§ 8 Laufzeit und Beendigung
Der Vertrag läuft auf unbestimmte Zeit.
Jede Partei kann mit einer Frist von
drei Monate zum Quartalsende
kündigen.
Die Kündigung bedarf der Textform.
AmaraCloud
A knowledge system for documents
AmaraCloud is a document-centred knowledge system with long-term memory for documents of every kind.
AI quick overview
What is the next step?
A conversation about use case and data posture. Onboarding is guided. There is no public self-service.
What is A.M.A.R.A.?
A.M.A.R.A. is Rieger Systems' intelligence layer – currently in development, not marketed as a standalone consumer product.
What does A.M.A.R.A. stand for?
A.M.A.R.A. stands for: A Machine Always Ready to Assist.
Where do AI and data run?
Operations and data flow are aligned to EU infrastructure. The core setup is designed without mandatory dependency on non-European cloud providers.
Do you sell our AI data?
No. Data from AmaraCloud is not sold. Data sovereignty remains fully in Europe, and production compute also remains fully in Europe.
Does AmaraCloud run autonomous agents?
No. People run the search, open the source and keep the decision. There are no free-running agents.
AmaraCloud
Every answer with the passage you can open.
Contracts, email, spreadsheets, images, policies: AmaraCloud keeps documents of every kind in long-term memory. Search does not stop at a sentence. It returns the evidence — snippet, place and the original passage.
Operated in the EU
Data storage and production compute stay in Europe.
No data resale
Data from AmaraCloud is not sold.
Guided onboarding
A pilot with selected organisations. No public self-service.
Knowledge sits in files. Decisions need the passage.
An answer without a place is still a claim.
The same question returns across drives, mail and tickets: which version holds, and where is the sentence?
A filename search and an AI sentence are not enough. Before a decision, the passage in the original has to be open.
The same file, readable three ways
AmaraCloud turns each document into a structure you can check and a machine can read. Automation, analysis and the answer use it. Every value keeps its source.
Machine-readable
The original becomes blocks with a place: heading, paragraph, table, cell, mail part, image region, time span. A spreadsheet stays rows and cells, not a paragraph.
- Word
- Excel
- PowerPoint
- Image
- Audio
- Text
Facts, grouped
Samples. The wording stays. Dates and amounts are also stored normalised. A missing value is not filled in.
Parties
Beispiel GmbH
Issuer
Rahmenvertrag-2024.pdf · Page 1
Amounts
€1,280 net
Offer, 4 September 2026
Angebot-Produkt-X.pdf · Page 2
Dates
2026-10-15
Due, wording 15 October 2026
Rechnung-XYZ-091.pdf · Page 1
Identifiers
XX-AM 204
Registration plate
Fahrzeugzuweisung.pdf · Page 1
Quantities
7 orders
Last one in September 2026
Auftragsliste-XYZ.xlsx · Sheet
Automation
Workflows read fields, not prose: amount, date, identifier, quantity.
Analysis
Cells, quantities and amounts stay comparable, with currency, unit and date.
Artificial intelligence
The answer cites the passage. It does not set a value without a source.
The proof is the passage in the original
AmaraCloud does not stop at a sentence. Every hit names its place and opens there. Exact applies to a text layer or a known time span. A scan and an image region stay Region. A missing page is not invented.
Questions from a working day
Samples. Names, customers, vehicles and amounts are invented. Not a live request.
Choose a question. It runs only after you send.
Which company car has Max Mustermann been driving since this week, and which one was it before?
Searching the sample files
- Fahrzeugzuweisung.pdf · Page 1Since Monday the VW ID.4, plate XX-AM 204
- Übergabeprotokoll.pdf · Page 2Before that the Skoda Octavia, XX-AM 118, returned 18 October
The assignment names the current car, the handover the previous one.
Since this week the VW ID.4, plate XX-AM 204. Before that the Skoda Octavia, XX-AM 118, returned on 18 October.
What is our business with customer XYZ, and how many orders were there between us?
Searching the sample files
- Rahmenvertrag-XYZ.pdf · Page 1Framework contract since March 2023
- Auftragsliste-XYZ.xlsx · Sheet7 orders, the last in September 2026
- Kontaktkarte-XYZ.pdf · Page 1Their contact: Erika Beispiel
The contract establishes the relationship, the list counts the orders.
A framework contract since March 2023. Seven orders in between, the last in September 2026. The contact at customer XYZ is Erika Beispiel.
