Zum Inhalt springen

Focus: local AI and Pseudomizer. Together. More

Industry partners wanted. Become a partner

EIGENKERNLocal AI
AI-generated

Explainer films

Understanding AI, step by step

Sixteen short films for everyone just starting with AI: from “What is a cloud?” to an AI of your own. A narrator explains in plain words. Plus nine films for your industry and everyday cases in detail. All examples are invented.

1 · Understanding

What is this, anyway?

Cloud, ChatGPT, local AI: the basics in one minute each, without jargon.

AI-generated0:46

What is a cloud?

Film text

Everyone talks about the cloud. But what is it, really?

Quite simple: a cloud is somebody else's computer.

No fog, no sky. A building full of computers.

When you ask a cloud AI something, your question travels there.

Your documents then sit on other people's computers, often in another country.

With a local AI, it's different.

Your documents stay where they are created: on your device, in your house.

The cloud is neither good nor bad.

The real question is: who holds the key?

More on that in the next film.

Also as a short film (vertical, German): Instagram

AI-generated0:52

What happens to my data when I ask ChatGPT?

Film text

You type a question into ChatGPT. What happens next?

Your question leaves your computer and travels to the provider's computers.

It gets answered there. And the conversation stays stored.

On the personal plans, one setting is switched on by default: your conversations may improve the AI.

You can switch it off. But only if you know about it.

Once you have sent something, it is no longer in your hands.

So here is the rule of thumb: don't type anything that couldn't also be written on a postcard.

Customer data, contracts and invoices don't belong there.

How to do it safely anyway, we show in the next films.

Also as a short film (vertical, German): Instagram

AI-generated0:46

What is local AI?

Film text

Local AI. What does that actually mean?

The AI does not run in someone else's data centre, but on a small computer at your company.

The computing happens at your place. Your documents never leave the house.

You ask, and the AI answers from your own documents. With the source.

You don't pay per question or per person. The whole team can use it.

Eigenkern sets up the device, connects your documents and looks after it.

Remember this: your data plus your knowledge make your own AI.

More on that in the next film.

AI-generated1:02

Local or cloud: the comparison

Film text

Cloud AI or local AI? Five questions make the difference.

First: where is your data? With cloud AI, at the provider, often abroad. With local AI, at your place.

Second: who learns from your data? In the cloud, that is possible depending on the contract. Locally, the AI only learns within your company.

Third: what does each question cost? In the cloud, a subscription per person or per request. Locally, no cost per question.

Fourth: does it work without the internet? Cloud AI does not. The device on site answers offline too.

Fifth: where is the computing done? In the cloud, at the provider. Locally, on your device, with the source.

Ask these five questions before you decide on an AI.

More on that in the next film.

AI-generated0:38

How does an AI read my documents?

Film text

How does an AI actually arrive at its answer?

Without looking things up, it answers from memory. Then it invents whatever sounds plausible.

That is where wrong numbers come from.

With look-up, it first searches your documents and then answers.

And the source comes with it.

Experts call this RAG. In plain words: answers bound to sources.

You can recognise it when a system also says what it does not know.

More on that in the next video.

AI-generated0:38

What is an AI of your own?

Film text

Your own AI. What does that mean?

Experts say: private language models with evidence-based answers on your own hardware.

Sounds complicated. It is three things at once.

First: a language model that belongs only to you.

Second: the answers rely on your documents.

Third: a device that stands at your place.

In short: an AI in your own house that only answers from documented sources.

Level two is about trust.

2 · Trust

What may go where?

Which documents may go to the cloud, which may not, and how to do it safely.

AI-generated0:53

Can this go to the cloud?

Film text

Can this go to the cloud?

Big cloud AIs are powerful.

Your documents don't belong there.

That's why you see in advance what goes out.

On the left, what you have: Ms Sample from Sample Limited.

On the right, what goes out: contact person Alpha from company Beta.

The meaning stays. The amount stays: four thousand eight hundred euros.

If the check finds something, nothing goes out.

And what goes out is recorded. Like a flight recorder.

You can prove it at any time.

Your documents stay behind your door.

We'll show you, using your own documents.

AI-generated0:53

Security and your documents

Film text

Before an AI comes into the company, the data protection team asks: where are the documents? And what goes out?

With local operation, everything stays on the device in your house.

Or on a server just for you, in a data centre in Germany.

A cloud AI only gets something if you explicitly book it. And then only a checked copy without names.

You get the documents for your approval on request: the data processing agreement, the technical measures, the deletion concept.

