Italiano 
Why the right question is not «which AI to choose»
Many companies start from the tool: they try one assistant, then another, and after a few weeks enthusiasm fades because it is unclear what it is for. The problem is not the technology. It is the lack of a concrete starting point.
A good first AI project is not the most ambitious one: it is the one where you already know what «done well» means and can check it.
That is why it helps to reverse the order: first choose the task, then the tool. The rest of this article helps you make that choice.
Five questions to choose the first process
Take a task that weighs on you and test it against these questions. The more positive answers you get, the better the candidate.
- Does it repeat? If it is different every time, it is hard to standardise. Answering the same requests, preparing the same type of document or updating the same report are good fits.
- Does the information needed exist and can you reach it? Emails, CRM, price lists, documents. If data is scattered or incomplete, the first job is to put it in order, and it must be done before any AI.
- If it gets it wrong, what does the mistake cost? Start where human checking is easy, such as a draft to review, rather than a reply that goes straight to the customer.
- Can the result be checked? Decide beforehand how you will tell whether it works, for example how long the task takes today. Measure the starting point: without it, any improvement remains an impression.
- Is there someone who knows it inside out? You need someone who can say what is right and what is not. AI works better when there is a clear reference.
Examples of first processes for marketing and sales
Among the areas where companies most often start are sales and management, content and communication, and customer support. In everyday marketing and sales work the most common candidates are:
- draft replies to requests that always arrive the same;
- summaries of emails and notes from customer meetings;
- first drafts of quotes to review before sending;
- first drafts of content, such as newsletters, product sheets and website copy;
- periodic reports assembled from data that already exists.
Each should be tested with the five questions. If content is your starting point, you may also find Il tuo sito non convince? Come scrivere testi web che portano contatti (Italian) useful and, for the relationship between AI and search visibility, IA e indicizzazione: come ottimizzare il sito (Italian).
If you work in events, we covered concrete examples in IA negli eventi: come rivoluzionare l'organizzazione (Italian) and in Segreteria eventi 4.0: come usare l'IA (Italian).
What not to choose as a first project
- A process nobody has ever described. If you cannot explain it in a few lines, you cannot check the result.
- A task with sensitive data and no clear rules. First decide which data may be used; for privacy matters it is worth talking to an adviser.
- An important decision with no human check. At the start AI prepares, the person decides.
- Everything at once. A project that changes ten tasks at the same time does not show what worked.
From the chosen process to the first test
Once the task is chosen, define a small perimeter: one type of request, one type of document, one group of customers. Test on real cases, compare results with the starting point and only then decide whether to extend.
If AI must work with the tools you already use, integration comes into play: connecting AI to CRM, email and management systems, instead of using it in a separate chat window. The two paths are described on the pages about AI agents and assistants and AI integration and workflows (Italian pages).
Before you start
Check these six points
A process described in a few lines, with a clear start and end.
A responsible person, who can say what is correct.
The data needed identified, with who owns it and where it is.
A way to check results, with today's starting point measured.
A defined human check, before any use with customers.
A small perimeter and a date to evaluate the results.
Frequently asked questions
No, the opposite. A small perimeter with one precise task shows early on what works and what does not. For trying out texts and ideas, yes. Once company data or links to CRM and management systems come into play, it is worth designing carefully which data to use and who checks the results. It depends on the process and the data available. That is why it is best to start with a small perimeter and decide in advance when to evaluate the results. At least one person who knows the task and can say what is correct, plus whoever manages the data and tools involved.Do I need a big project to start?
Can I start with AI tools that are already available?
How long does a first project take?
Who should follow the project inside the company?
Have a repetitive task in mind?
Tell us about it: we will assess together whether it is a good first candidate and how to set up a test on your case.
Tell us about your task
