Pick my brain!
Fill out a few details so I can prepare before we chat.
Every brand adopted AI. Almost nobody built the plumbing to make it work.
Someone on the team opens a chat window, types "write me an email for our spring sale" and gets back 200 words of competent, forgettable copy.
The prompt was one line. The machine knew nothing about the brand's voice, its customers, what worked last time or what the sale was for beyond shifting stock. So it did what any new hire does with a vague brief: it guessed, and landed in the middle. Safe. Publishable. Like everyone else.
Now run that across email, social, ads and product pages. More content, faster, with less idea than ever whether any of it works. A vague strategy produced at speed is just noise, sooner.
The tools are not the bottleneck. The inputs are.
AI does not have the idea. It does not know your customer, or why last year's campaign worked and this year's did not. What it does well is take what already exists and remix it: structure, format, language, twenty variations by lunch.
So the output is only as good as what goes in. Feed it your best sequences, your written voice, your customer research and your positioning, and the work is sharp. Feed it a sentence about a spring sale and you get exactly what you deserve.
AI doesn't replace the thinking. It multiplies it.
That is good news if you have been at this a while. Every email you wrote, every campaign you ran, every phrase you learned from a customer is the fuel. Beginners have none of it, which is why their output sounds like everyone else's.
Useful output is a question of what you build underneath.
Context, and a loop. The machine needs to know your position, your customer and your voice — not once, but every time. Output, feedback, better output. You still supply the idea; it executes against your direction. The longer and sharper the brief, the better the work.
Your SOPs are your prompts. The document you would write for a new hire is the one that tells an agent how to do the job. If your prompts are one line, your SOPs were always one line — you just had people filling the gaps.
Every email sent, every ad run, every sales call transcribed is fuel. Brands that have done the work for years are sitting on a seam of it. And it crosses channels: the sequence that converted can write the ad, and a sales call can write the landing page.
Every client here has a brand folder: not a shared drive of logos, but past campaigns, articles, style guides, voice documentation, product sheets, customer research and competitive position, in one place. Every workflow starts by reading it. That is the asset layer, and it is why the machine never has to invent a product benefit.
On top sits the workflow layer. Writing a Klaviyo sequence is not a one-line prompt. It is a documented procedure: pull the voice profile, pull the campaign parameters — segment, offer, length — then produce subject lines, hooks and calls to action in a set structure. The prompt behind it runs to several pages. That is an SOP with intelligence attached.
The hygiene layer keeps it from going stale. After each campaign we ask what converted, what fell flat and what sounded wrong, and the answers go back into the folder. The system improves because the context does.
If your brand voice lives in the founder's head, the machine cannot reach it. If your best sequences sit in Klaviyo with no note of why they worked, it cannot learn from them. If your position changes every quarter and is never written down, you will get a different answer every time you ask.
Writing down your voice, sorting your past work, defining how things get done: none of it feels like progress, which is why it keeps getting skipped. It is also what decides whether AI gives you an edge or just more of the same, faster.
The brands that did things the hard way for years are being rewarded for it. They have the data, the patterns and the judgment. They only need to arrange it so a machine can read it.
Nobody is aiming at full automation. What works is a person setting direction and a machine multiplying the output. You still have the idea, set the structure and decide whether the result is good enough.
The machine multiplies what you feed it. Feed it nothing and you get nothing, at scale. Feed it everything you have learned and you get something that sounds like you on your best day, on every channel at once.
The difference between AI as a gimmick and AI as an operating system is the work you do before you touch the tools.