Which AI tool should I use?

To start, a general chat assistant is enough. Knowing how to ask and review matters more than the tool.

Will AI replace marketing done by people?

It will replace repetitive tasks. Deciding who to talk to, what to say and why still needs someone who knows the business.

Does Google penalise AI content?

Google has said it assesses how useful content is, not how it was produced. Generic content produced in bulk and without review tends not to perform well.

Can I use AI to reply to client messages?

To prepare replies, yes. To reply on its own, only for simple, repeated questions and with clear rules. Clients quickly notice when they are talking to a machine that doesn't know what it is saying.

← All insights

AI·4 min read

Updated 1 November 2026 · Andreia Cirne

AI in small business marketing: what to use and what to avoid

AI won't decide what to sell or who to sell it to. But it can save you hours each week, if you know what to ask it and what never to hand over.

Everyone talks about artificial intelligence and almost nobody has time to find out what it actually does for a small business. Between the enthusiasm of people who think it will replace everything and the fear of people who won't touch it, there is a practical, useful and much less dramatic way to use it.

What AI does well

Used as an assistant, AI saves time on these tasks:

First drafts: website copy, posts, emails and answers to frequent questions. You review and adjust.
Variations: the same message in shorter versions, for different platforms or in another language.
Organisation: turning loose notes into a task list, a plan or the structure of a page.
Content ideas: the questions your clients usually ask, which can become posts.
Summaries: meetings, long documents, client reviews.

What AI does badly

Some things should never be handed to a tool:

Deciding your offer, your audience or your price. It doesn't know your business or your clients.
Writing in your voice without guidance. Without your own examples, the text comes out generic and sounds like everyone else.
Guaranteeing facts. It makes things up with complete confidence. Numbers, laws and data always need checking.

Examples by type of business

In practice, the most useful use changes from business to business:

Holiday rental: answers to guests' frequent questions, in several languages, reviewed before you save them.
Clinic or therapist: copy explaining each treatment in plain language, checked by the person who provides it.
Real estate agency: a first draft of a property description from your notes, without inventing what the house doesn't have.
School or trainer: exercises, review questions and lesson summaries.
Independent professional: proposals based on a template of your own and follow-up emails.

How to ask well

A good request has context, a goal and examples. Instead of asking for a post about pilates, try something like this: I run a pilates studio in Faro for people with back pain; write three post ideas for people who have given up on the gym, in a warm and practical tone; here are two texts of mine as examples. The difference in the result is huge.

How to keep your own voice

Gather three to five pieces of your own writing you like, such as a post, an email to a client or a service description, and always use them as examples. Say what to avoid too: words you would never use, stock phrases, exaggeration. Then cut. An AI's first draft almost always has too many adjectives and too little substance.

Things to be careful with

Four simple rules avoid most problems:

Don't put clients' personal data into tools you don't control.
Review everything before publishing, especially in regulated areas such as health.
Don't let AI reply to clients on its own without clear rules and review.
Save the requests that work well, so you can reuse them and share them with your team.

How to tell whether it is worth it

Time one specific task before and after. If writing the weekly post took an hour and now takes twenty minutes including review, you have gained. If you spend the same time fixing what the AI wrote, you have gained nothing: change the request or go back to doing it by hand.

AI is also changing how people find you

More and more people ask AI assistants instead of searching on Google. Those assistants summarise what they find on websites that answer specific questions clearly. A site with clear service pages, frequent questions and useful articles has a better chance of appearing in those answers. One more reason to write about what your clients actually ask.

Where to start

Pick one repetitive task that steals time every week, such as answering the same questions by message or writing the weekly post, and use AI only for that for a month. Then decide whether it is worth extending to other tasks.

If your team needs to learn to use these tools with judgement, hands-on training with examples from your own day-to-day work saves months of trial and error.

Common questions

Which AI tool should I use?: To start, a general chat assistant is enough. Knowing how to ask and review matters more than the tool.
Will AI replace marketing done by people?: It will replace repetitive tasks. Deciding who to talk to, what to say and why still needs someone who knows the business.
Does Google penalise AI content?: Google has said it assesses how useful content is, not how it was produced. Generic content produced in bulk and without review tends not to perform well.
Can I use AI to reply to client messages?: To prepare replies, yes. To reply on its own, only for simple, repeated questions and with clear rules. Clients quickly notice when they are talking to a machine that doesn't know what it is saying.

Who writes this

Andreia Cirne

Marketing consultant and certified trainer (CCP) based in Faro. She spent almost six years at a digital agency managing marketing projects, websites and teams, and providing marketing consulting to each client. Today she works directly with service businesses and trains teams so they can do the work without depending on anyone.

Read about my path

If your team needs to use AI with judgement, hands-on training saves months of trial and error.

See the training