3 timeless things you need to know about AI (WWI 13)

If you are somewhat aware of the trends within the media and marketing landscape, you have probably heard a lot about AI. You have heard about the things people do with it, the amazing things AI is already solving, and even saw your aunt used it to make you a birthday card that you didn’t really need in your life.

Yeah, AI is moving quickly.

Half a year ago, the AI newsletters I read were simple and straightforward. Now they’re about Recursive Synthetic Improvement and evaluating first-stage retrievers at scale. Wait… what? Yeah, that’s not what I signed up for a year ago, but it is a good representation of the speed AI is developing at.

I found out that (with the right hardware) AI can do almost anything, so this week’s WWI I decided to dive deeper into the things it can and can’t do, so you know what skills to double down on and which to delegate to your best friend Chad.

Yes, welcome to Wessel’s Weekly Insights (WWI), your favourite weekly newsletter that should probably be a monthly newsletter if I keep writing essays like this.

What AI is already doing (while you’re still using it as your personal therapist)

If you have been paying attention to the AI news in the last few months, you might have seen some crazy things come by:

• In June, a man used AI to develop a working personalised cancer vaccine for his dog.¹
• In August, Moderna and Merck had a successful trial doing the same for over 1100 melanoma (a form of skin cancer) patients.²³

This is huge. Now, when you are too stubborn to use sunscreen, you still have hope, yay! (Please just use sunscreen though. A not very marketing-related tip, but regular users of sunscreen reduce the risk of developing melanoma by 40%.⁴ And being alive and healthy immediately makes you a better marketer than when you’re dead and buried… So maybe it is a marketing tip after all. Use it to your advantage.)

And oh yeah, do you remember when AI was so bad at math it couldn’t do regular calculations? Well… that’s history. Now AI is knocking out math problem after math problem.

• From the ‘first mathematical breakthrough by AI’ in May, when AI solved an 80-year-old Erdős problem.⁵
• To AI settling 10 other complex, decades-old mathematical problems in August.⁶
• And the more recent settlement of an almost century-old, million-dollar ‘Millennium Prize Problem’.⁷

Even for math, AI is crazy now! And that’s just a few of the things it is doing at the moment. I am not even talking about all the things AI can do when it’s paired with the right hardware: it can sort packages⁸, play sports⁹, and even protest¹⁰.

What AI can’t do

So that’s when I thought, if AI can do almost anything, wouldn’t it be more interesting to know three things it can’t do instead of three things it can?

Because that’s where we can still add value to the world as humans right?(And it’s also where you can still add value for a company, if you really want to keep your job and/or income…😬)

So that’s when I advised the wisdom of an intelligent young man who had already philosophised about this half a year ago…

Yep, it was me.

When I was interning at WPP half a year ago I had seen firsthand how much AI could already do, especially in marketing and media. This led me to think about what it could and couldn’t do, and where I could still provide value as a human (especially in the creative space).

This is what I wrote down back then (translated to English):

“The few things AI can’t do:

1. Innovate (the innovation comes from the people, not the AI)

2. Colour outside of the lines (AI does what is asked of it, people dare to do other stuff and follow their intuition)

3. Foster real social connections (work is getting more social, I love to see sociality is increasingly valued in media and marketing)”

And to my surprise, these things are still accurate to this day. Here is why:

1. AI doesn’t innovate, it iterates.

When you look at the cancer vaccines and the math breakthroughs, it may look like AI is innovating, because it’s solving things that had never been solved before. But not every breakthrough is an innovation. Let me explain why.

Take the melanoma vaccines for example. The problem we had was already clear. Scientists already knew the target, and the lock-and-key mechanism needed to neutralise cancer cells. Finding the exact molecule that neutralised the cancer cells however, was the bottleneck. Finding it meant testing millions of possibilities, something that would take human researchers ages. AI could cycle through millions of combinations in seconds until it found the one molecule we needed, and that’s exactly what it did.

It’s not innovative, it’s just (a lot) quicker at sorting through the puzzle pieces than us.

2. Give AI a task and it does it, give a human a task and it does something with it.

Let’s say we get briefed with an awareness campaign.

