FAQ

• At its core, a local AI that runs entirely on your premises.
• Included are the hardware required, several local AI models and the central knowledge base.
• We first train the knowledge base on industry knowledge, then tune it to your data and processes.
• Your data stays local and never leaves your company.
• Further project work can be agreed individually where needed.

Cloud solutions come with three drawbacks.

• Data leaves the building: sensitive data is cached by external parties and used to train other people’s models. With a local AI it stays in the building, and nobody trains foreign models with it.
• Lack of precision: general models are trained broadly. Only your own knowledge base covers the special cases in which you differ from the crowd.
• Running cost: cloud services bill in tokens, either capped per subscription or charged per request. Local models know no such limit.

Barriers to entry have fallen considerably, and company size matters little at first. Typical use cases:

• Categorising and answering email traffic
• Capturing and allocating invoices
• Costing quotations on the basis of historical data

Where the benefit is greatest for you is something we clarify in two free, no-obligation conversations.

Not much is needed:

• One contact person with decision-making authority who knows your core processes.
• Read access to the central systems so we can build the adapters to the knowledge base.

What you do not need is AI expertise of your own, completed digitalisation or perfectly maintained data. Structures that have grown over time are our daily business, and we look at everything else together in the potential analysis.

• Time saved on repetitive work, which frees capacity for orders and strategy.
• Shorter waiting times: an email no longer sits unread for twenty minutes.
• Relief on work that generates no direct revenue, above all quotation costing.
• The effect accumulates over time and becomes visible on the cost side.

• Hardware: a dedicated server that stands on your premises once the project is complete. Several performance levels, purchase or financing over six months, yours either way.
• Setup: installation on site including an introduction, as a fixed lump sum depending on the option chosen, travel included.
• Both items are one-off. The server remains an asset with a resale value.
• Recurring: a fixed monthly amount for providing your local AI.
• Optional: further applications, bookable individually and never mandatory.

• The requirement follows from your tasks: data volume, number of users and the response time you want.
• A hardware package in one of several performance tiers follows from that.
• The investment consists of two one-off items, hardware and setup, plus a fixed monthly amount.
• The hardware remains your property and keeps a resale value.
• We name concrete figures in the second conversation, once we know your requirement. Both conversations are free, so you learn the order of magnitude without any obligation.

• Our target: first automation results within one to two months.
• Full implementation within three to six months.
• We put these markers in writing at the start, so that you can hold us to them.
• The range depends on your level of digitalisation and system integration.
• They rest on the founders’ experience from comparable projects and on the prototype we are currently building. COMPBRAIN is in its exploratory phase, and we do not claim a routine that does not exist yet.

• Potential analysis: needs and room for improvement, free of charge.
• Second conversation: implementation concept, also free of charge.
• After that we refine the approach together and the project phase begins.

COMPBRAIN is in the process of being founded. Behind it are the prospective founders Paul Spenlen and Christopher Mertens.

• Paul Spenlen is responsible for the technology, meaning the implementation of the local AI and the connection to your systems.
• Christopher Mertens is responsible for process analysis and automation approaches as well as the commercial side, including financial and liquidity planning.

• Your contracting party will be the company COMPBRAIN.
• A German GmbH is currently the intended legal form.

• Paul Spenlen: building automated workflows in email traffic.
• An automated application procedure for master’s students at a German university. It pre-screens several thousand documents per semester and saves several working days.
• Christopher Mertens: two years in transformation and performance consulting for German mid-sized businesses, from automotive to chemicals.

• All data sits on a local system on your premises.
• There is no permanent connection to the outside, only the link to your systems and users. For maintenance you open a limited access and close it again afterwards.
• That keeps access to your data in your own hands, without the detour via an external cloud provider.

• The server, your data and the knowledge base stay in your building.
• The local models keep running without us.
• There is nothing to switch off and no licence that expires.
• The backup is the same base, simply not in operation, and it sits in the same network under the same permissions.

• We treat your expertise like all other data, meaning locally and confidentially.
• All data can be accessed locally only.

• Mistral and Qwen are among the models used.
• A wide range of further local models is available alongside them.
• For daily use you pick the model that fits your task.

• We compile the industry knowledge ourselves, from our own research and not from customer data.
• Your system therefore starts out already knowing the terms and workflows of your industry.
• On site, the local model extends that base with your data, questions it and adjusts it.
• What results is a model that knows both your industry and your company.
• That adjusted state stays on your premises. We do not use it, neither for ourselves nor for other customers.

• The server sits in your network, on your hardware.
• The knowledge base and the models live on the same machine.
• The backup sits in the same network and follows the same permissions.
• None of it leaves your building.

• The local server holds COMPBRAIN, the knowledge base in which the information from your systems comes together.
• Only the condensed content is stored, not the original files.
• Individual adapters connect COMPBRAIN to your systems.
• Once set up, it keeps itself up to date.

• The productive system runs offline in your network and has no route outwards of its own.
• Alongside it sits a mirrored layer through which individual access points can be opened when needed.
• That is the route by which we bring in security updates and new model versions.
• You determine scope and access: you release what is needed and close it again afterwards.
• Without your release nobody reaches the system from outside, ourselves included.

• Selected managers feed in new information themselves through a straightforward interface.
• In the first month after go-live we stay in close contact and adjust the system where needed.
• Adjustments at go-live and afterwards are made by us via remote access. You grant that access and can close it again at any time.
• After that a dedicated contact person services the server annually and takes up your questions.
• For serious problems you contact your dedicated contact person directly, including outside regular business hours.
• Extensive training material is part of the application.