AI & ontology

An AI that knows your business. Not the internet.

Customer, project, document, payment and employee form one linked data model. Nothing goes out before a person approves it.

Request a demo How it works ↓
AI assistant in Pangea: question on the payment status of Hotel Lindenhof, answer with sources
Assistant. “Yes, €26,400 settled.” Source: invoice, bank transaction, project file Hotel Lindenhof.

The model

Every document knows where it came from and what comes next.

Ontology means: the things in a business are not tables but relationships. A payment knows its invoice, the invoice its order.

1Payment knows invoice

€4,816 match RE-0002. No searching, no manual journal entry.

2Invoice knows order

The progress invoice comes from the payment schedule of AB-0341.

3Order knows project

Project Lindenhof moves to the installation phase. The team gets a push notification.

4Project knows enquiry

The revenue is credited to the ad the customer came from.

Customer, project, quote, order, invoice, payment as one connected model
The model behind Pangea: six objects, connected. That's why the next step follows from the previous one.

Agents

The AI proposes. You approve.

Account coding and document recognition.

Photograph a supplier invoice or receive it by email: the agent recognises supplier, amount, account and project and puts the posting proposal up for approval.

AI approvals in Pangea: the agents' proposals with a reason on every line
AI approvals: 3 proposals, with a reason on every line.

Propose the reminder level.

Overdue items are detected, the reminder level proposed, the text prepared. You approve, the sending runs.

Dunning in Pangea: overdue items with the proposed reminder level
Dunning: payment reminder and 1st reminder, sent.

Draft quote from the enquiry.

From an email or a form, a draft is created with line items from the item master. You check it, then it goes out.

Quote editor in Pangea: draft with line items from the item master
Draft AN-0342 from an enquiry, 14 line items.

Answers with a source.

The assistant answers questions from your business data and names the document, file or posting. Only with what the person asking is allowed to see.

AI assistant in Pangea: answer with names, project and source
“Who hasn't clocked in yet today?” Answer with names and project.

Control and data

Four rules that always apply.

Proposal, not execution

Account coding, payment reminders, quotes: the agent prepares, a person approves. Logged, who and when.

A source for every statement

Every answer links to a document, file or posting. What isn't backed by evidence isn't claimed.

Permissions apply to the AI too

12 roles plus your own. The assistant only answers with data the person asking could see themselves.

Data in the EU

A dedicated instance per business, servers in the EU. Your data trains no third-party models.

Frequently asked questions

What businesses ask us about the AI.

What data does the AI see?
Only your business data in your instance, and only what the respective user is allowed to see. No data from other businesses.
Can the AI send invoices on its own?
No. Agents prepare, people approve. Every approval is logged.
Is our data used for training?
No. Your data stays in your instance and trains no third-party models.
What does the AI cost?
Nothing extra. Assistant and agents are included in every Office licence, with no quotas.
What's the difference to a chatbot?
A chatbot knows the internet. The Pangea assistant knows your business: customers, projects, documents, payments, employees, all linked. That's why it can tell you whether Lindenhof has paid, and how it knows.

Next step

A demo with someone who has issued progress invoices themselves.

30 minutes on your own example: one project from quote to payment, with your numbers.

Request a demo See pricing

Data migration via DATANORM, Excel and DATEV.