No digital filing cabinet: AI document analysis for real estate portfolios

The discussion about AI in the real estate industry currently has two main camps. On the one hand, there is the expectation that entire job profiles will disappear in a short time. On the other hand, there is the conviction that in the end everything will remain the same anyway. Sebastian Deppe, EVANA's VP of Sales considers both positions unhelpful. The reality, he says, "lies, as always, in shades of gray." This was his phrasing in a conversation with... Commercial Real Estate Today.

The real problem lies in the document inventory.

Those who hold real estate for decades don't manage a database, but rather a complex, organically grown portfolio: leases with amendments, appraisals, handover protocols, invoices, correspondence. Much of it is in PDF format, some as scans, and some of it is stored in a filing cabinet without being digitized.

The usual reflex is to file things away. Most companies use providers like SharePoint or Dropbox for this. Over the years, this leads to the growth of unwieldy folder structures and document graveyards. Even with standardized rules, these systems will eventually become unwieldy.

A document management system (DMS) creates order, but it doesn't answer any questions. Anyone wanting to know which contracts in their portfolio contain an indexation clause with a specific threshold still has to search manually.

Task	Classic DMS	EVANA
Storing and retrieving documents	Yes	Yes
Automatically extract content	no	Yes
Answering open questions about the inventory	no	Yes
Source reference for each statement	no	Yes
Transfer of structured data to ERP and existing systems	restricted	Yes
Mapping from the object level to the societal level	no	Yes

Nico Schröder Aachener Grundvermögen describes the initial situation precisely: They were looking for "not a digital filing cabinet," but a system that automates processes and can handle large volumes of documents. You can read about how the project was set up in the article. Data extraction via AI: Aachener Grundvermögen uses EVANA.

What AI document analysis can actually achieve

EVANA EVANA is an AI platform for the real estate industry that extracts, structures, and analyzes data from documents, processes, and external sources. With EVANA, we bring structure to data and documents. The AI assistant EVA analyzes documents, extracts the relevant content, structures it, and delivers it in a usable format to downstream systems. Instead of hours of searching through hundreds of documents, the result is a structured database.

This is made possible by domain knowledge, not by a language model alone. EVANA AI, The engine at the core of the platform classifies over 400 document types from the real estate and legal sectors, including land registry extracts, leases and service contracts, energy performance certificates, valuation reports, maintenance logs, and utility bills. It has been trained on more than 20 million relevant data points since its founding in 2015 and continues to be refined. This classification is essential for making the analysis precise enough for productive use.

For CEO Martin Teiber The next step is already commonplace. Speaking with documents is "no longer a vision." Users ask an open-ended question and receive an answer based on all stored documents, including correspondence.

The difference to general language models like ChatGPT, Claude, or Gemini lies in its traceability: Every statement in EVA is linked to specific documents, and every result is output with source citations within the document. Anything not substantiated is not claimed. The use case demonstrates the extent of this approach. Funding review: An unstructured stack of documents becomes a searchable data space, EVA answers the questionnaire, and the application is ready in minutes instead of days.

Clean data is a prerequisite, not a byproduct.

The point most frequently overlooked in the AI debate is that every analysis is only as reliable as the underlying data. Schröder's conclusion on this is clear.

Only with clean data can you analyze and evaluate it reliably.

Nico Schröder, Aachener Grundvermögen

That's why EVANA doesn't stop at extraction. About EVANA360 Structures from the object level to the company level can be mapped, existing partial solutions are merged, and the platform connects modularly to existing enterprise software. Multilingual processing is an added benefit for internationally distributed teams. EVANA functions reliably across multiple languages, regardless of whether a document is in PDF, scan, image, or email attachment format.

One current component is the AI-based transaction data room (EVANA TDR), Available as an add-on for EVANA360 or as a standalone SaaS application, it works across all types of real estate transactions, is ISO 27001 certified, and its application complies with GDPR, BaFin, and DORA regulations. A role-based access control system manages access for buyers, sellers, and advisors, while a central communication module structures Q&A queries. Anyone who has ever experienced due diligence with incomplete documentation knows how crucial data quality is in such situations.

What this means for the next few years

Martin Teiber's assessment is uncomfortable: Market participants need to understand the importance of clean data. Those who fail to grasp this will find themselves at a disadvantage in the medium term.

The competitive advantage doesn't come from the most spectacular model, but from the ability to keep one's own portfolio readily analyzable at all times. This may seem unspectacular, but it's the foundation for everything else.

Frequently Asked Questions

What is the difference between EVANA and a document management system?

A document management system stores and sorts files, but doesn't analyze their content. EVANA automatically reads the content, converts it into structured data, and uses this data to answer content-related questions about the entire archive. The two systems are not mutually exclusive: EVANA can be integrated with an existing DMS.

Why is ChatGPT insufficient for real estate documents?

General language models provide answers without reliable source verification and are not trained on real estate-specific document types. EVANA assigns each statement to the underlying document and works with over 400 document types from the real estate and legal sectors. This traceability is essential for decisions with legal or economic implications.

Which documents can EVANA process?

EVANA processes all common types of documents used in the real estate industry, including land registry extracts, leases and service contracts, energy performance certificates, valuation reports, maintenance logs, and utility bills. It processes digital files as well as scans, images, and email attachments, in multiple languages and without pre-sorting.

How reliably does the AI recognize a document?

The classification accuracy rate is over 98 percent. This is based on more than 20 million data points in the training and a combination of deep learning methods that analyze both the text and the raw image of a document. Classification only occurs when a user-defined threshold is reached.

Is the deployment compliant with GDPR and regulatory law?

Yes. The transaction data room is ISO 27001 certified and compliant with GDPR, BaFin, and DORA regulations. This makes it suitable for use by regulated entities such as asset management companies and institutional investors.

What is an AI-based transaction data space?

An AI-based transaction data room is a digital data space for real estate transactions that not only provides access to uploaded documents but also automatically classifies, extracts, and makes them available for analysis. For due diligence, this means that the entire database is searchable and content-queried from day one, instead of requiring manual review.

To the source

The full interview with Martin Teiber, Sebastian Deppe and Nico Schröder was published on July 9, 2026 in Handelsimmobilien Heute under the title „Speaking with documents is already a reality for us“. Read the interview on hi-heute.de