moresophy makes enterprise AI transparent and manageable with ContextChat 2.0.

A new skills architecture combines live data, organization-specific knowledge, and transparent reasoning within an enterprise AI platform.

Munich, September 10, 2026 – Munich-based AI specialist moresophy introduces ContextChat 2.0. The new version expands the enterprise AI application with modular skills, live queries from corporate systems, optional knowledge modules, and a step-by-step derivation of results. As a result, ContextChat no longer merely handles individual questions about company data, but now manages complete processing workflows: from analyzing individual pieces of information, through synthesizing them into an overall assessment, to planning and implementing specific next steps.

This addresses a key challenge in the enterprise use of AI: business-critical answers are rarely readily available in a single document. The relevant information is usually scattered across contracts, reports, databases, and operational systems. ContextChat brings this data together, organizes it by subject matter, checks against guidelines, and incorporates up-to-date information.

“The next phase of enterprise AI will not be determined by ever-larger models, but by whether companies can master their data and understand its significance,” says Prof. Dr. Heiko Beier, founder and CEO of moresophy. “Those who cannot demonstrate which data, rules, and calculations led to a result cannot responsibly automate business-critical decisions. ContextChat 2.0 makes precisely this line of reasoning transparent.”

Skills make technical expertise reusable

In ContextChat 2.0, companies set up assistants for specific task areas, such as requests for proposals, insurance claims processing, or framing analysis of media reports. Each assistant combines multiple skills into a seamless workflow. Each skill handles a clearly defined specialized task.

For each skill, you can specify which questions it should handle, what subject matter expertise and evaluation criteria apply, and what data sources, models, or additional resources it requires. Once set up, companies can reuse these building blocks in various assistants. The development environment supports automated testing and version control for robust, auditable AI processes.

Subject matter experts can adapt existing skills to their requirements using a no-code approach. DAPHY^®, moresophy’s control technology, handles the data-driven orchestration .DAPHY^® links the designated skills with the relevant data and combines them to match the request. To do this, the user does not need to program or define every process step as a workflow.

Live data complements the document analysis

Many enterprise AI applications rely on Retrieval-Augmented Generation (RAG). ContextChat 2.0 goes a step further: The platform connects documents to databases, ERP systems, and other line-of-business applications. While processing, it retrieves up-to-date information directly from the connected systems.

This multi-source fusion combines the relevant data points for a unified analysis. For example, a chemical company or industrial supplier can use a bid preparation assistant to link requirements from incoming documents with product data, supporting documentation, current inventory levels, and delivery information. Based on this, the assistant prepares a reliable evaluation and bid calculation.

From Case-by-Case Review to Portfolio Analysis

ContextChat 2.0’s multi-source fusion is not limited to individual cases. Thanks to the underlying ContextSuite architecture, the same business logic can also be applied to entire portfolios—across individual cases, contracts, or reports. This enables ContextChat to perform portfolio-level analyses that are difficult to implement reliably across large datasets using traditional AI agents.

This allows an insurer not only to review a single clause, but also to analyze entire contract portfolios for systematic risk or cost patterns. A media company can extend framing analysis beyond a single article to reveal reporting patterns across hundreds of articles and over extended periods of time. This portfolio-level analysis provides a depth of insight that was previously unattainable through document-based individual analyses.

The derivation provides more than just a list of sources

ContextChat 2.0 visualizes the processing path in a new derivation. It shows, step by step, which terms the system has clarified, which data points and sources it has used, which skills and rules it has applied, and which calculations it has performed. Users can thus trace individual statements back to their business and data-related basis.

A lack of evidence and conflicting information do not disappear just because the answer sounds plausible. If the company has defined a business rule for such a case, ContextChat 2.0 applies it and documents the step. If no rule applies, the application flags the uncertainty or conflict for further technical review.

This is particularly relevant in regulated processes, where companies must document and monitor automated or AI-supported processing steps in a traceable manner. This ensures that even complex decision-making processes remain verifiable.

Modules capture organization-specific meanings

ContextChat 2.0 enhances assistants and skills with optional modules for specific knowledge and governance tasks. The first available module is Glossary Lifecycle Management (GLM). It helps companies derive organization-specific terms from existing content, describe them in technical terms, and formally define them.

A general language model may know the common meaning of a term such as “copyright ownership.” However, it does not automatically know which technical or legal definition applies within a specific organization. Based on existing sources, the GLM suggests definitions, which subject matter experts review, supplement, and confirm in a human-in-the-loop process. The relevant skills and assistants can then access the approved terms.

The modular architecture is open to additional applications. In the future, standardized process descriptions, for example, could also be managed using the same principle.

The architecture limits the number of model calls and the computational load

ContextChat 2.0 uses AI models based on the specific task at hand. This allows powerful language models to be used for complex analyses, while smaller or locally run models can be used for simpler tasks.

The modular skill architecture keeps prompts short because each skill processes only the context it actually needs for its task. The platform performs database queries, filtering, and calculations analytically, without using generative AI. This allows ContextChat 2.0 to avoid unnecessary model calls and limit token consumption and computational overhead.

Cloud-agnostic operation and usage-based billing

ContextChat 2.0 runs independently of any specific cloud provider. moresophy offers the platform as Software-as-a-Service by default. Companies in regulated or security-critical sectors—such as operators of critical infrastructure as defined by the KRITIS regulation—can also run it in a dedicated private cloud or on-premises. This allows them to choose models, infrastructure, and data storage options based on their requirements and maintain their technological sovereignty.

The pricing model combines a platform fee with a usage-based component. The underlying DAPHY^® technology is already in use—or is being implemented—in media analysis, the insurance industry, music rights management, and small and medium-sized industrial companies. ContextChat 2.0 is scheduled to be available starting September 15, 2026, as a SaaS solution and in dedicated operating models.

About moresophy

moresophy develops enterprise AI for companies that want to reliably deploy AI even in business-critical processes. The technology combines corporate data, organization-specific knowledge, and domain-specific logic, making it transparent how results and recommendations are generated. At the technological core is the Reasoning Orchestrator DAPHY^®, which orchestrates data and AI models in such a way that conclusions can be traced back to their sources. The solutions connect existing enterprise systems and data sources across system boundaries to form a shared knowledge base—independent of individual AI models or cloud providers. The Munich-based company has been developing AI solutions for productive enterprise use since 2001.