You're running a seminar for two hundred people. It's 8.45am, the welcome desk opens in fifteen minutes, and you want to know who hasn't checked in yet.
Until now, you had to open your registration software, export the list, then paste it into ChatGPT so it could help you sort it.
Now you can simply ask your assistant: "Who hasn't arrived yet?" It will go and find the answer itself, directly in your software.
This small change is happening across thousands of tools. And it rests on a very simple idea.
A universal plug
Remember when every phone had its own charger. A drawer full of cables, and never the right one.
Then a standard plug arrived, and you could connect anything with the same cable.
For AI, it's exactly the same story. Every assistant had to be connected to every piece of software in its own particular way. An enormous amount of work, which nobody was doing.
One standard won out to settle this. It's called MCP, and it plays the role of the universal plug: any software that adopts it can be connected in one go to ChatGPT, Claude, Gemini or Copilot. We explain it in detail here.
This week, two more
Two companies announced their connection this week, and they perfectly illustrate the two ways of doing it.
Eventtia, an event management software, has opened up full access. The assistant can read and modify: attendees, speakers, the welcome desk, and even payments. It can also create registration pages from your data. The company promises a connection in under five minutes, without a single line of code, included in all its plans.
Stravito, a tool that gathers a company's market research, has made the opposite choice: read-only. The assistant can search reports, presentations and videos, but it can't modify anything. And it cites its sources, so you know where each piece of information comes from.
An assistant that can only read your data can, at worst, give you a wrong answer. You read it again, you correct it. An assistant that can write — modify a list, create a page, touch payments — can do real damage, and sometimes without you seeing it straight away. That's not a reason to refuse. It's a reason to know exactly what you're authorising.
The detail that reassures
Stravito makes an important point: its connection inherits each user's permissions.
In practice: if you don't have the right to see a confidential report in the software, your assistant won't be able to see it either. It has no more power than you do.
That's exactly what you should demand from any connection. An assistant should never be able to do what you yourself don't have the right to do.
Before you connect anything
The question to ask is no longer "can I connect it?". Soon, the answer will be yes for almost everything.
The real question is: once it's connected, what will it be able to do? Read or modify. My data or the whole team's. With or without asking me.
Two minutes to check these three points is the best investment you can make before all your tools start talking to your assistant. Our access guide shows you how.
Because they're all going to do it. Some are even starting to remove their buttons: if nobody clicks any more, what's the point of keeping them?