Everyone on LinkedIn seems to have an AI workflow that changed their life. Most of those posts are written by people selling AI workflows. I am not. I run a consultancy and help lead the team at OpenMinds, and I use AI the way I use any tool, only where it earns its keep. Nearly everyone in marketing uses it now, so the interesting question was never whether you use it. It is where you let it near the work, and where you keep it out. This is where I draw that line, in practice.
Audits start with me reading slowly, then AI does the second pass
I start most client audits the same way I have since 2009, by reading the website slowly, the way a customer would. What has changed is what happens next. I hand the raw material to an AI model and interrogate it. Which claims are vague? What would a first-time visitor still not know? What do the competitors say that this brand does not?
It does not replace judgement. It replaces the hour of note-taking that used to sit between reading and judgement. The rule I hold myself to is simple. AI drafts the observations, I decide which ones matter. A machine can spot that your homepage never mentions price. It cannot know that in your market, leaving price out is deliberate and the right call.
The unglamorous middle is where AI saves time
Most of consulting is turning findings into things a client can act on. AI has quietly taken over the boring half of that.
- Meeting recordings become action lists. Recorded calls turn into structured follow-ups in minutes. Nobody misses typing up transcripts.
- First drafts of ad copy. When we build ad variations, I generate many angles, keep a few, and rewrite those properly. The keep-rate is low. The time saved is still real.
- Spreadsheet cleanup. Campaign exports, messy category names, mismatched date formats. Work that used to soak up an afternoon now takes a prompt and a sanity check.
Notice what stays off the list: strategy, budgets, and anything a client signs off. Those stay human. Not out of principle, but because I have tested the alternative and the quality was not there.
I let AI write code, then review every line like a junior developer wrote it
My teams have built and maintained more than 30 websites over the years, on WordPress, Elementor Pro, WooCommerce, Shopify and custom builds. AI has changed that work. Boilerplate code, tracking snippets, small scripts, all of it is faster to generate and verify than to write by hand. I have also used Claude Code to build internal tools, including a dashboard that tracks social channels across many clients and writes a weekly summary. The verification part is not optional. AI-generated code is confident in exactly the way a junior developer is confident, and it deserves the same code review.
What I keep AI away from, and why it matters
I do not use AI to write client reports end to end. Clients can tell, and they should be able to tell. I do not use it to invent statistics, which it will happily do, and present with total confidence. Anything with a number in it gets checked by a human before a client ever sees it. And I do not use AI stock imagery that looks like every other AI stock image. For real work we use Gemini and our own Project Reverie for visuals, with a human deciding what actually ships. A visual that does not feel like the brand costs more than it saves.
The honest takeaway
AI has not made me a different consultant. It has made the consultant I already was faster at the parts that never deserved the hours. Treating it like a search engine, where you type a question and take the answer, is the wrong process. I am not an AI expert and I do not claim to be. I only recommend what I have already run in my own work. If you are wondering where to start, pick one repetitive task you already understand deeply, let AI draft it, and check its work like a sceptical boss. That is the whole trick. If you lead a team rather than only yourself, I turned that into a playbook I would hand any small marketing team.
