The gap between the demo and your inbox
Everyone has seen an impressive AI demo. Far fewer business owners have an AI tool they actually open every day. The difference is almost never how clever the AI is. It's whether it has been pointed at a real, repeated job in your business, and set up so you can trust what it does.
So skip the grand AI strategy. Start with one job that eats your week.
Five jobs worth handing to AI first
1. Replying to enquiries
Most enquiries ask the same few things. Do you cover my area? Roughly how much? When can you come? An AI assistant can draft a friendly reply in your tone, using your own price list and service areas, ready for you to check and send. You stay the one who presses send.
2. Turning site notes into quotes
After a site visit you have photos, scribbled notes or a voice memo. AI can turn those into a clear, itemised draft quote using your own prices, and you check the numbers before it goes out. On an operations platform I built for a remedial builder, every drafted line is checked against the company's approved price list, so the AI can't invent a price.
3. Chasing invoices and follow-ups
Polite reminders, "just checking in" emails and follow-ups after a quote are easy to put off. AI can draft them in seconds, so they actually get sent.
4. Answering questions from your own documents
Price lists, policies, warranty terms, product manuals. An assistant that reads your own documents can answer staff and customer questions from your information, not the internet, and show where each answer came from.
5. Paperwork and data entry
Summarising long emails, writing up job sheets, copying details from one app to another. It's dull work that AI is surprisingly good at, especially inside the apps you already use, like Gmail, Outlook or Xero.
Is your job a good fit?
A job is usually a good fit for AI when:
- You do it at least weekly, and each one looks a lot like the last.
- The information it needs already exists somewhere: your price list, past quotes, your emails.
- A person can check the result in a minute or two.
- A small mistake would be caught before it reaches a customer.
If a job needs fresh judgement every time, like pricing a tricky repair you've never seen before, keep it with a person and let AI do the typing around it.
How to start without the risk
- Pick one job, not ten. Choose something you do every week that follows a pattern.
- Try it on real work. Test it on last month's actual enquiries or quotes, not made-up examples. You'll see quickly whether it saves time.
- Keep a person in charge. AI drafts, you approve. Nothing goes to a customer without someone checking it.
- Keep private details out. Passwords, card numbers and sensitive customer or health details don't belong in an AI chat.
- Check it again when things change. New prices or services mean the AI needs the new information too.
Why checking matters
AI still gets things wrong sometimes, and it can sound confident while it does. That's why I test every AI setup against real examples from your business before it goes live, and check it again after any change.
The goal isn't to "have AI". It's less time spent on a job you do every week.
Start with one painful, repetitive job. Get that working well. Then expand from there.
Have it done for you, or learn it yourself
There are two ways to get there:
- Have it set up for you. I find the right job, build the AI around your real paperwork, test it and look after it. That's AI assistants and automation, priced per project after a quick chat (see pricing).
- Learn to do it yourself. If you'd rather use AI on your own work, AI coaching sets you up with Claude on your own computer and shows you how, starting with a 1:1 Kickstart.
To see what this looks like on a real job, read how AI reads a plan's legend while a person does the tracing. Or tell me about the job that eats your week, and I'll tell you honestly whether AI can help.