How To Turn AI Into A Real Business Tool (Guide for SMEs)

Most people use AI like a search engine. Type a question, get an answer. That works fine for “what’s the capital of France.” For running a business, it falls flat.

If you own an SME, AI can do so much more than answer questions. It can become your marketing assistant. It can draft your proposals and quotations. It can help you build investor decks, or reply to customer complaints. The kind of work you used to pay an agency or hire a dedicated staff member to do.

But only if you stop treating it like Google.

In this post, I’ll walk through a four-step method that turns AI from a search tool into a real business tool. It’s a framework called PAIR, originally developed by Professor Oguz A. Acar at King’s College London for teaching students how to use generative AI properly. It works just as well for SME owners trying to get real value out of these tools.

A Scenario: The Renovation Tender

Imagine you run a small renovation company. Four staff. Eight years of HDB and landed property projects. You know your trade inside out.

One morning, an email lands in your inbox. A condo MCST wants you to bid for their lobby renovation. Budget: $200,000. They want a formal tender proposal in three days.

The problem is, you have never written a formal tender document before. Your past quotations were always WhatsApp messages with a price and a few photos. And you know that your other competitors are used to submitting tenders every week, with proper BD teams and polished templates.

You know you can do the renovation work. You just need help producing a proposal that matches what the MCST expects to see.

So you turn to AI.

The Naive Attempt (And Why It Fails)

Most beginners would jump straight to ChatGPT and type something like:

“I want a renovation proposal.”

ChatGPT replies in seconds with a tidy template. Project Overview. Scope of Work. Demolition Works. Carpentry Works. At first glance, it looks decent (see below).

But read the above again and you realise something. It says nothing about your company. Nothing about this MCST. Nothing about the actual project. It is a generic template that could be used by any contractor in any country.

Submit this to the MCST and your bid gets binned immediately.

This is the first lesson of using AI well: AI is like a smart consultant who just joined your company one minute ago. It knows a lot about renovation in general. But it knows nothing about you, your business, or this specific project.

So how can you expect a proposal that actually wins the bid? You can’t, unless you give the AI the context to do the job properly. That’s what the PAIR framework is for.

The PAIR Framework

PAIR stands for Problem, AI, Interaction, Reflection. It was designed for students, but the four steps apply to any business task you want to solve with AI.

Let me walk through each step using the renovation scenario.

Step 1: Problem

Before you open any AI tool, define the task first.

This is the part most people skip. They want to jump straight to typing into ChatGPT. Resist that urge. Spend a few minutes asking yourself four questions:

1. What kind of help do I actually need? Sales? Marketing? Finance? Operations? For the renovation tender, the answer is business development.
2. What’s the task, exactly? Create a tender proposal. But have you ever read one? Do you know what sections an MCST expects to see?
3. What are the requirements? Look at the tender brief carefully. Did they ask for a water feature? Sustainable materials? Specific certifications? Compliance with BCA guidelines?
4. What’s your edge? Maybe it’s your response time. Maybe price. Maybe you have a strong residential portfolio with happy clients you can reference.

This thinking takes a few minutes, but it prevents the AI from making huge mistakes later. Skip this step and the AI has to guess what you want. Arrive prepared, and the AI can start working properly from the very first prompt.

Step 2: AI

Pick the right tool for each part of the job.

Many people default to ChatGPT for everything. Different tasks are often solved better by different tools. Here’s a rough guide:

• Deep research (e.g. understanding what a tender proposal looks like in Singapore): Gemini in deep research mode, or Perplexity
• Drafting documents (the actual proposal): Claude, ChatGPT, or Gemini
• Formatting and design (turning your draft into a polished document): Canva or Google Docs
• Image creation (e.g. a mood board for the lobby design): Nano Banana, Midjourney

You shouldn’t expect one tool to do everything well. I’ve written before about matching tools to specific tasks when building content workflows, and the same principle applies to any AI work for your business.

Step 3: Interaction

Now the actual work happens.

For the tender proposal, you can use Google Gemini. It’s convenient, since if you have a Gmail account, you already have access at gemini.google.com. Log in, switch to “thinking” mode, and upload the tender specs as a PDF.

Then write a proper prompt. Not “I want a renovation proposal.” Use what you figured out in Step 1:

“I need help with business development. I run a four-person renovation company in Singapore with eight years of experience in residential projects. I need to write a tender proposal for a condo lobby renovation with a budget around $200,000. The MCST wants it in three days. The proposal needs to address the requirements in the tender specs, which are attached. Use a professional but approachable tone.”

Look at what’s in that prompt. Who you are. Your context. The task. The tone. Everything from Step 1 is now baked into the prompt. You’ve also uploaded the tender specs so the AI can read them in detail.

The results are shown below:

This time, the AI responds with something much better than the generic template. A proper introduction. A section called “Understanding of Project Requirements” with specific bullets pulled from your brief: the water feature, the durable materials, the minimised disruption to residents.

The first draft still won’t be perfect. Maybe it uses filler like “state-of-the-art finishes.” Your job is to push back and refine with your actual expertise:

“Replace the generic material references with specific brands I use. Nippon Odour-less Premium for paint. Hafele hardware for fittings.”

The AI updates. You read again. You refine again.

That back-and-forth is the actual skill. You’re not accepting the first draft. You’re shaping it into something that reflects your business.

Step 4: Reflection

This step separates a good AI user from a beginner.

After the AI produces something useful, you need to carefully review it before using it. Sounds obvious. But many people skip this because the output looks polished and confident.

Think of it this way: if a new staff member who joined your company last week handed you this tender, would you submit it without reading? Of course not. Same logic applies here.

Three checks to run on every AI output:

When you work with AI properly, it can give you a 75% accurate draft in ten minutes. The final 25% comes from you adding the expertise and judgement that AI doesn’t have.

Why This Framework Works For Any Task

We used a tender proposal as the example. The same four steps work for almost any business task you can throw at AI.

Writing a quotation for a customer? Define who the customer is (Step 1). Pick the right tool (Step 2). Write a prompt with the context (Step 3). Check the numbers and tone before sending (Step 4).

Drafting an investor pitch deck? Same four steps.

Replying to a tricky customer complaint? Same four steps.

Building a marketing campaign? Same four steps.

This isn’t theoretical. It’s the same process I use when I build SEO content workflows for clients or generate case studies for B2B companies. Define the problem properly, pick the right tools, prompt with full context, and review the output carefully before shipping anything.

When I used AI to write a WordPress plugin from scratch, I didn’t just type “write me a plugin.” I had to research the specifications first, choose the right tools, prompt with proper context, and then debug everything carefully. That’s PAIR in action, except I didn’t know the framework had a name back then.

What Actually Makes The Difference

Most AI training I see online focuses on “prompt engineering.” Memorise these magic phrases. Use this template. Add this trigger word.

That misses the point.

The skill that makes AI useful for SMEs is the thinking that happens before and after the prompt. Defining the problem clearly. Knowing what good output looks like. Catching the mistakes that AI makes. Adding the expertise that only you have.

The prompt itself is just the middle bit.

So the next time you sit down to use AI for something important, don’t open ChatGPT first. Open a notebook. Spend five minutes on Step 1. Come back to the AI tool when you actually know what you’re asking for.

You’ll get dramatically better results. And you’ll stop wasting time on generic templates that don’t fit your business.

---

If you want help building AI workflows into your SME, whether for proposals, content, customer service, or back-office automation, that’s exactly what I do at Halfborg. Reach out and let’s chat.