What price did we last set for product X, and what was the delivery time?
Searching the sample files
- Angebot-Produkt-X.pdf · Page 24 September 2026, €1,280 net
- Preisliste-2025.pdf · Page 8Before that €1,190 net
- Auftragsbestätigung-X.pdf · Page 1Delivery in 3 weeks from the order
The newer offer replaces the list. The delivery time is in the confirmation, not in the price.
Last set at €1,280 net, in the offer of 4 September 2026. The 2025 list said €1,190. The latest confirmation says 3 weeks from the order.
Which invoice to customer XYZ is still open?
Searching the sample files
- Rechnung-XYZ-091.pdf · Page 1€4,600, due 15 October 2026
- Zahlungseingang.xlsx · Row 18No receipt for this number
The invoice names the amount and the due date. The payments sheet has no row for it.
Invoice 091 is open, €4,600, due 15 October 2026. No payment for it is in the list.
Which version of the travel policy applies?
Searching the sample files
- Reiserichtlinie-2026-10.pdf · Page 1In force from 1 October 2026
- Reiserichtlinie-2026-03.pdf · Page 1Replaced on 30 September 2026
- Freigabe.eml · Paragraph 1Management approves the October version
Two versions, one mail for the approval. The newer one replaces the older.
The version of 1 October 2026 applies. The March version is replaced. The approval is in the mail.
Until when does the warranty on the pump in hall 2 run, and which serial is on the plate?
Searching the sample files
- Typenschild-Halle-2.jpg · RegionSerial AB-2041, a region without a text layer
- Rechnung-Pumpe.pdf · Page 1Purchased 12 March 2024
- Garantiebedingungen.pdf · Page 424 months from purchase
The number is only a region in the image. Purchase date and term produce the end, which no line states as a date.
The plate shows AB-2041, as a region, not as an exact text passage. Purchased 12 March 2024, warranty 24 months. That yields 11 March 2026.
Who approved the discount on order 441, and how much was it?
Searching the sample files
- Angebot-441.pdf · Page 38 percent on item 2
- Freigabe-441.eml · Paragraph 2Lea Beispiel, 2 June 2026
The amount is in the offer, the person is in the mail.
8 percent on item 2. Approved by Lea Beispiel on 2 June 2026.
Did supplier Beispiel confirm the delay on batch 12, and until when?
Searching the sample files
- Mail-Lieferant.eml · Paragraph 3Confirmed, new week 42
- Bestellung-Charge-12.pdf · Page 1Originally week 38
- Besprechung-Charge-12.mp3 · 02:10–02:24The same week 42
Mail and recording name the same new week. The order keeps the old one.
Yes. Mail and recording say week 42. The order was week 38.
Did Max Mustermann confirm the March safety briefing?
Searching the sample files
- Teilnahme-März.xlsx · Row 4Confirmed on 14 March 2026
- Einladung-März.eml · Paragraph 1Safety briefing, 12 March
The invitation names the date, the list the confirmation.
Yes. The attendance list shows the confirmation of 14 March 2026 for the briefing of 12 March.
Which facts are structured in the framework contract with customer XYZ?
Searching the sample files
- Rahmenvertrag-XYZ.pdf · Page 1Party: Beispiel GmbH, issuer
- Rahmenvertrag-XYZ.pdf · Page 1Start March 2023, normalised 2023-03-01
- Rahmenvertrag-XYZ.pdf · Page 12Notice period: three months to quarter-end
The facts are grouped. Each value stays on its page.
Issuer Beispiel GmbH, start 2023-03-01, notice period three months to quarter-end. The wording stays on the page.
Can customer XYZ’s order list be read as a table for analysis?
Searching the sample files
- Auftragsliste-XYZ.xlsx · Sheet7 rows, columns order, date, amount
- Auftragsliste-XYZ.xlsx · CellLast order in September 2026
The spreadsheet stays rows and cells. Analysis reads the cells.
Yes. Seven orders as rows, the last in September 2026. Date and amount are machine-readable, and the cell stays the source.
No question selected
Which company car has Max Mustermann been driving since this week, and which one was it before?
What is our business with customer XYZ, and how many orders were there between us?
What price did we last set for product X, and what was the delivery time?
Which invoice to customer XYZ is still open?
Which version of the travel policy applies?
Until when does the warranty on the pump in hall 2 run, and which serial is on the plate?
Who approved the discount on order 441, and how much was it?
