If you sell your own creative work, AI for creatives is easy to get wrong in both directions. One camp says use it for everything. The other says touch none of it. Neither helps the florist answering her fortieth wedding enquiry on a Sunday night.
My position is simple. The making stays yours. The business around the making is where AI earns its place: enquiries, quotes, briefs, onboarding, content and the pile of small documents that eat your week.
What AI should and shouldn't do in a creative business
- AI drafts the first reply to an enquiry. You read it, fix it, and send it in your voice.
- AI turns a messy client message into a clear brief. You confirm it with the client before work starts.
- AI builds your quote from your price list. You decide the price and the scope.
- AI repurposes your own words into posts. You approve what goes out under your name.
- AI does not design, illustrate, style or make the thing you're known for. That is the product.
If you do want AI to generate images for any client deliverable, that's a separate decision, and a bigger one. Treat it as a conversation with the client, not a default.
The tension you should look at honestly
Wariness about AI among creatives is reasonable, so here are the real issues in plain terms.
- Copyright and training data. Arts Law Centre of Australia has a plain-English information sheet on AI and copyright. It says Australian law doesn't clearly settle whether works can be used to train AI, and that protection generally depends on a human author contributing independent intellectual effort. It was written in 2022 and the law is moving, so check for updates. The Australian Copyright Council also publishes an information sheet on the topic.
- Client contracts. Some clients, especially larger ones, now put AI clauses in agreements. Read yours before you paste any client material into a tool.
- Your own style. Don't upload your portfolio to a tool unless you've read what the terms say about how uploads are used.
- Disclosure. Decide your policy once, write it down, and use the same words every time.
I'm not a lawyer and this isn't legal advice. For anything contractual, ask a lawyer or your industry body.
What to tell clients, in three sentences
Put a short note in your onboarding pack. Something like: "I use AI tools for admin, such as drafting emails and organising project notes. The design and making is done by me. If a project would involve AI-generated imagery, I'll ask you first."
Adjust it to what's true for you. A clear sentence on day one is worth more than a defensive one after a client finds out. It also tells nervous clients that you've thought about it, which is often what they were hoping to hear.
Workflow 1: enquiry replies that sound like you
Collect your best five past replies. Give them to Claude or ChatGPT with your price range, your lead times and the questions you always ask. Ask it to draft replies to new enquiries in the same voice.
You check: the price, the dates, and that it hasn't promised anything you don't offer. Send it yourself.
Workflow 2: quotes and proposals from your own price list
Paste your packages, add-ons and terms once. For each enquiry, give the AI the details and ask it to assemble the quote, flagging anything not covered by your price list instead of guessing.
You check: scope, exclusions, deposit and revision rounds. Your terms should be pasted in from your own document, never generated.
Workflow 3: client onboarding and briefs
Turn your first-call routine into a short questionnaire: what the work is for, who sees it, deadlines, references they like, what they hate. After the client answers, AI summarises it into a one-page brief and a list of open questions.
You check: send the brief back for the client to confirm. That confirmed brief protects you in a revision dispute, and AI just made it quick to produce.
Workflow 4: project admin and SOPs
Write down how you run a typical project once, even roughly, by voice note. AI turns it into a checklist: handover files, invoice reminder, testimonial request, thank-you note. Next time, the process is a document, not something in your head.
You check: run the checklist on one real project before trusting it.
Workflow 5: content from what you already say
Creative businesses run on showing the work and the thinking. Rather than staring at a blank caption box, let AI interview you about a finished project, then draft from your answers. My post on the interview method for AI content explains how, and repurposing one video into social posts covers the production side.
You check: that it's your voice and that nothing about a client's project is shared without their say-so.
The small stack worth learning
- Claude or ChatGPT: drafting, summarising briefs, interviewing you for content
- Your existing quoting or invoicing tool: keep it, and paste its output into the quote workflow
- Google Drive or Notion: one home for your price list, voice examples and SOPs
- Canva or your design software: check what AI features are switched on inside tools you already pay for, and read their terms
When a tool does need building
Sometimes a creative's expertise deserves a bespoke tool. Sally Prendergast, a dance teacher and studio owner, built two AI apps without coding. It's worth reading for how she handled the sensitive parts: she used Claude to find the holes in her idea first, and her app includes an AI policy studios can publish plus an option to label images as AI-generated. That's the same disclosure habit I'd suggest for your own business.
Your first build: the enquiry reply
Pick the one thing you answer most often. Write down your five best past replies, paste in your price range, and ask for a draft reply to a real enquiry from last month. Edit it until you'd happily send it. Save the instructions that produced it.
Once it works three times, you have a system. If you'd like to build it with other business owners rather than alone, see Wright Mode.