Every brand "uses AI" now. But most of that usage is just shortcuts. The distinction matters because one compounds and the other doesn't.
Ask a founder whether they've adopted AI and the answer is almost always yes.
Every Shopify brand "uses AI" now. Product descriptions, ad copy, social captions — the usual.
But look deeper, and most of that usage looks the same: open ChatGPT, paste a prompt, get some text back, tidy it up, move on. That's not a workflow. That's a shortcut. The distinction matters because one of those things compounds and the other doesn't.
When we talk about AI workflows, we mean something specific. A repeatable system where AI handles execution within a structure that a human designed — and where the output feeds back into the system and gets better over time. A loop.
Most brands adopted AI the way most software developers did — layered it onto how they already work. There's a study from METR, a research group that ran a randomised controlled trial on experienced developers. The ones using AI tools completed their tasks 19% slower. And the part that should give you pause: those same developers believed AI had made them 20% faster.
The marketing equivalent plays out the same way. Brands producing more content, faster, with less clarity about whether any of it is working. Volume without direction. The AI is fast. The strategy it's executing against is vague. More output from a vague strategy is just more noise — produced quicker.
I keep seeing this pattern. A brand generates a month of social content in an afternoon, feels productive, and then watches engagement stay flat. The bottleneck was never the speed of the copywriting. It was the thinking behind it.
Let me get concrete, because "AI workflow" is the kind of phrase that sounds like a buzz term — but it's a fundamentally different way of working.
Host Digital's own open-source tools are a good way to show what I mean.
Email campaigns. The email-campaign-writer pulls in a stored brand voice profile, takes campaign parameters — audience segment, offer, sequence length — and generates a full email sequence with subject line variants, narrative hooks, and calls to action. It also produces a Figma email template with the brand's own colours and typography (we 'manually' export from Figma to Klaviyo). The interesting part isn't that AI writes the emails. It's what had to exist before the AI could write them well: a documented brand voice, a defined audience, a clear campaign objective, specified design standards. The workflow forced the strategy to become explicit. That's where the value lives.
Shopify themes. The shopify-theme-agent takes a Figma design and converts it into a production Shopify Liquid template — scaffolding, validation, deployment, preview link. A separate agent then compares the built template against the original Figma file pixel by pixel, and another audits the code for accessibility and Shopify best practices. Three agents, each with a defined job, each with clear success criteria.
The pattern: humans define what good looks like. Machines produce it. The system gets better because the specification gets better.
This is the idea I keep coming back to.
In the software world, the most advanced engineering teams discovered something counterintuitive. The bottleneck isn't writing code — it's describing what the code needs to do with enough precision that machines build it correctly. There's a three-person team at StrongDM running what they call a "dark factory" — specifications in, working software out, no human writes or reviews a single line. The whole thing runs on the quality of the spec.
Marketing is arriving at the same realisation. Most brands can't hand their positioning to an AI and get usable output back. Not because the AI isn't capable. Because the positioning was never written down with that level of precision. It lived in the founder's head, in Slack threads, in "you'll know it when you see it."
AI workflows force the strategy to become legible. A documented ICP. An articulated brand voice. Defined campaign logic. Customer feedback. Performance data. Clear success criteria. This work isn't AI work. It's strategy work — the kind most brands have been deferring for years because the humans on the other end of the brief could fill in the gaps with intuition and a follow-up message.
The AI builds what you described. If the description is vague, the output is vague. Every time.
A framework for building the specs that turn AI from a shortcut into a system. Three layers. Each with a defined job. Each with clear success criteria.
The question isn't whether to use AI in your marketing. That's settled. The question is whether you have the strategic layer — the specification — that turns AI from a shortcut into a system.
If the answer is no, the first step isn't a better AI tool. It's a clearer strategy. Documented positioning. A real brand voice — not a mood board, a working specification precise enough that a machine can execute against it. Defined audience segments. Campaign logic that exists outside someone's head.
Let's talk about turning your marketing strategy into a specification that machines can execute against.