The next era of AI is not another app in a browser tab. It is company AI that runs in-house: an AI that understands your whole business, meaning your knowledge and your systems. And it runs in your own environment: in your own rack or in a private cloud of your choice.
An ordinary Tuesday. A project manager wants to know what the expenses policy says about overnight stays. She searches the intranet, opens three PDFs and in the end emails HR anyway. A customer calls to ask about the status of order 4711. To answer, the colleague on the phone would first have to open the ERP and look up the order, so he asks the workshop instead. In the afternoon, someone needs to find a meeting slot with the project team. And somewhere, someone pastes the text of a new legal amendment into a public AI tool to work out what it means for the company.
The knowledge is there. It is just scattered across systems that do not talk to each other. This is exactly the situation HABB Switzerland from Rüegsauschachen offers HABB Core for. For a broader view of how AI automation is evolving at Swiss SMEs, see our article on AI automation for Swiss SMEs. This article is about the next step.
Why is the first wave of AI not enough?
The first wave falls short because public AI tools do not know your company, and your content has to leave the building to use them. Many employees use such tools today, often on their own initiative and with good intentions. Above all, that shows one thing: the need is real. But the first wave has three limits:
- The tool does not know your company. A public AI tool knows nothing about your expenses policy, your minutes or your ERP. It only works with whatever someone pastes in.
- Data goes outside. Whatever is pasted in leaves the building. Often nobody can see who sent what, or where.
- Costs grow with use. Cloud AI is often billed by usage, for example per token. The more your team uses it, the harder the costs are to plan.
The result: scattered tools, an uneasy feeling about data protection and no link to the systems where the real work happens. The next step turns this around.
What is company AI that runs in-house?
In-house company AI is a company-owned AI that knows your knowledge (documents, policies, minutes, file shares, email) and your systems (ERP, CRM, business applications). It is often also called on-premise AI or local AI. The language model runs locally, on your own hardware or in a private cloud of your choice. Documents and the knowledge base stay in your own environment.
«Local» means different things to different people today. So here is a clear distinction:
| Option | Where does the language model run? | Your documents | Costs |
|---|---|---|---|
| HABB Core in your own rack | On your own hardware, in a rack or server cabinet on your premises | Stay in your own environment | Fixed and predictable; tokens only when a question is deliberately delegated to a cloud AI |
| HABB Core in your private cloud | In a private cloud of your choice | Stay in your own environment | Fixed and predictable; tokens only when a question is deliberately delegated to a cloud AI |
| Public cloud AI | At the provider | Whatever is entered goes outside | Often by usage, for example per token |
Which of the two HABB Core options suits you is something we work out together, with a cost comparison over three years. Cloud AI does not disappear. It becomes a tool you use deliberately and under control.
One point of contact for everyone: how HABB Core works
HABB Core is the company AI in your own rack. The idea is simple: your employees have a single point of contact. They ask where they already work: in Teams, in the browser or on their phone. HABB Core knows who handles what best and passes the request on by itself.
Back to our Tuesday. Here are the same four questions with HABB Core:
- «What does our expenses policy say about overnight stays?» The answer comes from your company knowledge, with the source cited. The project manager can see which document it comes from.
- «What is the status of order 4711?» HABB Core reads the status directly from the connected business system, for example the ERP. Read-only, and within the permissions of the person asking. Nothing is changed.
- «Find a meeting slot with the project team next week.» What an existing assistant such as Microsoft Copilot already does well stays there. HABB Core hands the task over, and the handover is logged.
- «What does the new legal amendment mean for us?» This needs knowledge from outside. HABB Core brings in a cloud AI, but through the gateway. Only the approved question goes out. Your documents stay with you. Only in this case are tokens used.
Four questions, four routes, one point of contact. That is the real difference from the first wave: the AI is no longer just another browser window. It connects the systems that do not talk to each other today.
And it does more than answer questions. Later, HABB Core can also carry out actions in your systems, always approved by a person and logged. Verified solutions are stored internally as «skills».
Do I have to replace my systems to run my own AI?
No. HABB Core is built into the IT you already run. Your systems, tools and IT partners stay as they are. ERP, CRM and business applications are connected through open interfaces (APIs). On the knowledge side, that means documents, policies, minutes, file shares and email.
So you replace nothing. You make what you already have accessible to your team. If you would like to check what that means for your systems in practice: Book an AI and process check.
