Already paid for, but unused
Microsoft 365 or Google often include AI already. Without suitable templates, it stays at the odd experiment.
AI consulting and training · Zug
AI your team actually uses every day.
I introduce AI to teams that write a lot and handle confidential data: SMEs, fiduciary firms, medical practices, associations, foundations and municipalities. Chosen independently, with tested templates, clear rules and training. After 30 days, we measure how much time you save.
So far: AI internship with the Canton of Zug, pilot project with Alzheimer Zug, independent research in machine learning. Fixed price, no hourly billing.
The situation
The tools are there. Hardly anyone uses them.
In many organisations, quotes, emails, minutes and invitations eat up hours every week. AI could shorten that. Three questions hold most teams back.
Unclear what is allowed
Can minutes with names go into AI? Without a clear rule, some avoid it entirely and others enter everything.
No time to get started
Daily work goes on. Nobody has hours to compare tools and rethink processes.
The AI starter package
Set up in two weeks, part of daily work after 30 days.
A fixed scope and a fixed price. At the end, your team works with templates that fit its tasks, and you know how much time it saves.
Swipe sideways for all six services →
01
On-site assessment
Two hours at your office: which tasks repeat, how much time they take, which licences you already have.
Microsoft 365in place
Gemininot needed
ChatGPTnot needed
02
Independent tool advice
I don't sell licences. First I check what you already pay for.
QuoteReply to enquiryMinutesSummaryAnalysisInvitation+ 9
03
Template library
10–15 tested templates for your most frequent tasks: quotes and replies, minutes, document summaries and list analyses.
Quote without client namesYes
Payroll listNo
Minutes, anonymisedYes
04
Guideline with clear rules
One page that sets out what may go into AI and what may not, aligned with Swiss data protection law.
0:00What AI can and cannot do
0:30Templates on your own examples
1:30Guideline and questions
05
2-hour training
Your team uses the templates on its own examples, not on slides.
TrainingQuestions and adjustmentsFollow-up measurement
06
30 days of support
Questions by email, template adjustments and, at the end, a measurement of the time saved.
A closer look
What this looks like for you.
Can this go into AI?
The guideline answers this for your organisation, one line per case. Your team doesn't have to guess, and nobody enters confidential data by mistake.
- Quote without client namesYes
- Payroll listNo
- Invitation to membersYes
- Donor list with namesNo
- Minutes, anonymisedYes
Example from a guideline
What a template looks like.
Each template is a ready-made instruction for the AI. Your team only fills in the marked fields and gets a draft in your organisation's tone.
The templates stay with you, also after the project.
Write an invitation to the [event] on [date] at [time] in [place].
- Tone: warm and polite, 120 words at most.
- List the agenda items: [keywords].
- Ask people to register by [deadline] with [contact].
- Suggest two subject lines.
Do not enter members' names.
Tools
I work with what you already have.
I start with the tools you already pay for. If something new is needed, I set up a team version with a data protection agreement. I take no commissions.
Often already in your subscription, for writing and research, or for design and workflows:
- Microsoft 365 Copilot
- Google Gemini
- Google Workspace
- Zoom
- ChatGPT
- Claude
- DeepL
- Perplexity
- Canva
- Notion
- Zapier
- Make
- Infomaniak
- Proton
Brand names belong to their owners. No partnerships exist.
Who it's for
For teams that write a lot and handle confidential data.
The industry matters less. Typical uses look like this:
- Invitations and newsletters to members
- Minutes of board meetings
- Texts for flyers, website and annual report
Member and donor lists stay out.
- Answers to frequent questions from residents
- Minutes and action lists
- Announcements and texts for the municipal newsletter
Personal data from case files stays out.
- Standard correspondence with clients
- Checklists and internal processes
- Summaries of guidance notes and legal changes
Client data stays out; professional secrecy applies.
- General patient information and leaflets
- Texts for the website and notices
- Internal processes and onboarding material
Health data never goes into AI.
