TECHNOLOGY

AI that understands documents and gets work done

For more than 30 years we have been developing artificial intelligence that does not merely read complex documents but understands, assesses and makes them usable. From that experience we have built a technology that unites analytical AI, generative AI and agent systems into one whole.

PLANETBRAIN

The Foundation of Our IDP Suite

PLANETBRAIN is the foundation of our Intelligent Document Processing technology. It unites all of our core technology, from the patented PerceptionMatrix to modern agentic methods, and ensures the highest accuracy in analysing documents, images and speech.

Deep Learning

PLANETBRAIN combines over 30 years of experience with the most modern concepts of machine learning and artificial neural networks. Through our own fundamental research we continuously advance our technology.

Holistic Approach

PLANETBRAIN captures and assesses complex perceptions holistically: one approach for text, speech, images, video and other data, instead of many isolated point solutions.

Patented AI Technology

At its centre is the patented PerceptionMatrix, a search and extraction technology that has won multiple international awards and cannot be replaced by alternatives.

A Sustainable Vision

Our strategy is the consistent advancement of our AI through our own research and the integration of state-of-the-art methods. The goal: to evolve PLANETBRAIN step by step toward general AI for document-based work.

OUR THESIS

Knowledge Worker, not Chatbot

What is truly disruptive about artificial intelligence is not the chatbot — it is the Knowledge Worker: an AI agent that autonomously and reliably solves diverse, complex tasks. This is exactly what our technology is built for, applied to document-based business processes.

Chatbot

  • Answers

  • Reacts to questions

  • Needs guidance

  • Provides information

  • Conversation

Knowledge Worker

  • Gets it done

  • Understands the assignment

  • Works autonomously

  • Delivers the result

  • Value creation

Tool or actor: “summarise the quarterly report” is a chatbot question. “produce the quarterly report” is an assignment for a Knowledge Worker. Our pragmatic definition of general AI for document-based processes: autonomous with no hand-holding, diverse across domains, complex multi-step and multi-source, reliable and production-grade.

TECHNOLOGY-CONCEPT

Luna: Precise Reading and Thinking in One Platform

Building on PLANETBRAIN, we have developed a platform with which we implement our solutions. Its core is Luna: a technology concept that brings together two proprietary technologies. IDA handles precise reading, the DCA handles thinking and acting. Both work on top of modern foundation models.

IDA

Precise reading
Pixels → knowledge

DCA

Thinking & acting
Knowledge → result

LUNA

Knowledge Worker
on top of foundation models

IDA: Precise Reading

IDA is our heritage from more than three decades of research in document recognition. It turns pixels into structured knowledge: exact tables, correct numbers, preserved layout. This precision provides a reliable data foundation exactly where pure language models would hallucinate numbers. It is the basis for every downstream AI process.

DCA: Thinking and Acting

The Distributed Cognitive Architecture (DCA) is the technology core for Knowledge Workers. It augments interchangeable foundation models with what transformers cannot deliver on their own: memory, self-correction and guaranteed convergence to a stable result. A proprietary memory controller takes on executive control.

This advantage comes from architecture, not model size. It cannot simply be copied by larger models. Every advance in foundation models automatically raises our baseline; IDA and DCA add their value on top.

JAIDE is built on this platform: the Knowledge Worker that translates this technology into document-based processes.

DCA – DISTRIBUTED COGNITIVE ARCHITECTURE

An Architecture Modelled on the Brain

The DCA derives its design principles from the organisation of the neocortex. It is biologically grounded and formally analysable. At its centre is an atomic building block of two poles that converge together toward a stable result.

Pol Biological Analogue Function
World Model
(Foundation Model)		 Neocortex Knowledge and language. An interchangeable foundation model that can be swapped for a more capable one at any time.
Memory-Controller Thalamus, Hippocampus & präfrontaler Cortex Executive control. Orchestrates a three-level memory comprising working memory, episodic memory and long-term memory.

The Convergence Loop

Expectation, action by the world model, observation, reconciliation and update by the memory controller. This loop repeats until the system converges to a stable fixed point: a reliable interpretation or answer.

Fraktal Composition

Agents orchestrate agents, nestable to any depth. From the orchestrator through sub-orchestrators to specialists, the tree emerges per task and folds back when it is done. Every node is itself an atomic building block.

What Transformers Cannot Deliver Alone

Pure model size reaches its limits as soon as adaptive control is required. We deliver memory, self-correction and guaranteed convergence as formal theory and as implementation. This is precisely where the DCA comes in.

JAIDE

Platform Technology as a Finished Product

JAIDE builds on Luna and handles document-based processes autonomously. IDA delivers structured data, the DCA controls the process. The result is not a conversation partner but a completed work output.

IDA as the Reading Layer, DCA as the Thinking Layer

Other AI assistants work directly on incorrectly parsed documents. JAIDE does not. IDA reads first: structured data from scanned PDFs, handwriting and complex table formats. Only on this basis does the DCA draw conclusions, check and decide. That is not a detail but the core of the approach.

Traceable Results, Not a Black Box

Every output includes source references and highlights in the original document. Structured workflows replace open input fields: JAIDE guides users through industry-specific processes and delivers reproducible quality, regardless of who is working with it.

What JAIDE delivers in concrete processes is shown on the JAIDE product page.

TECHNOLOGICAL VISION

Step by Step Toward General AI for Documents

We built a platform on the foundation of PLANETBRAIN to power our solutions. At its core is Luna: a technology concept that brings together two proprietary technologies. IDA handles precise reading, DCA handles reasoning and action. Both work on top of modern foundation models.

Built for AI

Our platform consistently follows the #aifirst principle: concepts and structures are designed to be clearly understandable for AI as well. Every new generation of foundation models can be used directly, for new solutions and for advancing the platform itself.

Sovereignty by Design

Our technology is cloud-native and Kubernetes-ready, running on sovereign hosters as well as fully on-premises or in air-gapped environments. Processing stays where the data lives. Privacy-friendly from the ground up.

Model-Independent & Future-Proof

The foundation model can be swapped without touching agent logic or memory. Inexpensive agents handle routing and checking, while powerful models are used only when they are truly needed.

Knowledge Workers do not emerge from ever-larger models, but from complementary technologies on top: memory, controller dynamics and precise reading. Exactly what foundation models lack. This is the through-line of PLANET AI: not a pivot to AI, but a course pursued consistently over many years.

PERCEPTIONMATRIX

Deep Understanding of Text

The patented PerceptionMatrix is the core technology behind IDA. Its output is an abstract metadata format that is markedly superior to traditional OCR techniques.

  • Language-independent

  • Foundation for downstream processing

  • Patented, non-replaceable process

From Pixels to Structured Knowledge

Instead of plain text, the PerceptionMatrix delivers a rich, contextual representation: recognised content together with its spatial arrangement, relationships and meaning. It is precisely this data foundation that enables the reliable thinking and acting of the downstream Knowledge Workers.

Technology That Creates Value

From over 30 years of AI research, precise reading and a cognitive architecture, we build Knowledge Workers for document-based processes. Made in Germany.