Marketing Engineering with AI · Coming soon
Get reviews answered on time without letting AI post unchecked
For marketing and customer teams who own reviews and lifecycle email and want AI to do the drafting without speaking to customers unchecked. You build two drafting agents, one for review replies and one for lifecycle emails, that follow your brand rules and wait in an approval queue, with a log of every decision.
The course: Agents for lifecycle email and reviews, with approval gates
$25 $49 early-bird price
This course is being built now. Join the early-access list to reserve the early-bird price and shape what it covers.
Joining the list is free. You only pay if you buy at launch, at the early-bird price.
If this sounds like your week
- Reviews sit unanswered for days because nobody has time, and they sit in public.
- ChatGPT drafts sound like ChatGPT, and none of them mentions what the customer wrote.
- Lifecycle emails wait on one busy writer, so the welcome and win-back flows never get past their first version.
- The automation templates you found post straight to the review site with nobody checking, and one wrong promise would be public.
What you do now, and where it stops working
Most teams paste reviews into ChatGPT one at a time, or pay for a review tool with an auto-reply switch they are afraid to turn on. Copy and paste stops working past a handful of locations, and auto-posting puts the model’s mistakes in public — a Canadian tribunal held Air Canada liable for what its chatbot told a customer.
What changes after the course
- New reviews arrive as drafted replies, sorted so the one-star complaint sits at the top of the queue.
- Lifecycle email drafts come from a brief and your brand rules, so the writer edits instead of starting blank.
- Nothing reaches a customer until a person accepts, edits or rejects the draft.
- A log of every draft and decision shows which kinds of review the agent gets wrong, so you know what to fix.
How it works: Draft, check, approve, log
- 1Rules: write your brand voice and policy rules — what the agent may say, must never promise and must escalate.
- 2Draft: the review agent sorts new reviews by urgency and drafts replies that answer what the customer wrote; the email agent drafts from a brief.
- 3Check: automatic checks flag refunds, discounts and banned phrases before a person sees the draft, so approval is not the only safeguard.
- 4Approve: a queue in a spreadsheet or a simple app, where a person accepts, edits or rejects each draft.
- 5Learn: read the decision log to see where edits cluster, and tighten the rules there.
What you get
- Agent code for the email and review-reply drafters
- Brand voice and policy rules template
- Approval queue workflow you can run in a spreadsheet or a simple app
- Example review set for practice
- Video walkthrough of both agents end to end
This is for you if
- You own reviews or lifecycle email for one or several locations, or for a store, and fall behind on both.
- You want AI help and refuse to let it reach customers unchecked.
- You can follow and edit example code with a walkthrough, or you build in n8n, Make or Zapier.
Skip it if
- You want replies posted automatically with no person in the loop. The course is built around the approval gate.
- You need a finished review product with a support contract. This is code and templates you run yourself.
- You already run AI drafting with automatic checks, approval queues and decision logs in production.
Before you join
- Will the drafts sound like us, or like every other AI reply?
- Generic replies come from generic instructions. You write a voice and policy file — phrases you use, phrases you never use, what you can offer — and the agent drafts against it. The decision log then shows which drafts were edited, so the rules improve from your own corrections.
- What stops it promising a refund or saying something wrong in public?
- Two layers. Automatic checks flag any draft that mentions refunds, discounts or anything on your never-say list, and nothing is sent until a person approves it. The course treats approval as required, never optional.
- Will approving every draft just turn into rubber-stamping?
- It can, which is why approval is not the only safeguard. Drafts arrive sorted by urgency with flagged issues shown beside them, so the reviewer reads the risky ones closely and clears routine thank-you replies quickly.
- I do not write code. Can I still use this?
- You need to follow and edit example code with the walkthrough open, not write an agent from scratch. The approval queue runs in a spreadsheet, so the daily work happens in a tool you already use. If you never want to open code, a review tool with built-in drafting will suit you better.
- Who is teaching this?
- Jay West leads DevOps for the marketing organisation of a multi-location specialty retailer, where he builds AI agent systems and the quality checks around them. This course packages that approach — drafts, checks, approval and a log — with the working agent code included.
Rather have it built for you? Describe the project and the same work can be done for your team.