AI & automation from Switzerland · Aargau
We connect ChatGPT, Claude and local models to your systems, from Abacus, Bexio and SAP to the controls on the shop floor, and train your people so they keep building on their own. Your data stays with you.
Free and without obligation. Reply within 24 hours.
Official partner
Lecturers at
Services
Where do AI and automation pay off measurably for you? Answered with the calculation in Swiss francs.
ChatGPT, Claude or local models access Abacus, Bexio, SAP and your production systems.
Agents take over routine work. Where it matters, a person still decides.
We teach at ABBTS, Haufe Akademie and Klubschule Migros, and in your company.
A potential analysis typically takes two to eight weeks; an MCP integration from pilot to production four to six.
For your industry:ManufacturingFiduciary & finance
MCP & AI agents
Through the Model Context Protocol, ChatGPT, Claude or local models access your systems. Read-only by default, and whatever someone may not see in the source system, they will not see in the chat either.
Which open receivables above CHF 10,000 are older than 60 days?
Three items, CHF 48,210 in total. The largest: invoice 2026-0412, open for 74 days.
Source: Abacus, read-only · permissions of the signed-in user
And why did line 3 stop yesterday afternoon?
Two stops, 47 minutes in total: material shortage at the infeed (31 min) and a light-barrier fault (16 min).
Source: OPC UA / MES, read-only
Partners & data sovereignty
OpenAI
OpenAI Partner Network
ChatGPT Enterprise, the OpenAI API and agents
Anthropic
Claude Partner Network
Claude, Claude Code and MCP, which Anthropic created
NVIDIA
NVIDIA Partner Network
Local models on your own hardware when data must not leave the building
We work directly with the vendors and know new models early. We still recommend whatever fits your case.
Data stays in the source system
An MCP server reads only what each answer needs. Nothing is exported, copied or used for training.
Runs on your premises or in Switzerland
Integrations run in your infrastructure or in a Swiss data centre. If needed, entirely without cloud, with local models on your own hardware.
Swiss FADP and EU AI Act planned in
We provide the architecture details your data protection officer needs, and take the EU AI Act into account if you supply the EU.
Why Kapitel H
We make ourselves unnecessary, not indispensable. Source code, documentation and know-how go to your team.
We work in finance and in live production ourselves. We know what a month-end close and a machine stoppage cost.
Consulting that reaches both the books and the machines: data science meets electrical engineering and automation.
Official partner of OpenAI, Anthropic and NVIDIA. We recommend the model that fits your case.
Runs in your infrastructure or in Switzerland, local models where needed. Data protection under the Swiss FADP designed in from the start.
We teach AI and automation at ABBTS, Haufe Akademie, Klubschule Migros and IBAW.
Our approach
Lasting transformation only works when your people can drive it themselves. Our goal: you no longer need us.
Your company becomes independent. Staff are trained, MCP servers run in your environment, and source code and documentation are handed over to you.
In-depth interviews and process analysis. We listen before we act, and quantify the potential in hours and francs.
2 to 8 weeks
One clearly scoped use case goes live, with your systems and your data.
Feasibility 1 to 2 weeks, pilot 2 to 4
Selected staff are trained as internal AI champions and carry the knowledge into the team.
3 intensive days, then coaching
Your IT takes over, or we run and extend the integrations. You decide.
Handover or annual contract
References
Reference case
600 employees, around EUR 950 m revenue
Potential rated with the calculation disclosed and a conservative discount, before implementation.
The AI champions training sped up our digital transformation enormously. After just three days our team could identify and implement the first use cases on its own.
Kapitel H's hands-on seminars took our controlling team to a new level. We finally understand how to integrate AI profitably into our processes.
Not PowerPoint consulting, but real implementation. The structured method with in-depth interviews and MVP identification delivered clear results for us.
Quotes translated from German.
Team
You work directly with the two founders from Sisseln, Aargau, not with a rotating project team.
Principal Data Scientist · Applied AI & LLM agents
Finance, data & AI agents
Consulting on AI-driven process automation since 2017. At Kapitel H he owns the commercial side: connecting ERP, accounting and documents to language models without data leaving the building.
IT/OT project lead · Industrial automation
Production, IT/OT & automation
Electrical engineer who works in live production every day. At Kapitel H he owns the production side: making machine data from OPC UA, PLCs and MES queryable and, with an approval chain and audit log, controllable.
Questions
Your question is not listed? Write to us.
Kapitel H is a Swiss AI and automation consultancy based in Sisseln, Aargau, founded by Dr. Maher Hamid and Schady Hamid, B.Eng. Four services: AI potential analysis, MCP development and AI agents that connect your own systems, process and industrial automation, and training for staff.
Yes. Kapitel H is a member of the OpenAI Partner Network, Anthropic's Claude Partner Network and the NVIDIA Partner Network. For clients this means direct contact with the vendors and early knowledge of new models. Our advice stays neutral: we use the model that fits the use case.
That can be planned in. MCP servers run in your infrastructure or in a Swiss data centre and read only what each request needs. Where data must not leave the building at all, Kapitel H uses local language models on your own NVIDIA hardware. Swiss FADP requirements are considered from the start.
On the business side, among others Abacus, Bexio, SAP S/4HANA, SAP Business One, Microsoft 365 and, for clients in Germany, DATEV. In production, OPC UA, controllers such as Siemens S7, Beckhoff and B&R, and MES and control systems. The prerequisite is an interface with a clean permission model.
Usually wherever qualified people do routine work: checking receipts, preparing quotes, gathering data from several systems, writing production reports. The potential analysis quantifies this in hours and francs, with the calculation disclosed, before anything is invested.
It depends on scope: company size, number of departments involved and, for MCP, the source system and its permission model. Kapitel H therefore publishes no list prices. A quote in Swiss francs follows a free first call in which the scope is clarified.
In-depth interviews with key departments, analysis of processes and systems, prioritised use cases with ROI assessment, then pilot, implementation and on-site training. A potential analysis typically takes two to eight weeks; an MCP integration from pilot to production four to six.
That is explicitly not the goal. The principle is enablement over dependency: staff are trained, MCP servers run in your environment, and source code and documentation are handed over to you.
Yes. Kapitel H is based in Sisseln AG, UID CHE-350.522.012, with a focus on German-speaking Switzerland, and also works in Germany and Austria. Projects can be run in English; AI integrations are in production at companies in Switzerland and Germany.
Contact
In a free first call we work out where AI gives you the biggest lever and how your team can use it on its own.
Prefer e-mail? anfrage@kapitelh.ch