Your questions about AI in Doxis

Doxis offers several deployment models (cloud, on-premises & hybrid) to meet the needs of every type of organization, including those with the strictest security requirements (defense sector included). In cloud mode, data can be hosted in our partners' data centers, with the option for the customer to choose the geographic storage region to meet their data sovereignty requirements.

Yes. Doxis was designed to give organizations a holistic view by contextualizing all information related to their customers, projects, and cases. The platform connects to any ERP, CRM, or Microsoft 365 environment through certified integrations, with bidirectional data exchange: Doxis can both retrieve information from these applications and trigger actions or update data back in them (for example, a candidate status change in SuccessFactors can trigger a business workflow directly in Doxis).

Doxis offers dedicated, certified integrations for the platforms customers ask about most, including SAP, Microsoft Dynamics 365, Microsoft Office 365, Salesforce, Workday, and ServiceNow. A full overview of supported systems is available on the integrations page.

Yes. Doxis is recognized as a leader in document management, in large part thanks to its "metadata factory", the platform's foundational layer that structures and adds value to all information linked to documents. This architecture can process several hundred million documents per year, or even per day, for mission-critical needs. DHL Express demonstrates this ability to absorb very high volumes: with Doxis, it built the Global Imaging Archive, the world's largest commercial document archive, housing over 8.5 billion documents and growing by 130 million new documents every month. Read the DHL Express case study.

Credit management and AI usage can be finely configured by administrators. It is possible to define access to certain types of information by role or profile, enable or disable specific AI capabilities for certain users, and manage all of this at the department level. This addresses both security/governance needs and cost-control needs, ensuring AI credits are allocated to the teams where they generate the most business value.

The core Doxis platform for content and process management does not transfer any data outside an on-premises installation. When it comes to AI, many Doxis AI services rely on large language models, but these are typically privately hosted: the customer provides its own private LLM environment and configures Doxis to use it, so documents stay within infrastructure the customer controls. Data only leaves the customer's environment if cloud-based AI services are deliberately chosen and configured, and which service is used, and where that data is processed, depends on the use case and configuration.

Yes. When an AI assistant connects to Doxis through the Doxis MCP server, the connection authenticates as a specific Doxis user and works strictly within that user's existing permissions. The AI can only access what that user is allowed to see, and every action is traceable to that user account. In Ask Doxi, each answer also shows the documents it was based on, so users can see where the information came from and check it.

Doxis embraces the human-in-the-loop principle, recognized as a key lever for improving model performance and achieving efficiency close to 99% on certain processes. The availability of this human control can vary by product: on some, it is a configurable setting; in other cases, particularly where data is sensitive or classified, different approaches can be adapted to the customer's needs.

Hallucinations are a known risk of generative AI, so Doxis applies different safeguards depending on how AI is used. For document processing, such as classifying documents and extracting data, human-in-the-loop validation lets users review and correct results before they flow into downstream processes. For answers generated in Ask Doxi, responses are grounded in the customer's own documents using Retrieval-Augmented Generation (RAG), and each answer shows the documents it was based on, so users can verify the source.

Doxis takes a "composable AI" approach: rather than depending on a single model, the platform allows a specific AI model to be connected to a given business process, and lets you measure how effective that combination is. If another model proves more effective or better suited to the need, it can be substituted easily, without rebuilding everything. This lets Doxis always rely on the most relevant model for a given use case, rather than "putting all its eggs in one basket."

Thanks to Doxis' granular AI credit management, customers can define which teams, roles, or profiles use these credits, track their consumption, and redirect it to other teams if needed. This ensures optimized credit usage, aligned with workflow performance and the business value generated.

Yes, within clear limits that keep users in control. Through Ask Doxi, or through any AI assistant connected via the Doxis MCP server, users can ask questions in natural language. The AI then searches for and retrieves relevant documents, summarizes information, and answers questions based on the content the user is permitted to access. It works within the user's existing permissions, so it never exposes content beyond what that user could already see.

No. From their browser, within DoxisCube, the AI assistant already enables natural-language interaction. But thanks to the MCP server offered by Doxis, the platform can also be connected to any AI assistant on the market (Copilot, Claude, Teams, etc.). This makes it possible to combine the analytical power of a large language model with the tools made available by the MCP server, so that every action available in Doxis can be triggered, securely and with full governance, from the AI assistants users already work with every day.

No, Doxis AI is designed for three different profiles. End users and lines of business can create their own workflows without IT support, this is, in particular, the whole point of Doxis Business Studio, which allows workflows to be built in natural language, fully independent of IT. Doxis administrators / document management architects, meanwhile, can rely on AI to design an entire metadata and information architecture tailored to a new project or business team, somewhat as if the AI pre-structured a relevant folder tree before the team even begins working on it.

Doxis is designed to understand, search, analyze, and process unstructured documents (emails, images, scanned documents) in many languages. This relies in particular on IDP (Intelligent Document Processing), which combines several complementary approaches: ready-to-use pre-trained models, prompt building to describe the expected processing for a given document type, and, at a more advanced level, the ability to build custom AI models from a customer's own document corpus.

Doxis combines all three approaches, full-text, semantic, and metadata search. With Superhuman Search and Ask Doxi, the platform automatically identifies the key elements of a question (references, customer names, more open-ended queries, etc.) to trigger the most relevant search across the stored document corpus.

Yes, without needing IT. In response to a growing need for fast deployment of business workflows, Doxis lets tech-savvy business users configure their own workflows using low-code / no-code approaches, visual tools such as Node-RED, or directly in natural language via agents. This approach allows business teams, who know their processes best, to create and maintain their own workflows over time.