AI Can Do All Your Marketing. Don't Let It.

I’ve been doing this for 10 years. I’ve spent my whole professional career in some capacity, buying ads. Uploading images I didn’t like, transferring copy I didn't write, and spending on whatever platform I was told to as an exec lacking autonomy. 

I’d frankly fallen into the advertising world. During my post-uni job hunt I spoke to a close friend of my sister’s, who had started working at Jellyfish in its early days. She asked, “Could I do numbers, and could I do words? If so, I should look at the grad scheme”. Having just (about) graduated from the University of Bath with a Bsc in International Management & German, covering a variety of marketing, branding, statistics and economics modules, I shrugged and said “yeah i guess so”. 

And so I took the path of least resistance, kicking off my career as the most junior of juniors in the media buying team at Jellyfish in Reigate, before moving up the much more braggable Shard office. Upload the creative, analyse the creative, remove the creative. God life’s repetitive.

Now, I’m part of the 99% from our industry who wish those manual processes could be automated (some people enjoyed the process). If you ever tried to pull reports from DV360, or DBM for the dinosaurs amongst us, it was akin to pulling teeth. The pivoting, sorting and formatting performance reports was a painful process in a buggy system, and frankly looking back a real waste of literally everyone’s time. 

It’s rarely insightful, who cares how much CTR went up or that my CPM fluctuated 0.002% day on day.  The job was 90% manual lifting, 10% thinking. Honestly on some days I was probably 1% thinking.

At the time however that was how it worked. The big network agencies were in their prime with delightfully siloed process, lots of manual work and little joined-up thinking. 

Fast forward to early 2025, and the wonderland of AI labour saving is in full swing. Creative is the key to Meta Ads performance. Whack as many creatives into your ad account, let Meta figure it out, and you'll win. Way less work than a decade of banging my head against a wall.

But in the past year, it's gone further than anyone expected. Pull a report from Meta? Their recently acquired LLM, Manus, integrates into your ad account and will do this in seconds. Need to test 5 new headlines? Meta can generate these instantly using its Advantage+ features. In fact give it a chance, it will generate and test literally every element of an ad across headlines, descriptions, background music, you name it.

This has led to advertising lacking something. The humanity of advertising has been drained, feature by AI feature. We’ve been caught in the trap of thousands of shallow ad variations. Is a different colour CTA really going to make the difference to your ad, or is there a human spark of genuine thinking time lacking?

I truly believe there’s been another quiet evolution. To remain relevant now as an industry and a professional, you simply can not let AI take the wheel. It’s just part of the engine now, but you need to be there to steer away from the ditch, despite AI thinking it’s a shortcut.

In reality my focus as a media professional has taken a complete volt-face. I’m no longer a buyer in the traditional sense. I’m an insights consultant. I take context from a brand, and insight from an AI optimised campaign, and build the product and creative strategy in one process. The best strategists have taken the previous 10% thinking, 90% manual working and switched the numbers. Creative is what drives performance, but insight is what drives creative.

If you let AI, it will do everything for you, including building thousands of sloppy creatives in seconds. It's iterating, not originating.

Against some common modern day opinion, I actually like the idea of a customer persona to solve this problem. It’s how we give the AI the guard rails in which to optimise. The difference now is the quality of the data in building these personas, and the execution of speaking to them. The problem is no longer tinkering with targeting to try and reach this needle in a haystack, it's about crafting the best package of prompts, product messaging and imagery that will resonate with your audience. This all comes down to insights, understanding your customer, and knowing your business, to a depth that AI simply can not currently match.

So how can all this really help the modern marketer?

Let AI do the heavy lifting. Trust the outputs from your analysis prompts, however now your responsibility extends to overlapping these insights with business context. 

Compare the product performance to your brand’s marketing calendar. Where are the opportunities for creative diversity? 

We recently unearthed strong performance from 3 Gulf states for certain products of a premium watch accessories brand. AI would just spend more here. The human marketer can combine cultural nuance to know Eid, the biggest gifting opportunity of the calendar in the region is round the corner. We drove the top 5 giftable products into Nanobanana with strong performing gifting prompts from the previous 6 months and surprise, a new set of diverse creative variants, built with AI data and human comprehension.

Back to that question all those years ago. That question my sister’s friend asked actually holds a lot more value in reflection than I first thought. To be any good in the world of modern media buying, you need to be as adept at the quantitative work of data review, as you are translating that into qualitative messaging.

You can't optimise your way out of a hole. You have to think broader. The numbers, and the words.

Before you go,

Please let us know what we can improve, and what you really liked.

If you enjoyed it, why not forward this on to just one person who you think would find this helpful or interesting. 

Beyond that, have a marketing problem? Reply to this email and let’s chat!