When I led the repositioning and website redesign for IOA, I used AI the way most people in marketing now do, as a starting point. Ask it for layout conventions, ask it for messaging frameworks, ask it what a B2B intelligence firm's homepage usually gets right and wrong. It's genuinely good at this. It knows best practice the way a well-read junior strategist knows best practice. Value proposition near the top. Clear call to action. Reduce friction before the form.
The problem is that best practice is a starting point, not an answer. And almost everything that actually moved the bounce rate at IOA came from somewhere best practice doesn't reach.
A decade in this space teaches you something AI hasn't had the chance to learn yet, which is when to break the rule rather than follow it. Sometimes the right move is exactly what every guide tells you to do. Sometimes it's the opposite. Moving a value proposition higher up the page is good advice in general. Whether it's good advice for this specific audience, on this specific page, given what they already believe about your brand before they've even arrived, is a different question entirely. That's not something you can prompt your way into. It's something you test, watch, and adjust, with a level of judgment that comes from having been wrong about it before.
That's the part of the work AI genuinely struggles with, not the writing, not the layout suggestions, but the instinct itself. Without it, AI can't tell which suggestion actually applies to your brand, so it defaults to generic best practice. That's why A/B testing still matters. A change that looks correct on paper, the kind any AI tool would happily recommend, can still underperform once it's live, not because the advice was wrong, but because it was written for nobody in particular. Your brand is shaped by your audience, your market position, and the specific reasons people already trust you. General principles get you a reasonable first draft. They won't get you the version that actually converts for your business.
The second thing AI still can't do is read context the way a person who's sat across from the client can. It doesn't know what the last three campaigns taught us. It doesn't know that this particular organisation has a habit of underselling itself in writing, or that this particular market reacts badly to anything sounding too polished. It doesn't know what's already been tried and quietly failed. That kind of context lives in the relationship and in the history, not in the prompt.
None of this is an argument against using the tools. I use them constantly, for drafts, for structure, for getting past the blank page faster than I used to. But the IOA work, the actual repositioning that moved the number, came from sitting with the problem long enough to know which rule to keep and which one to break. That judgment is still entirely human, and I don't think that changes anytime soon.
If your marketing feels like it's following all the right advice and still not landing, book a call. That is the gap where the real work starts.
← Back to articles