AI implementation · automation · Switzerland
AI in day-to-day operations – not as a project, but as an operating system
Most companies know AI will change how they work. But there is a gap between strategy and real use in day-to-day operations. We close that gap step by step, in a hands-on way, with measurable results.
Forget the grand design. Focus on the many small wins.
Automation sounds like an IT project, a requirements document, and six months of waiting. The reality is different: the biggest gains do not come from one large system – they come from a hundred small improvements in day-to-day work.
Participants in our training programmes save 11.5 hours a week on average – while also improving output quality, increasing satisfaction, and freeing up more time for strategic and creative work.
The calculation: A team of 10, 11.5 hours saved per week, a CHF 65 hourly rate: over a year, that produces savings in the six-figure range. Well-executed AI automation is not a bet – it is maths.
Some of our clients
What does AI implementation look like?
Why starting small and growing is the right approach
The most common mistake is waiting for the big system. The team never learns, the solution only covers part of the use cases, and by the time it is ready, the technology has already changed twice.
Standard software
With ChatGPT, Copilot, Gemini or, better still, Claude, and CHF 30 per person per month, many repetitive tasks can already be automated — without writing a single line of code. Initial smaller workflows can be built and linked to third-party tools. SEO analysis from Claude, optimise copy and push it directly into the CMS? No theory — fully possible today.
No-code & advanced integrations
Existing processes are extended: more data sources, more tools, deeper automation – with Claude Cowork, Claude Code or n8n. Still without a major IT project.
Custom & On-Premise
Only when the use cases are clear and the team knows what works does a custom solution make sense. Then it needs a solid foundation: validated processes, a team that understands and adopts the solution, and a scalable architecture.
This way, the team grows into it naturally. The solution is fine-tuned live. In the end, you do not end up with a system people still have to learn. You end up with one they helped build.
«He strikes a good balance between genuinely challenging the status quo and ensuring the project is delivered. Mike can turn strategy into execution. He can break complex technical situations into clear, manageable parts, which makes him a strong and trusted adviser to senior management.»
Thomas Truttmann·Vice President, Marketing & Communications·McDonald's Switzerland
«An absolutely inspiring day, Mike Schwede. Thank you for all your insights and suggestions. The journey has begun!»
Jürg Beutler·Head of Communications·Accident prevention advisory centre
«Thank you, Mike Schwede. It was very insightful and valuable. AI will make us more efficient and has the potential to help deliver much-needed cost savings in healthcare.»
Simon Stettler·Marketing Management·Visana
Frequently asked questions about AI implementation
Where should you start?
With the concrete problem, not with the tool. Which task takes up the most time every day? Which process is error-prone and repetitive? Start there. If you start with the tool, you look for problems to fit the solution, not the other way round.
How much does AI automation cost?
The range is wide: from CHF 30 per month for standard tools to six-figure investments for on-premise solutions. The right starting point for most SMEs: a Claude or ChatGPT Team licence, one or two workshops, initial workflows. Total cost in the first year below CHF 10'000. With careful implementation, the ROI can be proven.
How long until the first results are visible?
For simple automations with standard tools: days to weeks. For a pilot: 10 working days to the first working system. For complex custom solutions: 2–4 months.
Do you need technical expertise in the team?
For stage 1 (standard software), no. For stages 2 and 3, a technically minded team helps, but it does not need to be a development team. We guide and upskill.
What happens if an employee leaves the company?
This is the most common objection, and it is valid. That is why we do not build black boxes: documentation, knowledge transfer, and a team that understands the system are part of every project.
We have had bad experiences with AI tools. What are we doing wrong?
Usually one of three things: the wrong tool for the use case, poor data quality, or expectations that are too high. A frank assessment quickly shows where the problem lies.
What is the difference between a chatbot and a custom GPT?
Marketing term versus architecture. What matters is not the name, but whether the system is based on clean internal data, knows the right limits, and responds reliably. Cheap chatbots do not do that. Well-built systems do.