This week at the KI Café, a statement was made that has stuck with many.
Manuel Pfiffner from sl.one showed a statistic: 70% of employees use AI today.
"A customer told me: nobody uses it for me"said Manuel.
His answer to the customer was direct: "Don't believe that. Please don't believe that."
And then the crucial point:
What happens if an employee enters company data with their private ChatGPT account?
🎬 Video: Manuel Pfiffner dispels cloud myths
The problem: three myths that slow down SMEs
I have been observing the same pattern for months.
Managing directors want to use AI. They have heard of Copilot, of ChatGPT, of all the possibilities.
But then come the concerns:
"Cloud is insecure."
"Everyone has access."
"That's far too expensive."
Manuel dispelled these three myths at the AI Café.
Myth 1: Cloud is insecure.
The reality: Microsoft employs over 3,000 people just for security. Security updates are rolled out immediately.
Manuel told us about a customer who sent him a vulnerability report.
His answer: "You're in the cloud, you've had the update for a long time."
Myth 2: Authorities have access.
Microsoft has contractually confirmed: No third party access without informing the customer first.
Myth 3: Cloud is more expensive than on-premise.
This is where it gets interesting.
Manuel showed the real comparison: With on-premise, most people forget the hidden costs.
Electricity. Air conditioning. Room rental. Maintenance. Hardware replacement.
An example from his practice:
A customer had made his last server update 273 days ago.
"That's quite sporting," said Manuel dryly.
He found "Call of Duty" installed on another customer's company notebook because his daughter supposedly needed it for school.
This is the reality of uncontrolled devices.
🎙️ You prefer to listen to the newsletter? Here is a short, AI-generated audio summary of this issue:
AI data security; private accounts; cloud myths
The framework: 3 steps to an AI-ready infrastructure
Zeno John then showed the second part: what Microsoft Copilot can actually do - and where the limits lie.
Step 1: Understand the difference between private and business co-pilot.
This confuses many people.
There is the Copilot, which sits in the Windows icon. And there is the M365 Business Copilot.
The Business Copilot costs extra (currently around CHF 30/month), but offers crucial data protection.
Why the price difference?
The complexity of security in companies is massively higher than in the private sector.
Step 2: Pay attention to the authorizations.
We have carried out an interesting test.
I asked Copilot for the payslip of a fictitious employee - once by e-mail, once on the SharePoint.
The answer: No information. Even with all the authorizations as managing director.
This shows that The safety rules in the company also apply to AI.
If the authorizations are set correctly, Copilot respects them.
Step 3: Combine the best of both worlds.
sl.one recommends a hybrid approach:
Standard services such as e-mail and Office via Microsoft 365.
Sensitive specialist applications in the private cloud.
The advantage: a direct connection from the office to the data center. No firewall required. The same speed as a local server.
The practice check: What came up in the discussion
The most exciting discussion came at the end.
One participant asked about the dependence on US tech.
Manuel confirmed: "It is technically possible to work without American products.
"But the problem is, you couldn't work together anymore."
If your compliance system blocks third-party software and you want to send someone a Teams invitation - then you have a problem.
Another point that came up: the energy consumption of AI.
Zeno called it "a huge thing that has not yet been published like this."
The server farms not only need electricity, but also water for cooling.
In the USA, nuclear power plants are being built to operate the data centers.
These are the realities that often get lost in the AI euphoria.
Take part in the MMIND.ai voting
The co-pilot with the most votes will be built in January - and exclusively presented here in the Flow Factor.
