Frequently asked questions

AI fundamentals

What are AI safeguards?

An AI guardrail is a mechanism put in place to supervise, limit or correct the behavior of an artificial intelligence system. It aims to avoid drifts, errors or inappropriate responses, by ensuring that the AI respects rules, business objectives or ethical standards. These safeguards can be technical (filters, checks), human (validation by an expert) or organizational (usage policies).

What is MCP (Model Context Protocol)?

The Model Context Protocol (MCP) is a protocol that structures the transmission of context, data and business objectives to AI models, improving the quality and relevance of responses. It enables conversational agents to better understand the global environment, user roles and business rules. At Wikit, we use MCP to design smarter agents, offering more natural, high-performance interactions aligned with each company's needs.

Can AI hallucinate or invent answers?

Yes, it's a risk inherent in all language models. That's why Wikit has chosen to strongly support generative AI with :

- Adding verified documentary content via RAG

- Defining strict response perimeters

- Temperature settings to reduce excessive creativity

In addition, a reformulation system or referral to a human agent can be set up if the AI cannot answer with certainty.

What is RAG (Retrieval-Augmented Generation)?

RAG is a technology that combines an internal search engine with a language model.

In concrete terms, before generating an answer, AI will first look for relevant information in your documents, then use it to respond.

This makes it possible to :

- Limit errors and inventions

- Rely on reliable, controlled content

- Cite sources if necessary

This is an essential pillar of reliability for professional AI applications.

What is NLP (natural language processing)?

NLP, or natural language processing, covers all the techniques used to enable a machine to understand, analyze and generate human language.

NLP is the foundation on which conversational AI is built: it enables us to understand a user's intentions, extract key information from a sentence or formulate a coherent response.

Wikit uses these techniques to enable AI applications to interact effectively with your users.

What is a large language model (LLM)?

A Large Language Model (LLM) is a type of artificial intelligence trained on huge volumes of text to understand and generate natural language.

These models, such as GPT or Mistral, are capable of :

- Understand a question asked in everyday language

- Formulate coherent, well-written, contextual responses

They don't "know" anything themselves, but learn linguistic structures and draw on data to produce relevant content.

What is generative AI?

Generative AI is a branch of artificial intelligence capable of creating content (text, images, code, etc.) from a simple natural language query. It relies on language models (LLMs, for Large Language Models) trained on huge volumes of data. These models can then :

- Generate answers to questions

- Summarizing documents

- Translate, reformulate, write texts or even produce lines of code

In Wikit, generative AI is used to understand users' intentions and produce personalized, contextualized and useful responses from your data.

What is an AI agent?

An AI agent is an autonomous entity orchestrated by AI, capable not only of answering a question, but also of intelligently executing complex actions. It can take different forms:

- Conversational chatbot,

- Job assistant,

- Interactive application integrated into a tool, and it can be simple (enriched FAQ) or more complex (multi-step scenario connected to your systems).

An AI agent can :

- Analyze a request

- Extract data in different tools

- Trigger actions (create a ticket, send an e-mail, access an API, etc.)

- Interact with multiple sources of information and make simple decisions based on business rules

Wikit Semantics platform

What is Wikit's pricing model?

The Wikit Semantics platform is offered as a flexible subscription, with several service levels tailored to your specific needs. The cost varies according to a number of criteria, such as the size of your organization, the number of use cases to be deployed, the features chosen (multi-IA, connectors, personalized support), etc.

We offer you a tailor-made pricing model, so that you can benefit from the most suitable solution at the right price. Contact us for a personalized quote.

Is it possible to test the Wikit Semantics platform?

Yes, we offer 3-month POCs (Proof of Concept). This allows you to test and evaluate the solution in your own environment before making a commitment.

During this period, you can explore the platform's functionalities, test its integration with your existing tools and check that it meets your needs. This test enables you to validate the solution in a real-life setting, with the support of our experts to accompany you throughout the process.

What are the advantages of using a unified AI platform rather than separate tools?

Using a unified platform like Wikit Semantics allows you to :

- Centralize the management of your AI agents and their data sources

- Avoid dispersion among multiple tools that don't communicate with each other

- Functional consistency and global supervision

- Reduce costs by industrializing the creation of AI agents

- Gain agility to rapidly deploy new use cases

Can Wikit Semantics be connected to various AIs (ChatGPT, Mistral, etc.)?

Absolutely. Wikit Semantics is a multi-IA agnostic platform, which means that you can choose the generative AI model(s) you wish to use according to your challenges (performance, sovereignty, costs or regulatory constraints) and change at any time according to your needs and market developments: ChatGPT (OpenAI), Mistral, LLama (Meta), etc.

Is the Wikit Semantics platform no-code?

Yes, completely! The Wikit Semantics platform is based on no-code logic, which means you can create your AI applications without advanced technical skills. Of course, the Wikit teams are always on hand to help.

What is an AI application in Wikit Semantics?

An AI application is a customized solution designed on the Wikit Semantics platform to meet a specific business need. It combines data sources and business rules to, for example :

- Automatic responses to user queries

- Assist an agent in daily tasks

- Helping users with administrative formalities

- Generate personalized content or responses

Each AI application can be deployed online, integrated with existing tools, or used in-house.

Who is the Wikit Semantics platform designed for?

The Wikit Semantics platform is aimed at all organizations wishing to exploit generative AI in concrete, operational ways: companies (HR, support, IT, customer relations, etc.), local authorities (citizen reception, communication, urban planning, etc.), business teams wishing to automate processes or provide intelligent assistance, IT departments looking for a secure, interoperable and sovereign solution, etc.

What are the main features of the Wikit Semantics platform?

The Wikit Semantics platform offers :

- A no-code editor for designing and configuring AI applications without writing a single line of code.