Did supplier Beispiel confirm the delay on batch 12, and until when?
Did Max Mustermann confirm the March safety briefing?
Which facts are structured in the framework contract with customer XYZ?
Can customer XYZ’s order list be read as a table for analysis?
Example · not customer data
Each question opens its own source.
Exact. The passage is marked on this page.
| Row | A · Region | B · Window |
|---|---|---|
| 12 | South | weeks 12–14 |
| 13 | East | weeks 15–17 |
| 14 | North | weeks 18–20 |
| 15 | West | weeks 21–23 |
Exact. Only cell B14 is marked, not the whole sheet.
Region. Without a text layer the source stays an area, not an exact character mark.
RegionRegion. The area on the picture is marked. A caption does not replace this source.
Nord kommt zwischen KW 18 und 20.
0:0001:42–01:58 · segment04:12
Exact is the time span. Single words inside that segment have no time of their own.
A place, not a claim
Page, cell, image region or time span. A hit without a place is not a hit.
Open the original
Preview jumps to that place. You read the source, not only the summary.
Honest precision
Exact for a text layer or a known time span. A scan and an image region stay Region. A missing page is not invented.
How a file becomes a structure
Ingest, a machine-readable structure, grouped facts, then the source. Relations between documents are the direction.
Upload
Documents of every kind are ingested, checked and stored with a content hash.
Machine-readable
PDF, Office, spreadsheets, email, images, audio and text become blocks with a place. A spreadsheet stays rows and cells.
Facts
People, organisations, identifiers, dates, amounts, quantities and places, grouped. The wording stays, the value is normalised.
The same structure
Search, automation, analysis and the answer read it. You open the source and decide.
Search becomes checking
Full text carries search. Semantic retrieval can join in without blocking it.
AmaraCloud ingests documents of every kind, writes them in a machine-readable form and attaches every value to a source location. Full text carries search. Semantic retrieval can join in without blocking it.
Relations between documents are the direction. The pilot promises the evidence you can open first — not a finished knowledge graph, and not one industry.
What AmaraCloud also carries
The source is the proof. The structure is what machines read.
Machine-readable
PDF, Office, spreadsheets, email, images, audio and text become blocks with a place.
Facts
Grouped as parties, amounts, dates, identifiers and quantities. The wording stays.
Content hash
Every upload is checked. The hash stays with the document, for files of every kind.
Visible index stage
Each document shows how far ingestion has got.
Same trail
Questions and search return the same evidence.
Tenants
Separate spaces and roles.
Relations
The direction between documents. The pilot delivers the source location first.
Not tied to one industry
Any organisation whose knowledge sits in documents.
Many places
Contracts, email, spreadsheets, images and policies.
Before the decision
Specialists who check an answer.
Repeated questions
Service, operations and backoffice.
Data location
Organisations with requirements on tenant separation.
First cut
Document-heavy teams. That cut is the entry point, not the product boundary.
Limits and responsibility
What AmaraCloud does not promise.
No guarantee of correctness
The evidence is there. The decision stays with people.
No autonomous professional decisions
Not in law, medicine or finance.
No polished precision
A scan is a region, not an exact text span.
No free-running agents
People open the source and decide.
Guided onboarding
A pilot with selected organisations, not public self-service.
The layer under AmaraCloud, not the product itself.
A.M.A.R.A. (A Machine Always Ready to Assist) is the in-house foundation. It is in development and not a finished consumer product.
Short answers
What AmaraCloud is, what counts as evidence, who it is for, and what happens next.
What is AmaraCloud?
A document-centred knowledge system: documents become machine-readable, facts are grouped, and every answer opens its source. For any industry and any use case.
What counts as evidence?
A snippet plus a place — page, cell, image region or time span. You open that place and check it yourself. A scan or a picture without a text layer stays a region. For audio the time span is exact, not the single word.
What does a file become?
Blocks with a place: text, table, cell, email, image region, time span. Grouped facts from that — people, organisations, identifiers, dates, amounts, quantities, places. Dates and amounts are also stored normalised. The same structure feeds automation, analysis and the answer.
Who is it for?
Organisations whose knowledge sits in documents, in any industry. The first pilot cut is document-heavy teams. That cut is the entry point, not the product boundary.
AmaraCloud is available now, on request
We look at the use case, the sources and what evidence has to mean in your day. Onboarding is guided, data stays in the EU, and nothing is sold.