Priya manages operations at a 25-person air-con servicing company. She already uses an AI chatbot every day to draft replies to customers. But every reply means copying a WhatsApp message into the chatbot, pasting the answer back, then updating the job sheet by hand.

"Isn't there a way for this to just happen?" she asks.

That's what AI automation is for. This guide explains how it works and how an SME can set up its first automated workflow, without hiring a tech team. If you're still deciding which task to start with, see our list of 7 business processes SMEs can automate with AI.

I. Four levels of AI help, in plain English

1. Using an AI chatbot by hand. You type a question or paste in some text, and you get a draft back. You move all the information yourself. This is where Priya is today.

2. Rule-based automation. "If this happens, do that." For example, when a web form is submitted, add a row to a spreadsheet and send a confirmation email. It's reliable, but it can't read a message or make a judgement.

3. AI-assisted workflow automation. Rules move the information, and AI does the reading, sorting and drafting in the middle. A person approves the result. For most SMEs, this is the sweet spot.

4. AI agents. You give the AI a goal and access to some of your tools, and it works out the steps itself, such as checking a calendar and proposing a time slot. Agents can do a lot, but they need tighter controls.

II. How it works: one enquiry, start to finish

Here's how Priya's customer enquiries could flow once automated:

  1. A customer fills in the website form: "Aircon leaking water, 3 units, Tampines."
  2. An automation tool picks up the form and sends the text to an AI model.
  3. The AI labels it (repair, urgent, three units, east) and drafts a reply using the approved price list.
  4. The draft appears in the team's inbox for review.
  5. A staff member checks it, edits it if needed and sends it. The job is logged in the spreadsheet automatically.

The AI never sends a quote on its own. Later, Priya might let simple acknowledgements go out automatically, while quotes still get a human check.

For step 3, the instruction given to the AI could be as simple as this:

"Classify this enquiry as Repair, Servicing, Installation or Other. Note the urgency, number of units and location. Draft a polite reply using only the price list below. If anything is missing, list the questions to ask the customer. Do not promise a time slot."

III. Getting started in four steps

1. Identify a repetitive workflow. Choose something that happens often, follows a pattern and is easy to check.

2. Map its inputs, decisions and outputs. On one page, write down what comes in, what someone decides and what goes out. If you can't describe the decision, AI can't help make it.

3. Choose a suitable tool or method. Start with what you already pay for, since many email, CRM and accounting tools now include AI features. No-code platforms such as Zapier, Make or Microsoft Power Automate can connect your apps. Consider a custom build only when off-the-shelf options fall short.

4. Pilot, test, measure and improve. Run the automation alongside your current process for two to four weeks. Track the time spent, any errors, and how often staff need to edit the drafts.

IV. Common mistakes to avoid

  • Automating a messy process. If the steps are unclear today, automation just makes the mess faster.
  • Sharing sensitive data without safeguards. The PDPA still applies. Use business accounts, limit who has access and keep personal details to a minimum.
  • Trusting every output. AI can sound sure of itself and still be wrong. Keep a person checking anything that goes to customers.
  • Automating too much, too soon. One working workflow beats five half-built ones.
  • Leaving employees out. The people who do the work know the exceptions.

V. Start with one workflow

For most SMEs, AI automation starts with one workflow that saves a little time every day.

If you'd like hands-on help designing and building your first automations, the AI for Business Automation programme covers workflow design, no-code tools and AI agents in practice.

Priya is an illustrative example based on a typical SME.