Axel Freeman

AI marketing engineer · build, not advice

AI marketing engineer: the machinery, not the deck

Most “AI marketing” is a subscription plus an opinion. An AI marketing engineer does something narrower and checkable: takes the path that is supposed to bring customers and makes it run — agent skills that produce and check the work, pages built so answer engines can quote them, sequences and routing that fire on their own, and an event that records what happened. Packages are published before the call: Sprint $900, Engine $1,900/month, Full build $2,900.

The service: marketing engineer · what a package buys: cost page · edges: scope of work · start: Telegram.

Four things that have to exist before “AI marketing” means anything

01 · Capture

A page a model can quote

A page that renders without JavaScript, answers one question in the first 60 words, states prices as numbers and carries Service/Offer or FAQPage markup. Writing the page for a human reader and a model reader is the same job. Free check: 23-point readiness checklist.

02 · Production

Skills, not prompts

Good output comes from a reusable operating file the agent loads every time — what to do, what never to do, which numbers are floors and not opinions. Ours is public: marketing-mindset, also on npm and as an MCP server.

03 · Decision

A kill rule written first

How many data points, at what result you stop, what happens to the winner. The sample-size floor for a two-variant email test is roughly 1,500–2,000 sends per arm; the kill-rule calculator returns the Wilson interval and the rule-of-three floor when conversions are zero.

What “AI marketing engineer” has to beat

AI marketing engineerThe usual “we use AI” retainerWhat you can open afterwards
The unit of outputAn AI tool subscriptionAn asset in your accounts, with a link
Where the AI sitsIn the chat windowIn the pipeline that writes, checks and ships
What a month leaves behindScreenshots of promptsPages, events, sequences, a changelog
Can a model quote you?Not addressedSchema, llms.txt, plain-HTML answers
When it stops workingYou cancel the seatsThe machinery keeps running; you own it

The test is one question: name something I will be able to open when the month is over. A subscription answers with a tool. An engineer answers with a URL, an event and an account.

Ships you can open right now (the method applied to itself)

How the engagement works

Sprint · $900

One path, live in a week

Pick the broken or unmeasured path: the page, the sequence or the reporting. It gets built, instrumented and running, and you keep it.

Engine · $1,900/mo

The continuing block

Build, measure, kill the losers, scale the survivor. Fourteen days notice to cancel — a retainer you cannot leave is a lease, not a service.

Full build · $2,900

Handed over running

Offer, pages, outbound, automation, measurement and documentation, so the system survives without the operator.

One line about your bottleneck is enough to start: Telegram.

Questions people actually ask

What is an AI marketing engineer?

The person who makes marketing work without a human in the loop: agent skills that write and check the copy, pages built so answer engines can quote them, sequences and routing that run themselves, and tracking that reports what happened. The word “AI” is not the job title of a prompt writer — it is the unit of output: a working asset with an event attached, not a plan about AI.

How is an AI marketing engineer different from an AI marketer or an AI consultant?

An AI marketer talks about tools. A consultant writes a strategy and leaves. An engineer ships inside your own accounts: the page, the schema, the llms.txt, the sending domain, the event, the kill rule. When the engagement ends, the machinery is still in your accounts and still running.

What does the work actually produce?

Openable artifacts. On this site the same method produced an offer page with Service/Offer and FAQPage markup, a 56-URL sitemap with IndexNow pings, an llms.txt written for model crawlers, three free browser tools, four published npm packages, a public playbook repository and two dozen pages written under real queries. Every one of them is listed with a live link on the proof page.

How do I know the AI part is not decoration?

Check whether the artifacts can be opened and run. The npm packages are on the registry, the tools run in the browser with no account, the playbook chapters are readable on GitHub, the sitemap and the snapshot of each page are public. If a vendor cannot show a link for the thing they claim, the claim is a deck.

How is it priced?

Three published packages, no hourly rate: Sprint $900 one-time (one path built, instrumented and live), Engine $1,900 per month (build, measure, kill or scale; 14 days notice), Full build $2,900 one-time (the whole system handed over running). The comparison with a freelancer, an agency and a full-time hire is on the cost page.

What is out of scope?

Ad spend management, content volume promises, hiring, and paid licences. Ad spend and subscriptions stay in your accounts, with no markup. The scope in writing: scope of work.