When the AI can write, what’s left for me?

Notes from a communicator learning to build with machines

Jul 06, 2025

If you work in marketing or communications and you’ve been wondering what AI means for your craft - I’m right there with you.

This isn’t a tech tips newsletter or a catalogue of prompt hacks. It’s a space for us - the people who shape language for a living - to think out loud about how our work is changing, and what that change might demand of us.

The first bad prompt

I wrote my first prompt in August 2021. It was on Jasper.ai, and I was exploring ways to accelerate content creation on the newsdesk of a sports app I was building called The AllStar (still available in the Apple and Play stores).

The newsroom was lean. Just me, one other editor, a handful of talented freelancers, and a wire feed. Budgets were tight. Turnaround times were tighter. So we explored automation and AI.

My first prompt was about the history of MMA gloves.

It sucked.

The prompt had no guardrails, no context, no thoughtfulness. The output was wildly inaccurate and bizarre. My next few prompts weren’t much better. Jasper didn’t stick. We let our subscription lapse and moved on to rules-based content automation, which served us better at the time.

But the idea stuck. Even if the tools weren’t ready, something about the approach - using structured inputs to generate consistent, on-brand outputs - felt like a glimpse of the future. Not a gimmick. A signal.

Leaving the newsroom behind

Before all that, I was at Bloomberg.

I spent nearly 20 years in journalism: reporting, writing, editing, leading teams. At Bloomberg, the bar for clarity, accuracy and speed was high. You didn’t get to hit publish unless every detail had been verified, every sentence sharpened.

But after two decades in the newsroom, I left. I joined a crypto startup as its comms lead, then moved into gaming and sports media. I founded a consultancy. I started building.

In that world, content didn’t sit in style guides and editorial calendars. It moved fast. It was measured in clickthroughs, shares, sign-ups, not column inches or bylines. And the work shifted too: from slow perfection to rapid iteration. From craft to system.

From content to systems thinking

At The AllStar, that shift became literal.

We weren’t just building a content pipeline. We were building an app. A product. The goal was to create the perfect second screen for sports fans to turn to while watching a game, whether on TV or in the stands. The content had to enhance that experience in real time: smart, contextual, snackable.

To achieve that, I had to start thinking in systems.

How do we go from data to insight to on-screen content in seconds? How do we streamline contributor workflows? Where do automation and human editing intersect?

And when I came back to generative AI tools later, it all snapped into place.

I wasn’t just trying to get the AI to write for me. I was designing a system around it, which eventually became the foundation of my work.

I use this system to produce long-form SEO content for gaming clients - in about a tenth of the time it used to take. A system that safeguards tone, logic, and structure. Fast, but not careless.

That’s when I realised I wasn’t really a writer anymore. Not in the old sense, anyway.

Not really a writer now?

Now, part of my process is prompting. Orchestrating. Reviewing.

I still care deeply about words - but I spend less time typing them from scratch, and more time thinking about the ideas behind them. More time guiding the systems that help shape them.

That doesn’t mean the output is fully AI-generated. Far from it. I use AI as a creative partner - to outline, explore, refine. The ideas are mine. The structure is mine. The judgment is mine. But the tools help me get there faster, and think more broadly than I could alone.

That’s how I created this article, too - through a back-and-forth with AI, shaped by my own edits, my voice, my experience and my personal anecdotes. It’s not hands-off, and it still took me about two hours to produce. A lot of that time is thinking about the nuances of what I want to express, especially with sensitive sections like this one. The robot can't do this for me.

So, writing with AI is still a highly hands-on process. Just... different.

What still belongs to us?

As a creator, your value hasn’t vanished. It’s just moved.

Tone, judgment, audience understanding, ethical control - all of that still rests with us. But the tools we use now ask us to work differently. Not just faster, but more structurally. More systemically. With more foresight.

In future posts, I’ll wrestle more with this tension: how much ownership or authorship should we claim when machines help us write? And how much do our audiences expect us to disclose? There’s no single answer - but it’s a question worth sitting with.

And I know I’m not the only one.

If you work in communications, maybe you’ve felt it too. You write less. You brief more. You don’t type every word, but you shape the outcome.

Why I’m writing here

This Substack is where I’ll try to make sense of that shift.

Not to offer hot takes or hype. Just to reflect, share, and occasionally question what’s changing in our field. I’ll write about prompting, sure - but also about process, identity, team roles, and the quiet panic that sometimes comes with all of it.

It will be grounded in real conversations with marketers and communicators navigating these shifts in real time. From team leads rolling out AI systems - or told they must - to content creators exploring GenAI by choice, or necessity.

This effort will be informed by my work. At UnMute, we take an AI-first approach to content creation. Full-stack prompting, custom GPTs, editorial workflows that blend speed with brand safety. But this isn’t a company blog.

It’s a personal one. For people like me. Maybe people like you.