Small business owners get two messages about AI at the same time. One says it will change everything and you're already behind. The other says it's an overhyped toy that writes bland LinkedIn posts.
Both are a little bit right, which is why they're both unhelpful. Here's a more useful way to look at it, from a studio that uses AI every working day.
Start with your week, not with a tool
Before you sign up for anything, take ten minutes and write down what actually eats your time. Typically it looks something like this:
- Answering the same customer questions by email, phone and DM.
- Writing posts, newsletters and product descriptions.
- Meetings, and the notes and follow-ups after them.
- Quotes, invoices, scheduling, admin.
- Running ads, or worrying that you should be.
Now circle the tasks that are repetitive and text- or pattern-based. That's where AI earns its keep. The tasks that need your judgement, your relationships and your taste stay with you.
Worth paying for now
These are mature, cheap relative to the time they save, and low-risk if you keep a human in the loop.
A general AI assistant
A paid plan of ChatGPT, Claude or Gemini is the single best-value AI purchase for most small businesses. Use it to draft emails, rewrite clunky text, summarise documents, brainstorm campaign ideas, translate (with a native check), and think through problems out loud. Treat it like a fast, well-read junior colleague: brilliant at first drafts, never the final word.
Meeting notes
AI note-takers record, transcribe and summarise calls, then list the actions. If you spend a few hours a week in meetings, this alone can give you an afternoon back. Do tell people you're recording.
Drafting customer replies
Most customer questions are the same twenty questions. Write your best answers once, then let AI help you adapt them to each message. You review and send. Faster replies, same human tone.
The AI already inside your ad platforms
Google's Performance Max and Meta's Advantage+ campaigns are AI-driven by design: they decide placements, audiences and bids automatically. They can work very well, but only if you feed them good creative, accurate conversion tracking and a clear goal. Give them bad data and they'll optimise efficiently towards the wrong thing.
Image and video clean-up
Background removal, resizing, upscaling, retouching and cutting one video into several formats used to cost hours of a designer's time. Now it's minutes. This is AI at its least controversial and most useful.
Worth testing, carefully
These can pay off, but they can also embarrass you in front of customers. Pilot them small, watch them closely.
Website chatbots
A chatbot that answers "Do you deliver to Crete?" at 2am is useful. A chatbot that invents a returns policy is a liability. If you add one, train it only on your own approved content, make it obvious that it's a bot, and make it easy to reach a human. In the EU, being open about AI interactions isn't just good manners. It's becoming a legal expectation.
AI-generated product imagery
AI can place your product in a lifestyle scene in seconds. Lovely for mood boards and social backgrounds. Risky if the product itself is altered: colour, texture, size or details that don't match what arrives in the box. That's a returns problem and, potentially, a consumer-law one. We cover this in Product photography for brands: DIY, studio or AI?
AI agents
"Agents" are AI tools that don't just write but act: book meetings, update your CRM, chase invoices, post content. Some of this works well already for narrow, well-defined jobs. Give an agent a small, low-risk task with clear rules first, check its work for a few weeks, then expand. Don't hand it the keys to your bank account or your brand's Instagram on day one.
Mostly hype (for now)
- "Fully automated marketing." Software that promises to plan, create, publish and optimise everything with no human input. What you usually get is a lot of content nobody asked for.
- AI instead of a team. AI makes a small team mighty. It doesn't make a team unnecessary. Someone still has to decide what's worth doing.
- Tools that promise results without your input. If an AI tool doesn't need to know anything about your business, your customers or your goals, ask yourself how good its output can possibly be.
A simple 30-day plan
- Week 1: Pick one general AI assistant. Use it daily for writing and summarising. Notice where it helps and where it doesn't.
- Week 2: Write down your twenty most common customer questions and your best answers. You've just built the foundation for faster replies, a better FAQ page and, one day, a chatbot.
- Week 3: Add an AI note-taker for meetings, or an image tool for content. Just one.
- Week 4: Look back. Count the hours saved. Keep what worked, drop what didn't, then pick the next task.
Notice what's missing: a big upfront investment, a twelve-month roadmap and a consultant's slide deck. Start small, measure, expand.
One rule above all
Keep a human between the AI and your customer. Not because the machines are malicious, but because when they're wrong, they're wrong confidently and in public. It's the same rule we run our studio by. Here's the long version.
If you'd rather have someone set this up properly across your brand, content, website and ads, that's our "Make it AI-ready" work. See our services, or start a conversation.