AI in a professional body: the four questions to ask before choosing the tool

AI in a professional body: the four questions to ask before choosing the tool

In short

  • For a professional body the right question is not “which AI should we buy”, but “which work repeats every week”.
  • All it takes is four questions on a sheet of paper: repetitive tasks, the time they take, where they get stuck, what we want to achieve in six months.
  • Before any tool: the DPO, a Council decision, an updated privacy notice and accounts in the organisation’s name, one per person.
  • The real change is moving from doing to checking: the assistant prepares, a person reads and approves.

⏱ Reading time: 6 minutes

A few days ago, during a call, someone who had read our article on prompts, connectors and AI agents asked me a simple, honest question: “We would like to understand whether and where to start. But without doing any damage.”

It is the right question. And for a professional body, the answer does not start with the tool. It starts with everyday work.

First of all: from doing to checking

Until recently, artificial intelligence was used like an oracle: I paste a text, ask for it to be made clearer, copy the answer. Useful, but it is the lowest step. Today assistants can carry out a task in several steps: read, search, compare, prepare a draft.

The easiest way to understand this is to think of a new colleague who has just joined. We explain four or five tasks, let them work and check everything before it leaves the office. With AI it is the same. Those who know their work well gain time. Those who do not risk approving mistakes faster.

The four questions to put on a sheet of paper

To anyone who asks us where to start, we always suggest the same exercise. One sheet, four columns, half an hour.

  1. Which tasks repeat every week? Only computer-based ones. For example: replies to members about fees and deadlines, checking registration applications, minutes, recurring communications.
  2. How much time do they take? An estimate is enough: “registration applications, about one hour a week”.
  3. Where do they get stuck? The real bottleneck is often not the task, but the waiting: a missing document, a reply that does not arrive, a step done by hand.
  4. What do we want to achieve in six months? Not “use AI”, but a concrete result: “draft replies to frequently asked questions ready every morning, to be reread and sent”.

With this sheet in front of you, it becomes immediately clear whether AI is really useful, and where. Sometimes the answer is no, and that is fine.

The rules to set before starting

A professional body processes personal data of members and citizens. That is why, before choosing any tool, there are some steps that must not be skipped.

  • Involve the DPO straight away. You need to understand which data the assistant will handle, where it will end up and whether it will leave the European Union. In many cases a data protection impact assessment (DPIA) will be needed.
  • A Council decision and a budget item. Using AI in the office is a choice of the organisation, not of the individual.
  • An updated privacy notice. Members and citizens must know whether, and for what purposes, the body uses these tools.
  • Organisation accounts, not personal ones. In free or personal versions, conversations may be used to improve the models unless set otherwise, and above all there is no contract between the organisation and the provider. Plans for organisations usually include rules on data processing: they must be read and checked.
  • One account per person. A shared account mixes conversations, confuses the instructions given to the assistant and makes it impossible to know who did what.
  • Start with the simplest plan. Plans are paid per user and consume “credits” based on use. Better to start at the basic level and move up only if really needed.

The European AI Act also asks those who use these systems to take care of staff training. The text was amended in 2026: what is actually required of a professional body should be checked with your own advisor.

Why the real risk is not reading

In the courses we run we often see the same scene. The assistant writes an accurate answer, complete with warnings. The person does not read it to the end and types “ok, go ahead”. The warning was there, but nobody read it.

Psychologists call it automation bias: the tendency to trust an automated system more than it deserves, especially when we are in a hurry (Parasuraman and Manzey collected the evidence in a 2010 review). The remedy is simple: the person remains responsible for what goes out, and reads it.

I was struck by the story of someone I spoke with: before trusting an assistant to identify plants, they asked it questions they already knew the answer to. Only when they saw it answered well did they move on to the real questions. Testing the tool on what we already know is an excellent first step, in the office too.

Nobody is late, but the pace is picking up

According to Istat, in 2025 16.4% of Italian companies with at least 10 employees use at least one AI technology. In Denmark, first in Europe, the figure is 42% (Eurostat). The gap does not depend on the tools, which are the same for everyone, but on skills and organisation.

It is a bit like the move from the typewriter to the computer: for a while both sat on the desks. Those who learned calmly, but in good time, did not fall behind.

How we work with professional bodies

We start from the four-question sheet, read it together and say frankly where AI can help and where it cannot. Then we suggest the right step, almost always drafts prepared by the assistant and approved by a person, with the DPO involved from the start. For staff we organise practical courses with a certificate of attendance.

Frequently asked questions

Where should a professional body that wants to use AI start?

From the tasks that repeat every week, the time they take and where they get stuck. The tool is chosen afterwards.

Can we use free versions for office work?

For texts that are already public, yes. For data on members and citizens you need an organisation account, with data processing terms checked by the DPO.

Is a formal decision needed?

Using AI is an organisational choice of the body: as a rule it goes through the Council, with a budget item and an updated privacy notice.

Is a shared account for the whole office better?

No. One account per person keeps conversations and responsibilities separate.

Can the assistant send emails on its own?

Technically it often can, but at the beginning it is not advisable: the assistant prepares, a person reads and sends.

Let’s talk

Would you like to do the four-question exercise together for your organisation? Book a 30-minute call or write to us at posta@swdweb.it.

Leonardo Pantaleo, CEO of SWD Group