The problem this solves
Most abandoned-cart emails are a coupon wearing a reminder costume. They restate what the shopper already knows - you left something behind - and then discount, training customers to abandon on purpose. What they almost never do is answer the actual reason the checkout stalled: a doubt about quality, sizing, shipping, durability, or whether the thing works as promised. The shopper left because of an unanswered objection, and the recovery email arrives carrying no answer.
The answers exist, in public, written by customers. Reviews name the exact hesitations - "worried it would run small, it fits perfectly" - in language more persuasive than any copywriter's, because it is verifiably not the brand talking. Forum threads and marketplace questions reveal the objections shoppers voice before buying. But mining that corpus, matching real proof to real doubts, and building it into a sequenced flow is research plus copywriting plus platform assembly, and that combination is why cart flows stay generic.
A recovery flow built on proof works differently: the first email answers the likeliest objection with a customer's own words, the second brings social weight, the third closes honestly. Building that is a research-then-draft-then-assemble mission, which is precisely the shape of work an agent runs end to end, with your review sitting between the drafting and the platform so nothing generic slips through.
How the mission runs
- Objections are mined before a word is drafted. Web Research goes through your reviews, marketplace questions, and forum mentions to build two lists: the doubts shoppers actually voice before buying, and the review lines that answer each doubt in a customer's own words. This evidence base, with sources, is the raw material for every email that follows.
- Proof is matched to the purchase moment. The agent pairs each major objection with its strongest available answer - a review excerpt, a pattern across many reviews, a concrete detail - and ranks the pairs by how often the objection appears. The flow will spend its limited attention on the doubts that actually stall your checkouts.
- Three emails are drafted, escalating honestly. Email one returns quickly and answers the top objection with proof. Email two widens to social weight: what buyers consistently say, drawn from the mined corpus. Email three closes with clarity about the cart expiring or stock reality, staying factual rather than manufacturing urgency the store cannot back.
- Subject-line variants are written for testing. Each email gets multiple subject lines built on distinct hypotheses - the objection named, the proof led, the plain reminder - so a test can reveal which framing your audience opens. The variants are labeled with their reasoning, making eventual results readable instead of a coin flip.
- The flow is staged in Klaviyo for your review. After you approve the drafts, the agent builds the three-email flow in Klaviyo with the abandonment trigger, agreed delays, and the subject variants attached, then reports exactly what it created. The flow is staged rather than live, so you inspect the built thing before any shopper receives it.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A three-email abandoned-cart flow staged in your Klaviyo account, argued from evidence: the top real objections answered with your customers' own review language, escalation logic that stays honest, and labeled subject-line variants ready for testing. Alongside it, the mined objection-and-proof inventory with sources, which remains useful for product pages and ads long after the flow ships.
Make it yours
- Split by cart value: high-value carts get proof-heavy, discount-free treatment, while low-value carts test whether an incentive in email three moves the number.
- Build per-category versions: mine objections separately for your two biggest categories, since sizing doubts and durability doubts deserve different first emails.
- Refresh quarterly: re-run the mining pass on new reviews and let the agent propose copy updates where the objection landscape has shifted.
Frequently asked questions
Does the flow send as soon as it is built?
No. The build lands staged in Klaviyo, deliberately inactive, and the completion report links what was created so you can walk the flow in your own account. Activation is your click, made after review - the default across these playbooks is that nothing reaches a real inbox on the agent's initiative.
Are the review excerpts used verbatim, and is that safe?
Excerpts are quoted faithfully from your own store's reviews with their sources recorded, and the draft marks each one so you can verify it in context during review. Publicly mined objections from forums inform the arguments rather than being quoted, keeping the flow grounded without republishing anyone.
What if my product has few reviews to mine?
The mining pass widens: marketplace listings of comparable products, category forums, and question threads reveal what shoppers in your space doubt before buying, even when the proof must come from your product details rather than quotes. The agent reports how thin the evidence is, so you know which emails argue from proof and which from clarity.