Pick my brain!
Fill out a few details so I can prepare before we chat.
Every brand uses AI now. Most of it is shortcuts. One kind compounds; the other doesn't.
Every Shopify brand uses AI now: product descriptions, ad copy, social captions. Look closely and it is nearly always the same move. Open ChatGPT, paste a prompt, take the text, tidy it, move on.
That is a shortcut, not a workflow.
An AI workflow is a loop: a human designs the structure, the machine does the work, and what comes out teaches the next run. That is the part that compounds.
Most brands bolted AI onto how they already worked, as most developers did. METR, a research group, ran a controlled trial on experienced developers using AI tools. They finished their tasks 19% slower. They believed they had been 20% faster.
Marketing goes the same way. A brand makes a month of social posts in an afternoon, feels productive, and watches engagement sit exactly where it was. The copywriting was never the bottleneck. The thinking behind it was.
Two of our own open-source tools show the shape of it.
Email campaigns. The campaign writer reads a stored voice profile, takes the parameters — segment, offer, sequence length — and produces the full sequence: subject line variants, hooks, calls to action, plus a Figma template in the brand's colours and type. The interesting part is not that a machine wrote the emails. It is what had to exist first: a documented voice, a defined audience, a stated objective, set design standards. The workflow forced the strategy into writing. That is where the value sits.
Shopify themes. The theme agent turns a Figma design into a production Liquid template, then deploys it with a preview link. A second agent compares the result against the original design, pixel by pixel. A third audits the code for accessibility. Three agents, three jobs, three definitions of done.
Humans say what good looks like. Machines make it. The system improves because the specification does.
The best engineering teams found something odd. Writing code was never the hard part. Describing what the code must do, precisely enough for a machine to build it, is. A three-person team at StrongDM runs what they call a dark factory: specifications in, working software out, with nobody writing or reading a line. It stands entirely on the quality of the spec.
Marketing is arriving at the same place. Most brands cannot hand their positioning to a machine and get anything usable back — not because the machine is weak, but because the positioning was never written down. It lived in the founder's head, in Slack threads, in "you'll know it when you see it".
A workflow forces that into the open: who the customer is, how the brand sounds, what the campaign logic is, what counts as success. None of this is AI work. It is strategy work, deferred for years because a human reading the brief could fill the gaps from memory and a follow-up message.
The machine builds what you described. Describe it vaguely and you get vague, every time.
Three layers that turn AI from a shortcut into a system. Each with one job and a clear test for done.
Whether to use AI is settled. The open question is whether you have the specification that turns it from a shortcut into a system.
If you don't, the answer is not a better tool. It is a clearer strategy: positioning on paper, a brand voice written as a working spec rather than a mood board, defined segments, and campaign logic that lives somewhere other than one person's head.
Let's turn your marketing strategy into a specification a machine can execute.