Most small businesses do not need an AI strategy. They need one page that says what they will try, who is responsible, what it will cost and how they will know whether it helped. A strategy document gets written once and forgotten. A one-page plan gets pinned above the desk and checked every Friday.
This guide walks through that page section by section. You can fill it in with a pen in an evening. If a section feels hard to write, that is useful: it usually means the idea is not ready yet.
Line one: the problem, in your own words
Start with a problem you already feel, not a technology you have heard about. Good problems sound like complaints from the shop floor:
- “We miss WhatsApp orders after 8pm and customers go elsewhere.”
- “My sister spends every Monday morning typing up last week’s sales.”
- “Half our cash on delivery parcels come back because the customer never really meant to order.”
Write one problem. Not three. If you cannot choose, pick the one that costs you the most hours or the most money each month, and park the others at the bottom of the page under “later”.
Then write one sentence describing what “fixed” looks like. “Every WhatsApp order gets a reply within ten minutes, day or night” is a fixed state you can check. “Better customer service” is not.
Lines two and three: the owner and the boundaries
Name one person who owns this. In a family business that might be you, a son or daughter who is good with phones, or a trusted manager. The owner does not need to be technical. They need to have the authority to change how the work is done and the patience to check results every week.
Next, write down what the AI tool is not allowed to do. This is the most important line on the page, and most plans skip it. For example:
- It may reply to customers with prices and delivery times, but may not offer discounts.
- It may draft a refund message, but a person sends it.
- It may read the sales book, but may not change it.
A simple rule works well: anything that moves money, reaches a customer in a way you cannot take back, or cannot be undone, needs a person to approve it. You can relax the rule later once you trust the system. You cannot easily undo a week of wrong refunds.
Lines four and five: the budget and the time limit
Put a number on it, in rupees, for two things: what you will spend to set it up, and what you are willing to pay each month to keep it running. Monthly costs matter more than people expect. Many AI tools charge by use, so a busy month costs more than a quiet one. Write the ceiling you are comfortable with and ask any supplier how they will keep you under it.
Then set a time limit for the trial. Six to eight weeks is usually enough to see whether something works on a small job. Longer than that and the trial quietly becomes permanent without anyone deciding it should.
A worked example helps. Say the Monday sales summary takes your sister three hours each week. Over a month that is roughly twelve hours. If her time is worth, say, LKR 1,000 an hour to the business, the task costs about LKR 12,000 a month. A tool that costs LKR 4,000 a month and saves most of those hours is worth trying. A tool that costs LKR 15,000 is not, however clever it is. You can run your own figures through our automation payback calculator.
Line six: how you will know
Choose one or two measures you can count without special software. Before you start, count them for two weeks so you have a starting point. Useful measures for small businesses include:
- Hours spent on the task each week (a notebook is fine).
- Number of orders missed or replied to late.
- Number of mistakes a person had to correct.
- Number of complaints about the new process, from staff or customers.
Also write down what would make you stop. “If customers complain about wrong prices more than twice in a week, we switch it off and go back to the old way.” Having a stop rule written in advance makes it much easier to act on it, especially if you have already paid for the tool.
What a finished page looks like
A finished plan for a small clothing shop might read: Problem, late replies to WhatsApp orders at night. Fixed means, every message answered within ten minutes. Owner, Nimali. Boundaries, can quote prices and stock, cannot promise delivery dates outside Colombo, cannot discount. Budget, LKR 20,000 setup, LKR 6,000 a month ceiling. Trial, eight weeks. Measures, late replies per week, wrong answers per week. Stop if, more than two wrong prices in a week.
That fits on half a page. The other half is for notes each Friday.
What a one-page plan will not do
Be honest about the limits. A one-page plan will not tell you which tool to buy, and it will not fix a process that is already a mess. If nobody agrees on how orders should be handled today, an AI tool will simply make the disagreement faster. In that case the first job is to write down the current process, not to automate it.
It also will not suit every business. If your problem is a lack of customers rather than too much work, AI tools that save staff time will not help much. Spend the money on getting found instead.
If you want a second view on your page before spending anything, Quadra Lanka offers consulting for exactly this stage. Otherwise, take a sheet of paper tonight, write the one problem that annoys you most, and see whether you can fill in the rest by Friday.