can guarantee inbox placement (no legitimate one can, and any that claims to is overselling), but because relevance and engagement are the levers that are actually connected to inbox placement in the first place.

The honest takeaway

No one, GoPrimy included, can promise an email will land in a customer's primary inbox. Inbox placement is decided by mailbox providers, based on sender reputation and recipient engagement patterns that shift constantly and aren't fully visible to any sender. What the data does support is more specific: static, identical, high-volume templates are the pattern most likely to lose the inbox-placement fight over time, and personalized, context-aware messaging is the pattern most correlated with staying out of that decline.

If your abandoned-cart flow is still running the same three-email, 10%-off sequence most stores have used since 2015, the open-rate problem you're seeing probably isn't a copywriting problem. It's a pattern-recognition problem, yours, and the spam filter's.

Sources

Klaviyo, Abandoned Cart Benchmark Report (https://www.klaviyo.com/blog/abandoned-cart-benchmarks)

Validity, 2025 Email Deliverability Benchmark Report (https://www.validity.com/wp-content/uploads/2025/03/2025-Benchmark-Report-FINAL-1.pdf)

TxtCart Blog (https://txtcart.ai/blog/)