• AI looks at all the data it has and comes up with the best idea to generate awareness (within the limits and boundaries we set for it).
• We (humans) think: “fuck it, wouldn’t it be funny if… [insert whatever stupid idea we came up with]. No complete data to back it up, just some crooked, biased or completely self-made-up logic (and a bunch of hope) behind it.

Chances are AI will consistently get good results, but it will never/rarely do amazing.

Humans will perform rather inconsistently, but once in a while they will absolutely knock it out of the park. Not based on data and/or skills, but based on pure courage, hope and ownership.

AI won’t come up with something that isn’t backed up with data and/or logic in any way.

People will (and even better/worse, they will be absolutely certain it’s a great idea). This ‘out-of-the-box thinking’ is the only way to get extraordinary results.

Call it uniformed optimism, but in the end, people show something AI does not have: the hope and courage to go into the unknown and do everything in their power to make something work in the end. The idea might not be genius, but we take responsibility for it either way. AI can’t do that.

In addition, sometimes requests are just so stupid that we need someone to push back against them. Let’s say you want to make a TV commercial to reach Gen Z and you ask for help. AI will immediately draft you a script and get you going, but a great marketer will push back and tell you maybe a TV commercial is not the best way to reach Gen Z at the moment. This way, our absolute inability to blindly follow directions can become a selling point instead of a bottleneck.

3. People want to work with other people, not with bots.

When you pay for a service, you expect it to adhere to a certain quality. AI doesn’t feel this responsibility, (most) humans do.

When you hire someone, you expect them to have a certain expertise. You evaluate the person behind the work before you buy anything from them. You expect them to inherit a certain responsibility of providing good work. You need to trust them. You can’t hold a bot or AI responsible for something, but the human behind it, you surely can.

I saw this within my internship as well. You can have the best ideas, but clients need to trust (and like) the people behind them first, otherwise you ain’t selling shit.

Conclusion

To conclude, our role as people will shift more and more towards curation and critical thinking. If you’re working in the creative industry, companies will probably be less and less interested in the things you can do, and more and more interested in the responsibility you take for your results.

As creatives, it’s increasingly important to push back against bad ideas because AI won’t. But it’s also increasingly important to trust your gut feeling, do things that have never been done before, and make it work yourself.

Ignorance and pure hope/courage are the only things that still differentiate us from AI.

How I am implementing this:

1. Running more fuck-it experiments and seeing what sticks

AI isn’t going to come up with the absolutely absurd ideas I am coming up with, so however bad they are, it’s at least a way to differentiate myself from all the AI slop.

2. Asking why

AI won’t question your ideas if you don’t ask it to. I am questioning everything more and more, and I actively take more time of to reflect on everything I am doing (and why I am doing it in the first place).

3. Doing hard shit

AI can do a lot, but I am still trying to do most things myself. Writing my own essays and doing my own research might be harder and less time-efficient, but I’m doing it anyway to keep my brain sharp.

(No one is taking my knowledge and skills by cutting the power.😝)

Sources:

1. https://www.independent.co.uk/news/science/ai-cancer-vaccine-dog-chatgpt-b2991945.html ↩︎
2. https://www.merck.com/news/merck-and-moderna-announce-phase-3-interpath-001-trial-of-intismeran-autogene-plus-keytruda-met-endpoints-of-recurrence-free-survival-rfs-and-distant-metastasis-free-survival-dmfs-in-patient/ ↩︎
3. https://www.youtube.com/shorts/vr6l_QNGtaM ↩︎
4. https://www.cancer.org.au/cancer-information/causes-and-prevention/sun-safety/sunscreen/latest-research ↩︎
5. https://openai.com/nl-NL/index/model-disproves-discrete-geometry-conjecture/ ↩︎
6. https://www.forbes.com/sites/jonmarkman/2026/08/03/openais-astra-solved-10-decades-old-math-problems-for-just-2000/ ↩︎
7. https://openai.com/nl-NL/index/navier-stokes-solution/ ↩︎
8. https://arstechnica.com/ai/2026/05/the-internet-cant-stop-watching-figure-ais-humanoid-robots-handling-packages/ ↩︎
9. https://apnews.com/article/china-humanoid-robot-games-us-86cb8e310843151a77057e4cb764b4e2 ↩︎
10. https://www.instagram.com/p/DdGIKiKgJPH/ ↩︎