Frequently asked questions

• Define the use case. Specify what need the agent addresses, for whom, and within what scope. For example: answering first-level questions submitted to IT support.


• Connect your data and tools. Knowledge bases, business applications via APIs, third-party tools via the MCP protocol, and user context: the agent has the information needed to take appropriate action.


• Build your agent in Studio. Drag and drop the blocks onto the canvas, configure them (select a template, set instructions and conditions), and then connect them to each other. This is how you define the agent’s knowledge, behavior, and actions: responding to inquiries, creating a ticket, updating a customer record, and more…


• Put your agent online. Save your workflow and link it to a Wikit Semantics application: your users can access it immediately, and you can track its usage in the platform's dashboard.

• Responding to requests submitted by employees or customers to IT support (for this purpose, the IT department may implement a suitable AI agent).


• Processing requests submitted to Human Resources (the HR Director defines which workflows to automate).


• Customer Relationship Management.


• Managing relationships with users of a public service.

Wikit Studio is integrated into the Wikit Semantics platform, which is available as an annual subscription. It is included in our most comprehensive plan, which allows you to design custom workflows. Contact us to learn more about Studio and find the plan that best suits your projects.

Agents created with Studio are Wikit Semantics applications just like any others. Their usage statistics appear on the platform's dashboard, alongside all your other applications.

Yes. The Studio is used solely to design your workflows: no user data passes through it. When your agents are running, they benefit from the same security guarantees as all Wikit Semantics applications: a platform hosted in France by OVHcloud, no model training using your data, and the ability to choose the hosting level for each agent’s models.

Governance takes place upstream. Each role has defined permissions: creation, modification, or view-only access. Administrators also choose which blocks are visible to each user. Your teams can therefore publish content independently, without a lengthy approval process, and using only authorized building blocks.

Yes. An AI agent created with Wikit Studio does more than just respond: it can also create a ticket, update a CRM record, send a notification, or trigger a process in a third-party tool. Since these actions affect your systems, we recommend coordinating them with your teams in advance. We’re here to help you with this.

• query your business applications via their APIs;


• connect to third-party tools using the MCP protocol (calendar, email, CRM, project management, etc.);


• take into account the user's context, such as their profile or the page on which they are asking their question.

For business roles looking to automate tasks and processes with AI—such as project managers, product owners, HR managers, or customer service representatives—they can create their own agents independently, within the parameters set by administrators. Other employees can view the workflows without the risk of accidentally modifying them. For the end user, nothing changes: they interact with an application better suited to their role, without seeing the workflow that drives it.

• Drag the blocks onto the canvas;


• Fill in the few fields provided, such as selecting a template from a list;


• Connect the blocks to each other.

Wikit Studio isthe no-code visual editor for the Wikit Semantics platform. It lets you design your own AI agents by assembling logic building blocks: querying an LLM, calling an API, setting a condition, searching a knowledge base… You’re no longer limited to the ready-to-use applications available in the catalog. Whether it’s a conversational assistant, an autonomous agent, or a RAG chatbot, you can build the agent that perfectly meets your needs.

• cite sources so that users can verify them,


• define what the agent does not handle, and forward the issue to a human if in doubt,


• monitor conversations to correct whatever needs to be corrected. 

Yes, you can certainly combine multiple AI agents from the catalog and have them work together on the same platform. Multi-agent orchestration enables specialized assistants to collaborate on complex tasks that a single agent would not be able to handle as effectively. Since each agent is an expert in its field, the process becomes more precise, and the results are generally better.

The number of agents you can deploy depends on your license: each Wikit Semantics license tier includes a set number of AI applications. You can therefore start with a few agents and then upgrade to a higher-tier license as your usage grows.

• Customer Support: advisors, call center agents, account managers.


• IT Support: help desk technicians, administrators, and application specialists.


• Human Resources: payroll managers, recruiters, HR administrators.


• Customer service: receptionists, administrative staff. 


• Management and Administration: executive assistants, general services, legal department. 

• Answer frequently asked questions about your products, delivery times, return policies, or procedures at any time. 


• Check the status of a case by querying your CRM or ticketing system directly. 


• Classify an incoming request and route it to the appropriate team, using the information that has already been collected. 


• Create and update a ticket from the conversation, without having to retype anything. 


• Notify the customer at each stage of the request processing. 


• Gather feedback on satisfaction after the exchange and identify any areas of concern. 

• number of employee users and number of conversations handled;


• resolving rate of requests without human intervention and average response time;


• average call duration; employee satisfaction rate;


• A reduction in the number of direct requests to the HR team (tickets, emails, calls).

