MILabs
Blog

Personalised offers without a data team: AI for small-medium enterprises

2026-11-18

Personalised offers without a data team are possible, and for most small businesses they are simpler than the software vendors make out. You do not need a platform. You need one export of past sales, three questions, and an AI tool to help you read the answers.

This is an opinion piece, so here is my position up front: most SMEs should ignore "AI personalisation engines" and do a smaller, plainer version by hand. It works better at your size, and you can run it this week.

Personalisation is mostly a list problem, not a technology problem

When big retailers talk about personalisation, they mean systems that rewrite a website for each visitor. That is not what a small-medium enterprise needs.

A café owner who texts regulars about a new Saturday special is personalising. So is a plumber who emails last year's hot water customers a reminder about servicing. The skill is knowing who should get which message. That is a sorting job, and sorting is something AI does well.

The Shopify guide to data-driven personalisation starts from the customer data a business already collects, and I think that is the right place to start. Most small businesses are sitting on orders, invoices, quotes and enquiry emails that nobody has looked at properly.

Personalised offers without a data team: the three-question export

Here is the method. It takes an afternoon. It is the thing I would want you to walk away with.

  1. Export your last 12 months of sales or jobs. From Shopify, Square, Xero, MYOB, your booking system or your invoicing tool. You want date, what was bought, how much, and a customer ID.
  2. Remove names, emails and phone numbers. Replace them with a number. You do not need identities to find patterns, and you should not paste personal information into a tool you have not checked.
  3. Give the file to an AI tool and ask three questions.
    • Who bought more than once, and what did they buy first?
    • Who bought once, more than six months ago, and never came back?
    • Which products or services do people buy together?

That gives you a rough picture: people who return, people who drifted off, and people with an obvious next purchase. Each deserves a different offer. A repeat customer does not need a discount to buy again. A lapsed one might.

Always check the AI's counts against a quick filter in your spreadsheet. These tools can miscount, and a wrong list means an embarrassing email.

Match the offer to the customer, and protect your margin

This is where I part ways with the usual advice. Personalisation is often sold as a way to discount more cleverly. I think that is the wrong lens for small businesses.

An offer is anything that makes the next purchase easier. For a repeat customer it might be early access, not a lower price. For a lapsed customer it could be a small incentive with a deadline. For someone who bought a drill, it might be the matching bits at normal price with free delivery.

A simple rule: discount only the people who need persuading. Everyone else gets access, convenience or a thank you. That keeps margin intact in a week when many businesses give it away. If you have not yet worked out who to aim at this season, our post on reaching the right customers covers that step.

Once you have two or three offers, ask AI to draft each in your voice, using two or three of your own past emails as examples. Then edit them. The drafts are a starting point. If you want the full campaign planned across channels, the afternoon campaign guide walks through it.

What this looks like for tradies, contractors and sole traders

Personalisation is not only for shops. Here is an example of a typical business, not a client.

Imagine a sole trader electrician with 300 past jobs in an invoicing tool. She exports them, removes the contact details and asks an AI tool which customers had a switchboard upgrade, which had air conditioning installed, and which have not booked in over two years.

Three groups appear. The switchboard customers get a note about a smoke alarm check. The air conditioning customers get a pre-summer service reminder. The two-year gap group gets a plain "we haven't seen you in a while" message with a free safety check. Three short messages, each relevant, none needing a data analyst.

The same logic works for a landscaper, a bookkeeper or a physio. Tradies and contractors in particular tend to have excellent history in their invoices and almost no habit of using it.

Where this goes wrong

A few honest warnings.

  • Small samples mislead. If you have 40 customers, treat patterns as hunches, not facts.
  • Creepy is worse than generic. Referring to something a customer browsed, without them knowing you track it, loses trust. Stick to what they bought from you.
  • Privacy and spam rules apply. The Spam Act covers any business that sends marketing emails or texts, whatever its size. Only message people who have agreed to hear from you, identify your business, and always include an easy unsubscribe. The Privacy Act exempts many small businesses under a turnover threshold, but not all of them, so check with the regulators (ACMA for spam, the OAIC for privacy) before you assume you are covered either way.
  • More offers is not better. Two or three that you checked beat ten that you did not.

If you want the wider case for why customers respond to relevance, the involve.me roundup of personalisation statistics collects a lot of figures. They come from a vendor that sells personalisation, so I would not build a plan on any single number. Your own sales export will tell you more about your customers than a headline percentage.

What to do differently this week

Stop thinking of personalisation as software you have to buy. Treat it as a few lists and a few messages. Run the export, ask the three questions, and send one tailored offer to one group before the sales rush. If it works, add a second group. If it does not, you have lost an hour. For the wider plan, see the BFCM playbook for small businesses.

If you would like help turning your own customer data into usable offers, that is what MILabs' AI adoption work is for. Discovery is free, no obligation from the first chat. Bring one export of your past sales or jobs, even a messy one, with names and contact details stripped out, and we will look at what it can tell you. Every enquiry gets a reply within 24 hours.

Sources

good questions, straight answers

Frequently asked questions

Do small businesses need a data team to personalise offers?
No. A spreadsheet export of past orders or jobs, plus an AI tool like ChatGPT or Claude, is enough to find patterns and draft offers that fit each group of customers.
Is it safe to upload customer data to an AI tool?
Only if you strip out names, emails and phone numbers first, and check the privacy settings of the tool you use. You rarely need identifying details to find patterns.
How many different offers should a small business send?
Two or three. More than that and you will not have time to check the copy or learn which one worked.
let's make something happen

AI adoption: how we can help

Choose how you want to work on this with MILabs. Both options start with an enquiry.

Discovery is free, no obligation from the first chat.

Let us build it for you

We build it and look after it.

Build it yourself, we'll help

You build it with our direction, teaching and a weekly check-in.