And we also say what we do not have: no ISO certificate of our own.

Security means: ask, check, and prove it.

More on that in the next video.

AI-generated0:48

Three ways: local, hybrid or your own server

Film text

Local, hybrid or your own server. Three ways to your own AI.

Local: everything runs on the device in your house.

With very many documents, the first read-in then takes longer.

Hybrid: the device stays the core. On request, a data centre in Germany does the first read-in of large collections.

Own server: the AI runs on a server just for you, in Germany. No device on site is needed.

With hybrid and your own server, a data processing agreement is part of it.

Local, hybrid or your own server: you decide.

More on that in the next video.

AI-generated0:54

Data, learning and data protection

Film text

Does the AI learn from my data? Yes. But only at your place.

The knowledge stays on your device. Other companies' models are not trained with your data.

You decide who sees what: only me, the whole company or certain groups.

The administration does not see the users' conversations.

Training only happens with your yes. Names and personal details are made unrecognisable beforehand.

If you stop, everything is deleted. On request, you get an export beforehand.

Remember this: learning only happens at your place. Your data stays your data.

In level three, you get to know Eigenkern.

3 · Eigenkern

Our solution

How the AI works in your house without your originals leaving it.

AI-generated0:50

What is Eigenkern?

Film text

A call, a query, an appointment. And you search: in folders, in emails, in people's heads.

The knowledge is already there. It is just spread everywhere.

Imagine someone who has already read everything. And who sits in your house.

They forget nothing and do not get impatient when you ask for the third time.

They read what comes in, find what you are looking for, and prepare. You look it over and approve.

That is Eigenkern: your AI that lives with you.

Our principle: we do not sell anything we do not use ourselves.

More on that in the next video.

AI-generated0:54

MLKI: local AI in your house, in four steps

Film text

One device. All employees. No cost per question. That is MLKI, the local AI in your house.

First: set it up. The device arrives ready configured. Connect power and network.

Second: connect your documents. You decide what the AI may read. We start with your biggest annoyance.

Third: choose specialist modules. Law, engineering, tax, HR and more.

Fourth: create your team. Everyone only sees what has been released to them.

Then you ask, for example: which clients have not yet submitted their receipts?

And the answer comes with its source. From your documents.

More on that in the next video.

AI-generated0:46

The strongest AI in the world. Without your originals.

Film text

The world's most powerful AI lives in the cloud.

Your documents live in your house.

Until now, that meant: either powerful, or secure.

Eigenkern separates the two.

The original stays in the house.

Only a twin goes out.

The key stays with you.

The cloud AI works with the twin,

and your answer comes back with the real names.

Two knowledge stores. One answer.

Not a concept paper. It's running.

We'll show you, using your own documents.

AI-generated0:41

Two knowledge stores, one answer

Film text

Some tasks need the strongest AI in the world. But your originals should stay in the house.

That is why there are two knowledge stores.

The first knows your originals. It stays in the house and works with your local AI.

The second only knows the twin, without names. Only that one works with the cloud AI.

Both contribute. And you get one answer, with the real names.

Two knowledge stores. One answer. And the original never leaves the house.

More on that in the next video.

AI-generated0:49

Devices and costs

Film text

Which device does my company need? And what does it cost?

For small teams: a compact, quiet computer on the desk.

For many documents: a small-format AI computer with 128 gigabytes of memory.

For larger companies: a device for continuous operation in the server room.

No device on site? Then the AI runs on a server just for you, in Germany.

You buy hardware and software once. After that, you pay nothing per question and nothing per head.

What it costs exactly, Eigenkern tells you after the preliminary analysis of your documents.

More on that in the next video.

AI-generated0:46

How to get started

Film text

How do you start with your own AI? Not with a big project.

Bring your biggest annoyance. We use it to show what AI can do in your company.

What costs the most time comes first. Then the rest follows.

Test with your own documents. Or talk to us first.

What it costs, we tell you after the preliminary analysis of your documents.

We accompany larger companies with an on-site analysis.

Remember this: first your biggest annoyance, then the rest.

In level four, you see AI in your industry.

4 · Your industry

AI in your industry

Nine short films, one per industry: what changes in daily work when the AI stays in your house. All examples are invented.

AI-generated0:49

Industry: the knowledge of production

Film text

Industry: the knowledge sits in manuals, maintenance plans and quality documents.

But when the machine stops, everyone searches in different folders.

With Eigenkern, you simply ask: what does the manual say about the error message on line three?