• Answering Employees' Questions


• Identify, locate, and share documents


• Guide employees through the process


• Automate the onboarding process for new employees


• Direct the request to the appropriate contact person when human intervention is required

An HR chatbot is a conversational assistant that can process natural language and answer common questions from company employees on topics such as time off, payroll, professional training, career development, health insurance eligibility, and internal procedures. This tool serves as a first-line HR support resource, available around the clock, without replacing the HR teams.

At Wikit, you can deploy a ready-to-use HR agent from our catalog, or create a custom, agent-based HR agent using the Wikit Semantics no-code Studio.

An AI chatbotfor IT support is not intended to replace technicians. The virtual agent handles recurring, low-value-added requests, allowing teams to refocus on tasks that require greater expertise: complex incidents, infrastructure development projects, and assisting employees with issues that require genuine human expertise. The goal, therefore, is not to replace your support technicians, but to enhance their roles!

Check out CPage's REX on AI for Business Support

• the quality of the existing documentation,


• the desired level of automation (simply answering questions or performing actions in the information system),


• the necessary integrations with ticketing tools (IWS, GLPI, etc.) or corporate directories.

Yes, the Wikit Semantics platform is hosted in France, on an infrastructure where the development, pre-production, and production environments are strictly separate. Your data is isolated on your own tenant, encrypted at rest and in transit, and access is restricted according to the principle of least privilege (granting users only the rights strictly necessary to perform their tasks).

Depending on your needs, granular permission management (RBAC) and detailed access logs can be implemented, along with strong authentication (MFA) and single sign-on (SSO) for all users.

In addition, we conduct an annual security audit and penetration test carried out by a specialized third-party provider.

We significantly reduce the riskof inaccuracies using RAG (Retrieval-Augmented Generation): the chatbot retrieves information from a defined knowledge base rather than drawing generalizations from its internal knowledge. The sources used remain accessible, allowing technical teams to verify the origin of each response. However, no AI technology can guarantee absolute reliability.

In addition, Wikit incorporates several complementary safeguards: query filtering, verification of generated responses, definition of the business scope, and continuous monitoring.

• Automated incident management, including ticket classification and prioritization, diagnostic suggestions based on IT history, etc.


• IT self-service to resolve simple issues: password resets, managing access to tools, hardware requests, etc.


• Answers to frequently asked questions about tools and procedures are available in a dedicated IT FAQ. 


• Directing employees to the appropriate request and/or contact person through an IT services catalog. Centralizing and continuously updating the IT knowledge base.

• password reset,


• hardware failure, 


• equipment request, 


• question about an internal procedure, etc. 

Yes. We treat public data as a strategic asset: hosting in France, encrypted data streams, and strictly controlled internal access.

An AI agent does not collect or store personal information from citizens; interactions focus on your administrative procedures and public information, not on sensitive personal data.

Furthermore, this security is based on verifiable commitments: a documented security assurance plan, an annual risk analysis, and an annual penetration test conducted by a specialized external service provider.

For government agencies, it is essential to understand the criteria on which an AI agent bases its responses. The use of such an agent to manage interactions with users of a public service inevitably raises questions about the reliability of the responses provided to citizens, particularly regarding sensitive topics (personal or professional circumstances, social assistance programs for which they may be eligible, and specific administrative procedures).

• query filtering,


• definition of the business scope,


• referring the citizen to a human agent when the level of confidence in the response is insufficient.

• Level 1, in France: The models are hosted by OVHcloud or Scaleway (Mistral AI, Qwen, DeepSeek, GLM, etc.). This is the highest level of data sovereignty, suitable for the most sensitive data.


• Level 2, in Europe: The models are hosted on Microsoft Azure in European data centers (OpenAI, Mistral AI).


• Level 3, globally: The models are available directly from their developers (Anthropic with Claude, OpenAI) for uses that require the most recent models and do not involve sensitive data.

1. The library of native connectors covers the software available on the market: ITSM, EDM, CMS, CRM, digital workplace, directories, and collaboration tools. Some give your agents access to your knowledge bases, while others allow them to take action (create a ticket, send an email, transfer the conversation to an agent, etc.).


2. The MCP (Model Context Protocol) opens up possibilities for agent-based applications. The platform connects to both the MCP servers provided by Wikit and those you make available yourself, allowing your agents to call your own tools and perform actions within them.


3. Finally, REST APIs cover specific integrations.

• the number of AI applications you can deploy;


• the number of data sources you can connect to your agents;


• connectors to your business tools, including the MCP protocol in the most comprehensive offerings;


• distribution interfaces: chat, API, Microsoft Teams, and even white-label solutions; annual usage volume, in terms of the number of users and requests;


• the level of support, up to unlimited support.

• To centralize the management of your AI agents and their data sources


• To avoid spreading resources across multiple tools that do not communicate with each other


• To ensure functional consistency and overall oversight


• To reduce costs by streamlining the development of AI agents


• To become more agile in order to quickly deploy new use cases