- FAQ
Frequently Asked Questions
Discover MIA, our AI-powered software platform, along with our secure and highly customizable local AI API. Explore our training programs and AI conferences designed to help your teams successfully adopt and leverage generative AI technologies.
Software
Internal AI Chatbot
MIA is an artificial intelligence software designed to help organizations fully leverage their internal data in a simple, secure, and efficient way. Built on an advanced RAG (Retrieval-Augmented Generation) architecture, it acts as a conversational assistant capable of searching, analyzing, and retrieving information from your documents and business systems.
Deployed on-premise, MIA ensures complete data confidentiality while delivering high-performance capabilities tailored to professional and enterprise use cases.
What is MIA software?
MIA is an AI-powered conversational assistant designed to query, analyze, and leverage your company’s internal data securely and efficiently.
Built on a RAG (Retrieval-Augmented Generation) architecture, MIA retrieves information directly from your business documents and data sources — including PDFs, databases, knowledge bases, and enterprise files — to generate accurate, context-aware responses in real time.
Fully deployed on-premise, MIA ensures complete data privacy and sovereignty, with no information ever sent to external cloud services or third-party AI providers. This makes it an ideal solution for organizations requiring secure, compliant, and fully self-hosted generative AI infrastructure.
What is the difference between your solution and ChatGPT or other SaaS tools?
The main difference is that MIA is a solution designed for professional, secure, and fully customized use, unlike tools such as ChatGPT, which remain general-purpose.
It is configured according to your specific business needs and relies on your own data to deliver accurate, actionable, and context-aware responses.
Unlike SaaS solutions, MIA is deployed on-premise, ensuring complete data confidentiality with no exposure to foreign legal frameworks such as the Cloud Act. Finally, it provides a collaborative interface that is better suited to enterprise teamwork and internal use cases.
What use cases are covered (customer support, HR, legal, IT, etc.)?
MIA covers a wide range of enterprise use cases, from support functions (HR, legal, IT, customer support) to highly specialized business domains.
It leverages your internal documentation to adapt to your organization and deliver precise, context-specific answers, even on complex technical or expert-level topics unique to your industry.
Thanks to its fully on-premise deployment and complete data confidentiality, there are no functional limitations on usage. You can safely leverage MIA across all your data, including the most sensitive information, without compromising security or compliance.
What types of companies can use your solution?
MIA can be used by any type of organization, including private companies and public institutions, with a particular fit for SMEs and mid-sized enterprises (SMEs & mid-cap companies).
The solution is fully scalable and can support use cases ranging from a few dozen users to several hundred, or even thousands, depending on organizational needs and infrastructure requirements.
It is designed as part of a broader effort to democratize enterprise AI adoption, with a lightweight infrastructure and cost-efficient pricing model, making advanced generative AI accessible while still meeting the high standards required in professional environments.
How are my data and their security handled with your solution?
Your data remains fully secure and never leaves your environment.
MIA is deployed on-premise, meaning no information is ever sent to external servers: all data is stored and processed directly within your own infrastructure.
This ensures complete confidentiality, with no exposure to foreign legal frameworks, and enables full compliance with GDPR requirements, particularly regarding the control, storage, and processing of sensitive data.
What types of documents can be integrated (PDF, Word, databases, etc.)?
MIA can integrate a very wide range of document types — more than thirty formats in total.
It supports all standard business file formats such as PDF, Word, PowerPoint, and Excel, as well as their open-source equivalents (OpenDocument formats). It can also process more technical files including JSON, SQL, JavaScript, Python, and accounting-specific formats such as FEC (French Audit File).
Images are also supported, including JPEG and PNG formats, thanks to built-in vision models and OCR capabilities that allow text and information to be extracted directly from visual content.
In addition, MIA can connect directly to your existing environments such as SharePoint, Google Drive, or internal servers, enabling it to leverage your existing file structures, repositories, and databases without requiring data migration.
Is the chatbot multilingual?
Yes, MIA is fully multilingual.
The interface can be used in any language, and the chatbot is capable of understanding and responding in more than 180 languages, as well as in most programming languages.
This makes it an ideal solution for international organizations, enabling seamless communication and consistent access to information across global teams.
Which AI models are used?
At Artemia, we are fully model-agnostic when it comes to artificial intelligence. We continuously monitor the AI landscape and select the most high-performing models available on the market, regardless of the provider. Our choices are based on rigorous internal and external benchmarking conducted during both testing and production phases.
For our clients, this means their solutions continuously evolve over time: as soon as a more advanced or better-suited model becomes available, we can seamlessly upgrade their system to ensure it remains state-of-the-art, without any additional cost.
How does the software installation process work?
After subscription, a technical scoping meeting is organized to precisely size the infrastructure (GPU requirements, concurrency, service quality, etc.) and to prepare the integration of the solution within your environment.
The deployment is then very fast: it typically takes place over a single day. The morning is dedicated to the technical installation (infrastructure setup, connection to the company network, testing, and validation). The afternoon focuses on the human aspect, with end-user training to ensure immediate adoption and effective use of MIA from day one.