The answer comes from your documents, with the page in the manual.

Inspection reports become searchable, without going to an AI provider.

Operating data from controls and sensors can be connected too. The figures come from fixed evaluations.

Everything stays in the house. Eigenkern sets up and looks after the system.

Test it with the documents from your own company.

To the industry page

AI-generated0:54

Tax firm: AI for the tax firm

Film text

Tax firm: the knowledge sits in files, tax notices and correspondence.

But client data is confidential. That is why AI stays out of many firms.

With Eigenkern, you ask: what did we agree in 2024 on the depreciation of the farm shop?

The answer comes from the file and the tax notices, with the source.

Incoming receipts are sorted, gaps are spotted, and reminders are prepared as drafts.

Deadlines for VAT returns and objections are shown. The professional check stays with you.

No client file goes to an AI provider. Eigenkern sets up and looks after the system.

Test it in your own firm.

To the industry page

AI-generated0:47

Trades: site and office in one place

Film text

Trades: quotes, photos and records end up everywhere, just not in one place.

The master craftsman is on the building site, and the office is searching.

With Eigenkern, you ask: which paint did we use for the Meyer family?

The answer comes with photo and invoice, even by voice message from the phone.

Measurements and photos from the site become a draft quote. Approval and sending stay with you.

You ask data sheets and manufacturer documents like a colleague.

Your customer data stays in the business.

Test it in your own business.

To the industry page

AI-generated0:45

Law: AI with client confidentiality

Film text

In law, client confidentiality always comes first.

Today, files, briefs and rulings are searched one by one.

With Eigenkern, you ask: where did we argue limitation in similar cases?

The answer comes from the file, the briefs and your own collection, with the source.

Client letters and summaries are prepared as drafts, in the tone of the firm.

Deadlines are found in the documents. Assessment and signature stay with you.

No document goes to an AI provider.

Test it in your own firm.

To the industry page

AI-generated0:45

Medical practice: AI for the practice, data stays inside

Film text

Medical practice: patient data is the most sensitive thing a practice has.

And new staff ask the same questions every day.

With Eigenkern, you ask: how is the reprocessing of dental handpieces regulated?

The answer comes from the quality manual and the hygiene plan, with the source.

Duty and holiday planning for the whole team. Sick notes and cover in a few clicks.

For organisation and administration, not for diagnosis or therapy.

Patient data does not go to an AI provider.

Test it in your own practice.

To the industry page

AI-generated0:50

Hotel: the knowledge of the house

Film text

In a hotel, promises, allergies and manuals are spread across emails, lists and heads.

And the guest asks: is there nut in today's dessert?

With Eigenkern, the kitchen asks exactly that.

The answer comes with recipe and data sheet, or with a note that a data sheet is missing.

An enquiry and the price list become a draft reply. Checking and sending stay with reservations.

Each department sees its own part. The kitchen sees the allergy, not the staff file.

Everything runs on one device in your house.

Test it in your own hotel.

To the industry page

AI-generated0:47

Administration: bylaws, plans, minutes

Film text

In offices and administrations, plans, contracts and bylaws sit in folders and emails.

And every citizen enquiry starts with searching.

With Eigenkern, you ask: may I build a carport on this plot?

The answer comes from the development plan and the bylaw, with the source.

Ask plans, expert reports and project files. Minutes and reports as drafts.

Contracts, service charges and appointments for each unit at your fingertips.

Everything runs in your own network, without a third-party cloud.

Test it in your own office.

To the industry page

AI-generated0:45

Agriculture: your farm, your data, your AI

Film text

On the farm, contracts, operating data and plant knowledge are kept, and that is where they should stay.

But the answer is often in a folder from ten years ago.

With Eigenkern, you ask: when does the lease end for the meadow by the stream?

The answer comes from the lease, with the source.

For biogas plants on the farm, the machine learning system answers from your control system's measurements.

Calculators help with feedstocks, quantities and gas yield.

Your data stays on the farm, not in someone else's cloud.

Test it on your own farm.

To the industry page

AI-generated0:44

Education: learn AI without giving away data

Film text

Schools and education providers need AI that keeps their documents in the house.

Because today it is often unclear where inputs end up.

With Eigenkern, you ask: what do our exam regulations say about retakes?

The answer comes from your own regulations, with the source.

Courses and videos on local AI guide the team step by step.

No prior knowledge needed: you get a ready-to-use tool.

The device stands in the school. The documents stay on the device.

Test it in your own school.