- An AI agent orchestration system for combining multiple intelligences in a single workflow.

- Ready-to-use connectors to access your data (SharePoint, GLPI, websites, etc.).

- A supervision interface to monitor interactions and improve performance.

- The ability to choose from several AIs (OpenAI, Mistral, etc.), according to your preferences and needs.

IA use cases

How are local authorities using Wikit Semantics?

Many local authorities use Wikit Semantics to :

- Set up citizen chatbots to provide information on administrative procedures, municipal services, opening hours and events.

- Lighten the load on receptionists by automating responses to frequent requests.

- Offer a service that's accessible 24/7, without having to mobilize teams outside opening hours.

- Facilitate access to information on specific topics: town planning, civil status, education, etc.

- Some go even further, with internal assistants for agents, or integration into citizen portals.

What are the most frequent use cases for the Wikit Semantics platform?

The Wikit Semantics platform is designed to meet a wide variety of business needs. Among the most common use cases :

- HR assistance: answering employees' questions about payroll, leave, training, etc.

- IT support: automate ticket management, help resolve level 0 to 2 technical problems.

- Customer or citizen relations: provide information, guidance and support for online requests (public services, procedures, opening hours, etc.).

- Customer relations: providing commercial or technical answers, guiding customers through the purchasing process.

- Enhance internal documentation: interrogate a body of documentation (procedures, regulations, internal guides).

Chatbot AI

Is it possible to combine chatbots and AI agents in the same application?

Absolutely. The chatbot can be one of the entry points to a more global AI application, in which several agents collaborate (for example: a receptionist, an HR agent, a business agent...). Wikit lets you orchestrate this collaboration to deliver an optimal user experience.

On which channels can I deploy my AI chatbot?

The Wikit Semantics platform enables you to deploy your AI chatbot on a wide variety of channels, depending on your organization's needs and your users' preferences: Website (via a widget or dedicated page), desktop application, ticket management systems (via ITSM integrations like GLPI or Jira), internal portals (Intranet, collaborative platforms), messaging systems (Teams), and so on.

Wikit gives you total flexibility to choose the channel or channels that best suit your users, while guaranteeing a smooth, seamless experience on every platform.

Can you supervise or correct your AI chatbot's responses?

Yes, you can monitor conversations, spot misunderstandings or possible improvements, and adjust content accordingly. The platform offers an administration space for supervising, correcting and developing your chatbot over time.

How does Wikit ensure the reliability of my AI chatbot's responses?

Wikit integrates advanced technologies such as RAG (Retrieval-Augmented Generation), which enables AI to draw on your reliable data to generate its answers. You can also configure rules and safeguards, and visualize the sources used to respond, to guarantee quality and consistency.

What are the most common AI chatbot use cases?

The most widespread use cases include: assistance to citizens with administrative formalities (local authorities), HR support for employees (access to procedures, vacations, teleworking, etc.), IT support (help with resolving incidents, hardware requests), customer support, dynamic FAQs, customer guidance, or even business co-pilots.

Are Wikit chatbots multilingual?

Yes, the AI models used by Wikit understand many languages. So you can design chatbots capable of interacting with your audiences in French, English, Spanish or other languages, depending on your needs.

What's the difference between an AI chatbot and a traditional chatbot?

A traditional chatbot often relies on rigid decision trees and predefined responses. A Wikit chatbot, on the other hand, leverages the capabilities of generative AI (such as GPT, Mistral, etc.) to provide contextualized, natural responses, drawing on your internal information sources. It can also be part of a larger AI application, with several orchestrated intelligent agents.

Can the Wikit platform be used to create AI chatbots?

Yes, the Wikit Semantics platform makes it easy to design chatbots powered by generative artificial intelligence. These chatbots can be used to answer users' questions, automate support or guide users through the process. Thanks to Wikit's no-code approach, you can create a chatbot without any technical skills.

Connectors and integration

Are external data synchronized automatically?

Yes. Thanks to the Wikit Connect platform, you can program the frequency of synchronization of your data sources: every night, every week, or on demand.

The platform automatically checks for changes (new documents, modifications...) and updates the databases used by your AI agents. This ensures that your answers are always up to date, without manual intervention.

How do AI applications access internal data?

AI applications designed with Wikit access data via secure connectors that ensure :

- Intelligent content indexing

- Automatic data updates at defined frequencies

NB: Data is never transmitted "raw" to the AIs: it is pre-processed to guarantee the relevance of responses, while ensuring confidentiality.

What kind of connectors does Wikit offer?

Wikit offers two main connector families:

- Data connectors, which automatically exploit knowledge sources (SharePoint, websites, document databases, intranets, etc.).

- "Action" connectors, which enable interaction with business tools (GLPI, IWS, HR tools, CRM...), for example to create a ticket, retrieve a status or trigger an action.

These connectors make it easy to seamlessly integrate your AI applications with your existing tools, without any special development.

Can we connect a Wikit chatbot to our internal databases (SharePoint, website, etc.)?

Yes, thanks to the connectors on the Wikit Semantics platform, you can integrate your knowledge bases (SharePoint, websites, documents, business databases, etc.) so that the chatbot can access the right information in real time. This ensures reliable, up-to-date answers, adapted to your context.

Security, compliance & sovereignty

Is the Wikit Semantics platform RGPD compliant?

Yes, the Wikit Semantics platform has been designed with privacy by design and full compliance with the RGPD in mind. This implies:

- Fine-tuned management of personal data,

- The ability to anonymize or pseudonymize content,

- Treatment traceability,

- Controlled conservation options.

Where are the data processed by the Wikit Semantics platform hosted?

The data processed by the Wikit Semantics platform is hosted in France by OVHcloud.