What are the technical requirements (servers, GPU, etc.)?
The technical requirements mainly depend on the number of users and the expected use cases. MIA can be deployed on your existing infrastructure if it already includes compatible GPU resources. Otherwise, dedicated servers can be set up and properly sized according to your needs.
However, the solution remains highly accessible thanks to a “frugal AI” approach, optimized for efficiency and resource usage. This significantly reduces hardware requirements and makes deployment possible even for small and medium-sized organizations.
During the technical scoping phase, we assess your environment and define the most appropriate configuration to ensure the best balance between performance, scalability, and infrastructure cost.
How many users can use the chatbot at the same time?
There is no fixed limit to the number of concurrent users that can use MIA simultaneously.
Capacity depends entirely on the underlying infrastructure, in particular the sizing of GPUs and servers allocated to the deployment.
These parameters are defined during the onboarding and technical scoping phase, before deployment, in order to precisely match system performance to the expected usage volume. This ensures a smooth and responsive user experience, even with a large number of simultaneous users across the organization.
Is there a user administration system?
Yes, MIA includes a comprehensive user administration system.
It allows you to manage access rights and permissions, as well as licenses, passwords, and user groups. It also supports integration with authentication systems such as Single Sign-On (SSO) and Active Directory.
In addition, you can monitor platform usage (number of queries, documents added, activity levels, etc.) to track adoption, identify power users, and support employees who are less active in using the solution.
This provides full visibility and control over user engagement while ensuring secure and centralized identity management across the organization.
How much does MIA software cost?
MIA software is based on a two-stage pricing model.
First, a one-time initial investment covers setup, development, customization, deployment, integration, and user training.
Second, a recurring subscription fee applies for the software license. This includes all updates, new features, as well as ongoing improvements and upgrades to the underlying artificial intelligence models.
The pricing model is based on a per-server flat rate rather than a per-user fee. This allows clients to maximize the use of their infrastructure, benefit from full flexibility in the number of users, and achieve a lower overall cost compared to traditional per-user or consumption-based pricing models on the market.
Do you offer a demo or a POC?
Yes, we offer a demo of MIA.
First, we organize an initial discussion followed by an online demonstration, based either on generic use cases or on the specific use cases of the interested company. If needed, we can also work directly on the client’s real data, with a non-disclosure agreement (NDA) in place when required.
Second, we do not offer a traditional POC as the solution is fully custom-built. However, our contracts are standardly set on a 24-month basis with a three-month exit clause, allowing clients to validate the solution in real operating conditions, without functional limitations or usage restrictions.
Automate
Custom AI API
Our custom artificial intelligence APIs enable the integration of advanced generative AI capabilities directly into your business tools (ERP systems, CRMs, internal applications, and more). Designed on a tailor-made basis, they adapt to your specific use cases in terms of performance, speed, and business domain specialization. Deployed on-premise, they provide full control over your data, improved cost management, and seamless integration into your existing information system.
What is a custom AI API?
A custom AI API is an interface that enables you to integrate generative artificial intelligence capabilities outside of the MIA interface, directly into your business tools such as an ERP system, CRM, IDE, or any other application.
It allows you to add features such as analysis, summarization, content generation, translation, and automated data processing. It can also be used to build workflows and automations between applications using tools like n8n.
In this way, it replaces generic APIs such as those from OpenAI or Mistral AI, offering a solution that is tailored and optimized for your specific use cases, with improved cost control and enhanced confidentiality thanks to on-premise deployment.
The local API is fully customized according to your needs: choice of AI models, level of reasoning, response speed, and scalability depending on the number of users and expected workloads.
When is it better to use an API rather than a turnkey software solution?
It is preferable to use an API rather than a turnkey software solution when you want to industrialize a use case — meaning when you need to automate a task that is performed frequently and at scale. In this context, the API allows you to structure the process and significantly increase the level of automation.
Another relevant use case is when you want to bring AI as close as possible to the data itself, without going through an intermediate interface. This helps eliminate back-and-forth between different tools and enables seamless integration of artificial intelligence directly into your existing systems and workflows.
How do you design a custom API?
When we design a custom API, we start with a scoping discussion with the client to clearly understand the need and the specific use case. This allows us to identify the expected capabilities of the AI models, particularly in terms of reasoning ability, response speed, and possible domain specialization (for example, code generation or a specific industry domain).
We then size the required infrastructure, especially GPU resources, to ensure the expected performance levels based on user volume and request load. Based on these elements, we select one or more suitable AI models, whether general-purpose or specialized, lightweight or high-performance.
We are model-agnostic and work with multiple providers to always choose the best combination for each use case, ensuring optimal performance, efficiency, and relevance.
Can the API evolve over time with new models?
Yes, absolutely. The API evolves over time, and this evolution is included in our service offering.
As a model-agnostic provider, we continuously monitor the entire AI ecosystem. We rely on both internal benchmarking and external benchmarks to evaluate model performance.
When relevant, we update or replace the models in use to ensure our clients always benefit from a state-of-the-art solution, aligned with the latest advancements in artificial intelligence.