To the industry page

5 · In detail

Everyday cases in detail

For each industry the most common tasks from real job ads, played through step by step: from the question to the approval. More industries will follow.

AI-generated1:25

Medical practice: questions at the front desk

Film text

The front desk of a medical practice: this is where most questions arrive.

A patient asks by email what she needs to bring to her first appointment.

The question is simple. But it comes every day, and every time it interrupts the work.

Let's see how the AI in the house handles it, step by step.

First: the question arrives by form or email. The AI reads it.

Second: it recognizes what it is about. Here: documents for the first appointment.

Third: it searches the practice's own documents, for example the information sheet for new patients.

Fourth: it writes a draft reply and names the source.

Fifth: the staff member at the front desk reads the draft and approves it.

For medical questions the AI decides nothing. Those go to the practice team.

All data stays in the practice, on the device in the house.

The result: the reply is ready quickly, and the front desk has time again for the people in front of them.

The takeaway: The AI prepares. The person approves.

Try it with the documents of your practice.

To the industry page

AI-generated1:28

Medical practice: the room for a stress ECG

Film text

The treatment room: every examination starts with preparation.

A new medical assistant asks: What belongs in the room for a stress ECG?

The answer is in the practice's work instruction. But it sits in a binder, and her colleague is with a patient.

Here is how the AI in the house helps, in five steps.

First: the question arrives, on the tablet in the treatment room. The AI reads it.

Second: it recognizes which examination is meant.

Third: it searches the practice's own work instructions, not the internet.

Fourth: it puts together the checklist and names the source.

Fifth: the practice management has approved the work instruction. New versions are reviewed before they apply.

Whether and how to examine is decided by the doctor. The AI gives no medical instructions.

The work instructions stay in the practice, on the device in the house.

The result: the new colleague quickly feels sure, and the experienced one is not constantly interrupted.

The takeaway: The AI finds. The practice decides.

Try it with the work instructions of your practice.

To the industry page

AI-generated1:24

Medical practice: moving an appointment

Film text

Appointment scheduling: this is where the calm of the day is decided.

A patient writes: I can't make Tuesday. Can I move my appointment?

Now someone has to check the appointment book, find free slots and write an email. For a single appointment.

Here is how the AI in the house goes about it, in five steps.

First: the message arrives. The AI reads it.

Second: it recognizes the patient and the appointment.

Third: it checks the practice's appointment book for free times the following week that fit the type of appointment.

Fourth: it writes a draft reply with two suggested times.

Fifth: the staff member at the front desk checks the suggestions and approves them.

Nothing is booked until the patient has agreed and the front desk enters it. The AI books nothing on its own.

Appointments and patient data stay in the practice.

The result: the patient quickly gets two suggestions, and the front desk keeps a clear head.

The takeaway: The AI suggests. The front desk books.

Try it with the appointment book of your practice.

To the industry page

AI-generated1:26

Medical practice: the blood sample and the pickup

Film text

The practice lab: here everything depends on the right time.

An employee asks: By when must the blood sample reach the lab to be picked up today?

The pickup schedule hangs somewhere, and it has exceptions. A mistake means the sample sits overnight.

Here are the five steps the AI in the house uses to help.

First: the question arrives. The AI reads it.

Second: it recognizes that it is about pickup by the lab.

Third: it searches the practice's own documents, the lab's pickup schedule and the sample instructions.

Fourth: it writes the answer with the time and names the source.

Fifth: the practice management has approved the pickup schedule. If it changes, it is read in again.

How a sample is taken and handled is decided by the specialist staff. The AI only says what is in the schedule.

The schedule and all lab documents stay in the practice.

The result: the sample goes out on time, and nobody has to search the schedule.

The takeaway: The AI shows what the schedule says. The person decides what to do.

Try it with the lab documents of your practice.

To the industry page

AI-generated1:20

Medical practice: the insurer's request

Film text

Practice administration: this is where the requests land that nobody likes to handle.

A health insurer asks in writing for proof regarding a treatment case.

That means: find the file, find the right template, write the letter. And all of it in line with data protection.

Let's see how the AI in the house prepares this.

First: the letter arrives. The AI reads it.

Second: it recognizes what the insurer asks for and by when.

Third: it searches the practice's own templates for the matching reply letter.

Fourth: it fills in the draft and names which template it came from.

Fifth: the practice management checks the draft and approves it before anything goes out.

Which details about a patient may be sent is decided by the practice. The AI sends nothing on its own.

All patient data stays in the practice, on the device in the house.

