Almost everything written about AI for small businesses stops at the list: "a chatbot", "automate your replies". What nobody covers is what comes after — what you need to build it, how long it takes, what it costs to keep running, and when it is not worth doing at all.
This is that. Seven specific jobs, and for each one: what it solves, what you need in place first, how long it takes to get working, and when to leave it alone.
If what you are after is deciding where to start and how much to invest, that is in the AI for small businesses guide. This is the detail of each piece.
First: how to know what you will actually save
You need no tool to work this out, and it is the only number that matters.
Over one normal week, note every time you do one of these tasks and how long it takes. At the end you will have, per task: times per week × minutes each time. Multiply by four and you have hours per month.
That figure is your ceiling. If an automation costs more than those hours valued at your rate, do not build it, however good it sounds.
1. Answering the same questions, at any hour
What it solves. The enquiries that repeat: prices, opening hours, whether something is in stock, directions, where an order is. Above all the ones that arrive at night and at weekends, which today are lost entirely.
What you need first. Your answers written down somewhere. Nothing sophisticated: a document with the twenty questions you get asked most and their real answers, prices and conditions included. This part is 80% of the work and it decides whether the assistant is useful or makes things up.
How long. Two to four weeks, and most of that is gathering the document. If you already have it, days.
When it is not worth it. If you get fewer than five or six enquiries a day, maintenance will cost you more than the time you save. And if your answers change weekly, you will end up with an assistant giving stale information with your name on it.
2. Draft replies for email and WhatsApp
What it solves. The ten questions that arrive each week have the same ten answers, but each needs personalising. AI leaves the draft with the customer's details and your terms; you check it and send in a minute instead of twenty.
What you need first. Your current templates, even if they live in someone's head. You pull them out by reading the last fifty emails sent and seeing what repeats.
How long. A week. It is one of the fastest to set up because nothing needs connecting: it works on the email you already use.
When it is not worth it. If every reply you write is genuinely different — consulting, bespoke projects — the draft will get in your way more than it helps.
3. First drafts of content
What it solves. Product pages, posts, review replies, campaign emails. The blank page stops being the bottleneck.
What you need first. Someone who knows the product and checks before publishing. Without that, do not do it: one wrong detail on your own product page does more damage than having no page.
How long. Days. It is the most immediate of all.
When it is not worth it. When nobody will review. Search engines have spent years getting better at spotting mass-produced content with no judgement behind it, and the cost of being caught is greater than the time saved.
4. Capturing and following up leads
What it solves. Every enquiry logged with what they asked, when and where it came from, and an automatic follow-up two days later if nobody replied.
What you need first. One place where all enquiries land. If some arrive by WhatsApp, some by the form and some by phone and go unrecorded, the first job is not automating: it is bringing them together.
How long. One or two weeks if you already have somewhere to store them.
When it is not worth it. Frankly, never. Most small businesses do not lose customers on price: they lose them by never writing back. This is almost always the first thing we recommend.
5. Turning scattered notes into something searchable
What it solves. Notes, meeting minutes and loose documents become procedures you can search and read. Less "where was that?".
What you need first. The documents existing in digital form. If they are on paper or in two people's heads, that is the prior task.
How long. Depends on volume, but usually one or two weeks.
When it is not worth it. In a team of two or three who already understand each other, the benefit is small. This pays off when new people join and need bringing up to speed.
6. Forecasting with your own data
What it solves. Which product will move most next month, what you are going to run out of, when the workload peaks. Planning with data instead of instinct.
What you need first. History. At least a year of sales or orders, and clean. Six months of patchy data produces nothing reliable.
How long. Weeks, and it is the only one on the list that can fail even when done properly: if there is no pattern in your data, there is no forecast.
When it is not worth it. If your business is strongly seasonal and you already know your peaks by heart, this tells you nothing new.
7. Chaining a whole process
What it solves. An order comes in once and appears by itself in the store, the tracking sheet, the invoice and the warehouse notice. AI barely features here: this is plain automation, and it is where the most hours come back.
What you need first. The process working by hand first. This matters: automating a broken process only makes it fail faster and in more places at once.
How long. Two to six weeks depending on how many systems need connecting.
When it is not worth it. While the process is still changing. Wait for it to settle or you will be rebuilding the automation every month.
What it costs to keep running, which nobody mentions
All seven have a monthly cost, and it does not go away:
- AI service and connection tools. For a small business, 20 to 60 USD a month depending on volume. Check it on the day you decide: rates change.
- Keeping it current. You change a price, add a service, close for holidays. If nobody flags it, within three months your assistant is giving false information with your name in front of it. Nobody needs to write code, but someone does need to be responsible. It is a few minutes a month, not a job.
Where to actually start
Pick one: the one you repeat most often each day according to what you noted in that first week. Build it, use it for a month, and measure whether it gave back the hours you expected.
If it did, move to the next on the list. If it did not, you switch it off having spent a month instead of a quarter. Both outcomes are fine; the bad one is building five things at once and not knowing which failed.
At FenixL we build this kind of automation from 250 USD for the simple ones, orders or leads, and 500 USD for an assistant working from your real business information. Before proposing anything we look at your process: if the answer is that it does not pay off for you, we say so.
Want to know which of the seven suits you? Tell us which task eats most of your day and we will come back in writing, free and with no commitment.