Is the API hosted locally or in the cloud?
The API is systematically deployed on-premise, which is at the core of our expertise.
This approach ensures complete data confidentiality, particularly for sensitive use cases such as customer data processing, internal information, or source code. All processing remains within the client’s own infrastructure, with no exposure to external environments.
It also provides a significant cost advantage, as intensive usage is not billed per request or per token like traditional cloud APIs (e.g., OpenAI), but instead relies on a dedicated and fully controlled infrastructure.
How are my data and security managed with your solution?
Your data remains fully secure and never leaves your environment.
The API is deployed on-premise, meaning no information is ever sent to external servers: all data is stored and processed directly within your own infrastructure.
This ensures complete confidentiality, with no exposure to foreign legal frameworks, and enables full compliance with the GDPR, particularly regarding the control, storage, and use of sensitive data.
Is it easy to integrate into an existing application?
Yes, absolutely. Our APIs are built on a standardized format compatible with OpenAI-style APIs, which makes them very easy to integrate into most existing applications that already support API interfaces.
During deployment, we carry out validation tests to ensure the integration works correctly. We also support our clients with dedicated training sessions and comprehensive technical documentation to facilitate onboarding, understanding, and effective use of the API.
How is a custom AI API billed?
Like the MIA software, our custom AI APIs operate on a two-stage model.
First, an initial investment covers the entire setup phase: requirements scoping, infrastructure sizing, API development, installation, testing, validation, and training.
Second, a low recurring fee covers API usage as well as ongoing updates to the underlying artificial intelligence models. This model also includes unrestricted usage, with no token-based billing and no limits on the number of requests, ensuring smooth, predictable, and fully controlled usage of the solution.
Learn
Training & Conferences
We offer artificial intelligence training programs and conferences tailored to businesses, public institutions, and academic organizations. Our sessions cover both the fundamentals of AI, its challenges and limitations, as well as practical approaches such as prompting techniques and real-world integration into business functions. Available both in-person and remotely, our formats are designed to suit all levels and objectives — from AI awareness and foundational education to hands-on upskilling and operational proficiency.
What types of training programs do you offer?
We offer training programs ranging from a few hours to several days, tailored to the needs of companies and organizations.
Our sessions can be both conceptual and practical. On the conceptual side, we explain the origins of artificial intelligence, how it works, and its key challenges through a technically accessible approach. On the practical side, we cover prompting methodologies, real-world use cases, and best practices for using generative AI effectively in professional environments.
We also address the limitations of generative AI as well as regulatory topics, particularly the EU AI Act. As part of every MIA deployment, training is systematically included to help ensure compliance, especially with Article 4 of the European AI regulation.
Training sessions can be delivered either in-person or remotely, depending on your organization’s needs.
What is the difference between a training session and a conference?
A training session and a conference do not serve the same objectives and are not designed in the same format.
Training programs are longer and intended for smaller groups, usually around 10 to 12 participants, in order to encourage interaction, hands-on practice, and personalized support. They are highly participative and focused on concrete use cases and operational application.
Conferences, on the other hand, are shorter and more lecture-oriented. They can accommodate much larger audiences, ranging from a few dozen to more than a hundred participants depending on venue capacity. The topics covered are generally broader and more conceptual, focusing on the societal, organizational, and strategic challenges of artificial intelligence.
Do you offer training programs for beginners?
Yes, absolutely. Since generative artificial intelligence is still a relatively new technology, it requires the development of new skills, and a large proportion of users are still beginners.
We therefore offer training programs adapted to all levels: beginners, advanced users, and experts (including topics such as API usage or tokenization). However, most of the training sessions we deliver are aimed at beginner or intermediate audiences.
We know how to adapt the content to the group’s level, drawing in particular on our experience with MIA training programs, where participants are also mostly newcomers to AI at the beginning.
What group size do you recommend?
The ideal group size depends on the type of training and its objectives.
For more theoretical formats, groups can include up to 12 to 15 participants. However, for more practical sessions, we generally recommend groups of around 10 people in order to maintain a strong level of interaction and engagement.
When the training involves extensive hands-on practice — such as prompt engineering, AI workflows, or practical exercises — we recommend approximately 5 to 6 participants per instructor to ensure high-quality guidance and personalized support.
For conferences, there is no specific limit other than the capacity of the venue.
Are your training programs eligible for funding schemes (e.g. Qualiopi)?
Yes, absolutely. Training programs related to the deployment of MIA, as well as our general generative AI training courses, are eligible for funding through French skills operators (OPCO, formerly OPCA) under the Qualiopi certification framework.
We invite you to contact us to verify your eligibility and assess the level of financial support that may be available depending on your organization and situation.
How much do a training session or a conference cost?
The cost of a training session or conference depends on several factors, including the duration of the session, the number of participants, the location, and the format (on-site or remote).
Another key factor is the level of customization: whether it is a general training, an industry-specific program, or a fully tailored session based on your use cases, pricing will vary accordingly.
We invite you to contact us to receive a customized quote adapted to your specific needs.