The One About The AI, Bro!

AKA: "Where The Heck Has This Newsletter Been, Bro?"

Feb 06, 2026

Sometime in Q3 last year, I began to feel… left behind.

LinkedIn, Threads, and Reddit were swarming with posts about AI. Everywhere I looked, another “AI-bro creator” was crawling out of the woodwork.

They seemed to know all the models with the intimacy of someone who’d spent thousands of hours with them. (At that time, I hadn’t calculated that those models hadn’t existed for thousands of hours!)

They knew all the tools. From the hits to the obscure. As someone deeply rooted in the Google ecosystem and tied strongly to Gemini, I suddenly felt like a geriatric Millennial in a world of Gen Alphas.

They knew the prompts. They’d hacked productivity. Hacked consulting. Hacked creativity. Hacked life.

All I’d hacked was… well, I still remembered most of my passwords.

As someone obsessed with learning as a way to grow, this was unacceptable.
So I lit a fire under my bum.

I put a daily, 30-minute block on my calendar called “Meddle With AI.” And followed it diligently for weeks.

I’d face a problem during the workday, wonder if AI could solve it and note it down. Then I’d spend the session figuring out how to solve it with AI.

Most days, that 30-minute window would stretch into 45, then 60… and before I knew it, three hours had flown by.

Then, out of the blue, I had my first breakthrough.

Interlude
I’ll be using the words “coding” or “programming” quite liberally below. For the purposes of this discussion, you may assume that the coding language I’m referring to is English.
We now return you to your regularly scheduled… er… programming.

The Breakthrough

It was September 21st. I was camped near our office barista, gulping coffee and hammering away at my keyboard, crafting some Gems — Gemini’s version of custom GPTs. Five, to be precise.

I’d write some code, test it in action, make tweaks. Switching between Gems, chaining their outputs together.

Two hours later, I got goosebumps.

I sprinted down to my wife Saav’s desk (she’s also in Google Marketing) and started babbling about what I’d built. We both knew I was on to something big.

Cut to October. A series of presentations later — first to my manager, next to my skip — and I had a project.

A creative AI tool, built entirely in Gems. Today, it’s being formally incorporated into our creative workflows across APAC. It’s used by nearly 150 marketers and comms pros, with the potential to fit into our Global MarTech stack.

All of a sudden, I’m a prompt engineer / creative AI architect / product manager.

Who’d have thunk?

It’s been quite the journey since.

Beyond the creative tool, I’ve been experimenting with my own “Chief of Staff.” Artoo (yes, I’m a geek) is coded as an on-demand Gem and a series of automated Workspace Studio Flows. It pings me with reminders and content, helping keep me organised and laser-focused.

I’m now rethinking our team’s entire operating system using no-code platforms.

But this post isn’t about chest-thumping. (Okay, maybe a little.) It’s about what I’ve learned. So you can learn in turn.

You need to learn how to ask questions. Christopher Smith taught me this, through his excellent book To Question Is To Answer, and a session he did for our team last year. The better you question the AI, the better it helps you think.

AI is seductive. It responds to you in exhausting detail, in highly articulate language, with a tone of great authority and confidence. You want to buy what it’s telling you. Be warned — there’s generally quite a bit of snake oil hidden in those responses. How you question it, how you challenge it, is critical. This is where judgement — creative and otherwise — comes into play. Bringing me to my next point…

Slop in, slop out. Antti Toivanen, MD of Superson APAC, came up with this phrase in a recent fireside chat we did together. It’s probably the most eloquent way to say it. The quality of output depends on having the right people in the room, with the right data for the AI to work with, asking the right questions of AI’s answers, and really directing it as opposed to accepting it blindly. Which leads me to…

AI isn’t creative on its own. Ben Affleck said it perfectly: It can get you some ideas, some starting points, but it can’t write you Good Will Hunting. Because AI defaults to the average. And our job as creatives is to escape the trap of the average. So…

Being good at prompting alone isn’t enough. You’ve got to teach AI to think the way you want it to. My colleague (and fellow AI hobbyist) Debbie refers to this as “meta-cognition”. To get the AI to think the way you want it to, you have to understand the way you think. And then arm the AI to do so, in turn. What I’ve been able to do with the creative tool I’ve built is teach it how I think.

There are some things it’s impactful at. And some things the corporate world believes it’s impactful at. It took my team and Antti’s five days of being holed up in a conference room, with thousands of generations, to deliver our first AI-generated piece of content. And that was merely 80 seconds. I know other teams have taken as much as two months to deliver a fully AI-generated commercial. Consuming millions of tokens — at a cost — through hundreds of thousands of generations. Did it save time as opposed to traditional production? Probably not. But can it save you time — with quality output — in the upstream thinking? Yes. We’ve seen that in the pilots we’ve run with my creative tool.

As a creative, it’s not cheating to use AI to help you think. It wasn’t cheating back in 2003 to scroll through Getty Images for creative inspiration for a print ad. So why should it be anathema to use AI to support you today? It’s simply a tool, like your computer, like those creative frameworks all of us grew up with. A tool that can help you break the shackles of a blank page, get you started, get those creative juices flowing faster.

You will spend more time scrutinizing and critiquing than creating. This is true. And I do feel a bit of sadness here. However, in that scrutiny lies the genesis of further creation. Nothing has stopped me from blending a few thought starters from AI, or building on them, or playing Creative Director and bombing them…or simply starting with my own idea and speedily refining it to sell internally with AI. And the end of the day, I’m paid to deliver great work. Nobody’s paying me less (yet) if I use AI to do it.

Finally, AI is going to reshape you think. I read about a Harvard-MIT study recently. There are three kinds of AI users:

The Delegators, who outsource everything, including judgement and agency, to the AI. And lose those skills over time.
The Cyborgs, who get good at using AI to improve their work product or workflows, but don’t really evolve on the whole.
The Centaurs, who use the AI to do better work, and learn in the process.

I’m trying hard to fall within 2 and 3 for the most part. Working with the creative tool has taught and re-taught me things about marketing, creativity and effectiveness, enabling me to build fresh knowledge and skills.

To paraphrase Thanos: AI creeping into the workplace is inevitable. You either accept it, adopt it, and adapt — or fade away. 💨

The way I think about my work has changed significantly. I no longer manage just a human team — I manage and creative-direct a hybrid squad of humans and machines.

Understanding people isn’t enough.
Understanding the tech isn’t enough either.
The key to success is understanding your own ways of thinking.

That is the insight.

Samit