There's a lot of noise about AI right now. One-person billion-dollar businesses. Every job being automated. Entire industries disappearing overnight. Most of it is either breathlessly optimistic or quietly terrifying, and very little of it is useful if you're running a small business trying to get more done with a lean team.

So here's our honest take. Not a prediction about the future. A practical account of what we're actually doing right now — and what it's changed.

The shift that matters

For most of the last decade, AI in business meant chatbots and autocomplete. Useful in a narrow way. Nothing that fundamentally changed how you operated.

That's changed. The shift from AI that answers questions to AI that completes tasks is a genuinely significant one. The difference between "what should I put in this email" and "draft, review and schedule this email based on what you know about this client" is not a small step. It's a different category of tool entirely.

Claude — specifically Claude Code and Claude's Skills framework — is what we use. Not because it's the only option, but because it's the one that's made the most practical difference to how we actually work.

What Claude Skills actually are

Skills are reusable workflows. Think of them as standard operating procedures that an AI can actually follow — reliably, repeatedly, without being reminded.

The way it works: you write a CLAUDE.md file for a project that tells Claude everything it needs to know. Who the client is. What their tone of voice sounds like. What the deliverable looks like. What good looks like and what to avoid. Then you build a skill — a set of numbered instructions — that Claude follows every time that task runs.

The result is that a task you used to spend 45 minutes on every week now takes five minutes to review and approve. Not because the quality dropped. Because the groundwork is done.

What we've actually built

Here are five things we use regularly that have changed how we work.

Client briefing summaries. After every new client conversation, we feed in our notes and Claude produces a structured briefing document — problem statement, objectives, suggested service mix, questions to resolve before work begins. What used to take an hour takes ten minutes of editing.

Social media content drafts. For clients on social media retainers, we've built skills that understand their brand voice, their audience, their content categories and their posting rhythm. First drafts come back formatted, captioned and ready to review. We edit, approve and schedule. The creative judgment is still ours. The mechanical production isn't.

Proposal first drafts. When a new enquiry comes in, we feed in what we know about the client, what they're asking for and what we'd recommend. Claude produces a structured proposal draft in our voice, with the right service mix and the relevant context already included. We refine it. We don't start from a blank page.

Weekly priorities. Every Monday we run a brief that pulls together what's outstanding across active projects, what's due this week and what needs a decision. It functions like a chief of staff — not because it makes decisions, but because it organises information in a way that means we start the week knowing exactly where to focus.

Research synthesis. Before we start any strategy project, we need to understand the client's market, their competitors and their customers. What used to be half a day of browser tabs and notes is now a structured research brief that Claude compiles and organises. We still make the judgments. We make them from a much better starting point.

What it doesn't do

It's worth being honest about the limits, because the hype tends to paper over them.

Claude doesn't replace creative judgment. The reason our work is good is because we know what good looks like — from thirty years of doing this. AI can produce a competent first draft. It cannot produce the insight that comes from genuinely understanding a client's business, their market and their customers.

It doesn't replace relationships. Clients work with us because they trust us. That trust is built in conversations, in the way we handle difficult feedback, in the fact that we answer the phone. None of that is automatable and none of it should be.

It doesn't replace strategy. Knowing what a client actually needs — as opposed to what they asked for — is the most valuable thing we do. That requires experience, honesty and judgment that no AI currently has.

What it does do

It removes the mechanical work that used to sit between having a good idea and executing it. The formatting, the structuring, the first drafting, the repetitive admin. The things that took time and attention without requiring creativity.

For a small studio, that's significant. It means we can take on more work without taking on more people. It means the work we do deliver gets more of our actual attention. It means we spend less time on production and more time on the things that actually make a difference.

The practical starting point

If you want to start experimenting, here's where we'd begin.

Pick one task you do repeatedly every week that follows a similar structure each time. Write down the steps you go through. Give Claude the context it needs — who it's for, what the output looks like, what good looks like. Build a simple skill file. Run it once. Refine it.

You don't need to understand code. You don't need to restructure your business. You need one task, one hour, and a willingness to iterate.

The businesses that will get the most from AI over the next few years aren't the ones that wait for a perfect solution. They're the ones that start small, learn what actually works, and build from there.

We're still very early in this. But early is a good place to be.