Not a coder
I’ve written before about the night I built my first website. A young artist I know had made a brilliant piece of work, and I wanted to give it a home online. I opened Claude Code in the terminal, understood almost nothing, and screenshotted every error into Claude chat to ask what it meant.
I’m still not a coder. A lot of what scrolls past in the terminal still confuses me. I’ve got better at it, but I still go back and forth between Claude chat and the terminal, asking for things in plain English. And I still think the terminal is the right place to build. It’s where the real work happens, and it makes you understand what you’re asking for.
Pen and paper
What I didn’t expect was how much of my old job would carry over.
Any time I put a strategy deck together, I started with pen and paper. I mapped out the sections. Sometimes I drew every slide and wrote what each one had to say. You’re building the blocks of a story, and the logic has to hold from one block to the next.
A website is the same. Strategy helps me thread the logic together: how someone moves through the site, what they need at each point, and where it’s likely to go wrong before it does. It also helps with the unglamorous part, which is managing the process well.
How a build goes
Every build now follows the same order.
- A skeleton first. What are the key sections?
- Then the sections within the sections.
- Black and white, content first. All the words down before any design.
- Then the look. I review designs and aesthetics I like, and pick a palette.
- Then the extra layer. The features, animations and functions that make it feel alive.
The foundation is always the content. The clever features come last.
It’s tempting to start with the shiny part. But a beautiful site with weak content is a nice frame with nothing in it.
When it breaks
What goes wrong is usually a glitch. Something that should work doesn’t, and fixing it takes far longer than building it did.
When we finally get to the end of one, I ask the same question every time: what did we learn, and how do we make sure it doesn’t happen again? The answer goes into a guidelines file that the AI reads before every session. Every mistake becomes a rule. The file gets longer, and the builds get smoother.
No solo runs
The rule I care about most is simple. The AI doesn’t go off on a tangent or take a solo run on anything.
I’m the arbitrator. If there’s a choice to make about the approach, it comes to me. I decide how we do it, and I decide whether the quality is there. That isn’t about control for its own sake. It’s that the AI can be very confident and still be wrong, and I’m the one who answers for the result.
I dictate the approach. I dictate the quality.
A team that reports back
These days I don’t just build with one AI. I work with a small team of AI agents in a platform called Buzz, each with its own role, from research to copy to design. A project manager agent coordinates them. I give the instructions, and I ask the project manager for status.
What I like is that I can also go straight to any one of them, even mid-task, and change course. It feels less like using software and more like running a team. Which, as a strategist, is the part of the job I already know.
Technology strategists
I think every strategist coming into the business now has to accept something. We’re technology strategists as well as everything else.
You still have to do the reading and learn the trade. You still have to learn how to think about strategy. That doesn’t change. But the delivery of strategy is changing fast, and a lot of it for the better. Communications will change with it. The strategists who stand out will be the ones who can think with this technology, and use it creatively.
Who controls the gears
Agents will take on a lot of the admin and the lower-level work. They might even do some of the design. But a person will always be needed to communicate the strategy, iron out the creases and answer the hard questions.
Someone has to sit between the strategy and the creative work. Someone has to inspire the team. That’s still a human job, and I don’t see it changing.
It reminds me of Eric Morecambe telling André Previn: “I’m playing all the right notes, but not necessarily in the right order.” The agents can play all the notes. Human expertise decides the order.