The result: the reply is prepared quickly, and the deadline is in view.

The takeaway: The AI drafts. The practice sends.

Try it with the templates of your practice.

To the industry page

AI-generated1:22

Tax firm: what is missing for the tax return?

Film text

The tax firm: here every tax return begins with a stack of documents.

A client writes: Is anything still missing for my tax return?

Finding out means: open the file, compare receipts, check the list. From scratch every time.

Here is how the AI in the house goes about it, step by step.

First: the message arrives. The AI reads it.

Second: it recognizes the client and the year in question.

Third: it compares the firm's own documents, the client's checklist and the receipts already received.

Fourth: it writes a draft reply with the missing documents and names the source.

Fifth: the tax assistant checks the draft and approves it.

The tax assessment stays with the tax advisor. The AI decides nothing about it.

Client data stays in the firm, on the device in the house.

The result: the client knows at once what is missing, and the firm saves itself the searching.

The takeaway: The AI prepares. The person approves.

Try it with the documents of your firm.

To the industry page

AI-generated1:17

Tax firm: how was this booked last year?

Film text

The annual accounts: months of bookings come together here.

An employee asks: How did we book the garage invoice last year?

The answer is in last year's file. But for every client that means hundreds of pages.

Here is how the AI in the house helps, in five steps.

First: the question arrives. The AI reads it.

Second: it recognizes the client and the previous year.

Third: it searches the firm's own file, the booking notes and the audit report.

Fourth: it writes the answer and names the source.

Fifth: the employee sees the source and decides whether to book it the same way.

How a transaction is treated for tax purposes is decided by the tax advisor.

The file stays in the firm, on the device in the house.

The result: the answer is there quickly, and the accounts move forward without long searching.

The takeaway: The AI finds the passage. The expert decides.

Try it with the files of your firm.

To the industry page

AI-generated1:17

Tax firm: when is the advance payment due?

Film text

Client care: this is where the trust a tax firm lives on is built.

A client asks by email: When is my next advance payment due?

He expects a quick answer. But the due date is in the notice, and the notice is in the file.

Here are the five steps of the AI in the house.

First: the email arrives. The AI reads it.

Second: it recognizes the client and the type of question.

Third: it searches the firm's own documents, the latest tax notice and the deadline calendar.

Fourth: it writes a draft reply with the due date and names the source.

Fifth: the tax assistant checks the draft and approves it.

The AI gives no advice on amounts or tax planning. The tax advisor does that.

Client data stays in the firm.

The result: the client feels well looked after, and the firm needs hardly any effort.

The takeaway: The AI searches and drafts. The firm is responsible.

Try it with your clients' documents.

To the industry page

AI-generated1:17

Tax firm: assigning an incoming invoice

Film text

Bookkeeping: a stack of invoices arrives here every day.

A client's incoming invoice arrives as a scan.

It has to be read, assigned to a supplier and given a booking suggestion. Invoice after invoice.

Here is how the AI in the house prepares this.

First: the scan arrives. The AI reads the text.

Second: it recognizes supplier, invoice number and date.

Third: it compares with the firm's own records, the supplier list and earlier bookings of this supplier.

Fourth: it makes a booking suggestion and names what it was based on.

Fifth: the bookkeeper checks the suggestion and books.

Booking happens only after approval. The AI books nothing on its own.

Invoices and master data stay in the firm.

The result: the groundwork is done, and the bookkeeper checks instead of typing.

The takeaway: The AI prepares. The bookkeeper books.

Try it with your clients' invoices.

To the industry page

AI-generated1:16

Tax firm: incoming payment without a reference

Film text

Financial accounting: in the end every figure has to match a receipt.

On the bank statement there is an incoming payment with no reference.

Who paid, and for which invoice? Finding out often takes half an hour.

Here is how the AI in the house goes about it.

First: the booking line arrives. The AI reads it.

Second: it recognizes amount, date and sender.

Third: it searches the firm's own records for open invoices with the same amount.

Fourth: it suggests the matching invoice and names the source.

Fifth: the bookkeeper confirms or picks a different one.

If several invoices could fit, the AI does not commit. It shows all of them and asks.

Bank statements and invoices stay in the firm.

The result: searching becomes choosing, and the monthly close goes faster.

The takeaway: The AI shows the options. The bookkeeper decides.

Try it with the bookings of your firm.

To the industry page

Bring us your biggest annoyance.

One conversation, one real task from your daily work. We use it to show what AI and automation can do for you.