Where to start with AI in your business: three tasks you can automate in a month

Most small businesses start with AI at the wrong end: by buying a tool. How to choose the first task, three that almost always deliver results within weeks, and how to tell whether it was worth it.

Almost every company that asks us about artificial intelligence starts the same way: "which tool should we sign up for?". It's the wrong question, or at least the second one. The first is different: where does your team's time go each week? AI isn't a project in itself; it's a way of taking hours away from specific tasks. If you don't know which ones, any tool you buy will end up as one more subscription nobody opens.

How to choose the first task

The first task you automate has to meet four conditions. It should be frequent, because a task you do twice a year doesn't justify the effort. It should be repetitive, always with the same structure even if the data changes. A mistake should be cheap, because at first the AI will get things wrong and someone has to be able to fix them before they reach a customer. And it should be measurable: how long it takes today and how long afterwards.

If you're not sure where your time is slipping away, start with the repetitive tasks test we published recently. In two minutes it tells you which areas of your business have the most room for improvement.

Three tasks that almost always deliver within a month

1. Answering the questions that keep coming back. Opening hours, rough prices, lead times, how to book, what paperwork is needed. In almost every business, half of the emails and messages ask the same thing in different words. AI can prepare a draft reply using your company's information, and a person only has to review it and send it. The point isn't for it to answer on its own from day one, but for nobody to have to write the same reply for the twentieth time.

2. Moving data from one place to another. Supplier invoices someone types into the accounting software, orders that arrive by email and get copied into a spreadsheet, website forms that are entered into the CRM by hand. These are dull, error-prone tasks and perfect for automation: the AI reads the document, extracts the data and puts it where it belongs for someone to check.

3. Summarising and turning things into tasks. A call with a client, a team meeting, a work report scribbled at the end of the day. AI can turn it into a clean summary with the pending tasks, who owns them and by when. It's one of the biggest time savers and one of the cheapest to set up.

What you shouldn't automate yet

Some things are better left for later. Anything that involves deciding about money or commitments to a customer without a person reviewing it: approving quotes, accepting returns or promising delivery dates. And anything that handles sensitive personal data before you're clear about where it's processed and with what safeguards. It's not that it can't be done; it's that it isn't the first step, and a bad first step takes away the appetite for a second.

How to tell whether it was worth it

Before you start, note how much time the task takes each week today and how many errors slip through. Then run the automation for four weeks and measure again. If you save less than it costs to maintain, stop it and try another task. If you save time, you now have both the argument and the method for the next one. That's how AI genuinely gets into a small business: one task at a time, with numbers and without big projects.

At Teix we design and implement AI systems for small businesses with exactly this approach: first we understand where the time goes, then we automate what pays off most and we measure it. If you'd like us to look together at where to start in your case, get in touch.

Sources: General Data Protection Regulation (GDPR, EUR-Lex) · European Data Protection Board: guidelines on automated decision-making and profiling · European Data Protection Board: data protection guide for small business