- Quotes, replies and standard letters
- Minutes and summaries
- Job ads and website texts
Customer and staff data only under a clear rule.
Possible later: Automating workflows, for example with Power Automate. From CHF 3'000, after a starter package, from summer 2027.
The process
Three steps: two weeks of setup, 30 days of support.
Week 1
Understand
A free 30-minute first call, then the on-site assessment. I measure how much time the tasks take today.
Weeks 1–2
Set up
Choose the tool, build templates and test them on real examples, write the guideline. You give feedback, I adjust.
Week 2 + 30 days
Embed
Training with your team, then 30 days of support. At the end we measure again and compare.
References
Experience from Zug.
My first assignment: finding, assessing and testing AI use cases in a cantonal office. My first pilot project: a non-profit organisation in Zug.
Civil Engineering Office, Canton of Zug
- Identified, assessed and tested AI use cases
- 8 working prototypes
- Guideline for drone imagery
- Presented the results to the whole office
Alzheimer Zug
- AI templates for invitations, announcements and flyers
- A rule on which information must not go into AI
- Team training and measurement of time saved
Case study with measured results
The pilot project's evaluation will appear here in December 2026.
The concepts, background papers and recommendations he produced form a valuable basis for future development […] in the areas of drone use and artificial intelligence.
Nils Leutenegger worked through the tasks assigned to him very thoroughly and impressed with his quick understanding, his independence and his open, curious manner.
About
Nils Leutenegger
I do independent research in machine learning and write my own research papers; in September 2026 I presented my work with a poster at the Bernstein Conference in Frankfurt. So I know these tools from the inside and know where their limits are.
At the Canton of Zug's civil engineering office I saw what matters: the problem is rarely the technology, it is the rollout. The tools have to fit the tasks, and the team has to back them. That is where I come in.
From autumn 2027 I will study at ETH Zurich. Until then I deliberately take on only a few projects, so each one gets the care it needs.
Notes (German)
Notes from practice (in German).
- Can this go into AI? Three questions before your team enters anything.Personal data, sensitive data and the right account: a rule of thumb for daily work.Read (German)
- Copilot, Gemini or ChatGPT: which tool fits your team?Why the answer almost always starts with your existing subscription.Read (German)
- The problem is rarely the technology. What I learned in a cantonal office.Four lessons from eight prototypes at the Canton of Zug's civil engineering office.Read (German)
FAQ
Frequently asked.
Do we need new software or licences?
We clarify that in the initial review. Depending on your tasks, what you already have is enough, such as Microsoft 365 or Google Workspace, or a licence pays off. I don't sell licences and don't recommend anything in advance.
What happens to our data?
For the templates I work with examples that contain no sensitive personal data. The guideline sets out what your team may enter. Everything I see at your organisation stays confidential.
How much time does our team need?
About five hours over two weeks: two for the assessment, two for the training and about one for feedback on the templates.
Which organisations is this for?
For teams of about 3 to 50 people with a lot of recurring writing, for example associations, foundations, municipal administrations, fiduciary firms, practices and SMEs in the Zug area. What matters is not the industry, but that there is a lot of writing and confidential data is involved.
Who is responsible if the AI makes a mistake?
The AI writes drafts; it decides nothing. Every template and the guideline state that a person checks the text before it leaves your organisation. So your team stays in control, just as with a draft from the office. In the training we practise spotting typical AI mistakes.
What if it's not a fit for us?
That's what the first call is for. It's free and takes 30 minutes. If I don't see a clear benefit, I'll tell you.
Contact
Is AI worth it for your team?
Let's find out in a free 30-minute call, in person in Zug or by video. I work in German and English.
- nils.leutenegger@leutenegger-solutions.chReply within one working day
- 077 202 01 02Weekdays, otherwise I call back
- Hertistrasse 9, 6300 ZugMeetings at your office