How do you stay in control of company AI?
Documents and the knowledge base stay in your environment, and your existing permissions still apply. An AI that knows your whole company has to follow rules. In HABB Core, they are built in:
- The gateway. It controls what leaves the building, with a data filter, a log and a budget. At most an explicitly approved request goes out. Every delegation can be traced.
- Permissions stay as they are. The AI answers with the rights of the person asking. Each person only sees what they are already allowed to see today. The AI does not open doors that were closed before.
- Read-only first. At first the AI only reads. Writing to systems comes later, and only with approval.
- No lock-in. HABB Core builds on open building blocks. Configuration, learned procedures (skills) and knowledge belong to you and can be exported at any time.
We are also clear about where the boundary lies: when a question needs knowledge from outside, a cloud AI is used. Not secretly, but deliberately, filtered and traceable. An AI that is allowed to do everything is not progress for an SME, it is a risk. An AI that is allowed to do exactly what you approve is progress.
If you are interested in the wider question of who owns data and infrastructure, read more about data sovereignty in our article on the topic.
How predictable are the costs of local AI?
The costs are fixed and predictable because the language model runs locally. The local language model has no token costs, however intensively your team uses the AI. Tokens are only charged when a question is deliberately delegated to a cloud AI. That request then goes through the gateway, with a budget and a log.
That changes the logic: more use in-house does not mean higher costs. Requests to a cloud AI are covered by the budget in the gateway. Whether your own hardware or a private cloud fits better is shown by the three-year cost comparison in the AI and process check.
And what does it deliver? There is no one-size-fits-all answer. But here is a sense of scale:
Illustrative example: Suppose 60 employees each save 20 minutes a week. That is around 20 hours a week. At internal costs of CHF 80 per hour, that comes to around CHF 6'900 a month.
Example figures, not a guaranteed saving.
Which use cases are really worthwhile for you only becomes clear once we look at your processes.
How do you introduce AI in your own company?
In stages rather than as a big-bang project, and starting with the process rather than the technology. Our principle: process first, then technology. In workshops, we ask the same four questions about every process:
- What comes in (input)?
- What should come out at the end (output)?
- Which steps lie in between?
- Which step stays with a person, and which one does the machine take over?
After that, HABB Core is introduced in four stages:
- Knowledge (6–10 weeks): The AI answers questions from your company knowledge, with sources cited.
- Read systems (+4–8 weeks): ERP, CRM and business applications are connected through open interfaces, read-only at first.
- Act (+6–12 weeks): Actions in your systems, always approved by a person and logged.
- Learn (ongoing): Verified solutions are stored internally as skills.
Each stage has its own result. After each stage, you decide whether to continue. No leap into the unknown, but a path you can see from start to finish. That is how we understand «Swiss Quality · Precision · Reliability»: clean, step-by-step work rather than a noisy big project.
Frequently asked questions
Can AI run inside my company instead of at a cloud AI provider?
Yes. With HABB Core, the language model runs locally: on your own hardware in a rack or server cabinet on your premises, or in a private cloud of your choice. Documents and the knowledge base stay in your own environment.
Does my data leave the building?
No. Documents and the knowledge base stay in your own environment, in your own rack or your private cloud. At most an explicitly approved request goes out to a cloud AI, through the gateway with a data filter, a log and a budget.
Will everyone be able to see all the data?
No. Your existing permissions stay in place. Each person only sees what they are already allowed to see today.
Do my employees need a new tool?
No. They ask where they already work: in Teams, in the browser or on their phone.
Why are there no token costs?
Because the language model runs locally. The costs are therefore fixed and predictable, however intensively the AI is used. Tokens are only charged when a question is deliberately delegated to a cloud AI.
How can I start without taking a big risk?
With the AI and process check. It is a workshop with your management and key people, at a fixed price, credited towards follow-on work and with no obligation. You receive prioritised use cases and a recommendation for your own hardware or a private cloud, with a cost comparison over three years.
The next era starts with a workshop
The first wave of AI showed how big the need is. Now that need becomes something that belongs to your company: an AI that knows your knowledge and your systems, runs in your own environment and follows your rules. The next era of AI is not louder, it is closer: closer to your data, closer to your processes, closer to your team.
Find out more about the company AI in your own rack on the product page. Or take the first step right away:
HABB Switzerland, Rüegsauschachen. Swiss Quality · Precision · Reliability.