Simsoft Industry’s analysis of the CNIL white paper

CNIL White Paper on Voice Assistants

Industrial voice assistants and RGPD: a cross-eyed look with the CNIL

Faced with the increasing development of the uses of voice assistance technologies, the CNIL (National Commission for Information Technology and Freedoms) proposes an analysis of the use of these technologies in compliance with the RGPD (General Regulation on Data Protection) in the form of a White Paper. The CNI’s main argument is based on section 25 of the RGPD, which recalls “the importance of designing systems by building that respects the privacy of users” (CNIL, page 42). The link with voice assistants is related to the fact that ” the voice may contain a great deal of information about his identity or inferred characteristics such as emotional state, socio-cultural background, ethnicity, or health status. From then on, the voice reveals information that touches on the intimate. (CNIL, page 39).

The CNIL focuses on the evolution of the relationship of individuals with information, linked to the use of “mainstream” voice assistants, and proposes good practices regarding the implementation of these, with the RGPD as a common thread. SIMSOFT INDUSTRY provides an expert perspective on the industrial uses of intelligent voice assistants.

Today, the known uses of mainstream voice assistants are linked to the “simplification and fluidization” of certain everyday tasks. Nevertheless, new uses of voice support services are identified and of “interest in professional environments where it is difficult to handle computer tools and use written commands (e.g. manufacturing worker work)” (CNIL, page 20). This is the niche SIMSOFT INDUSTRY chooses: intelligent voice support services for industry technicians.

Relationship with information

In the professional and private worlds, the use of voice to access digital information changes the relationship with it. Indeed, “voice assistants propose a major paradigm shift by transforming the relationship to digital tools, through the transition from visual to vocal” (CNIL, page 37).

Removing the visual interface in the majority of home voice assistants eliminates the ability to access a large amount of data following a search, for example: the assistant will give 1 single answer to a question, where a visual search will provide a series of classified answers.

This removal of the visual interface also raises the question of “memorizing information and its appropriation by the user” (Mathieu Gallet, 2020). The visual presentation of the information, in parallel with its verbalization, contributes greatly to its assimilation.

Solution for industrial use

For SIMSOT INDUSTRY in a professional world, the recommended solution is to develop the multi-modality of user interfaces. A multimodal interface allows you to interact with your digital environment (with data in the broadest sense) in several ways: keyboard, smile, visual, voice, tactile, gestures, etc. To remain pragmatic and acceptable to a user in a professional setting, SIMSOFT INDUSTRY is working on the development of its intelligent voice support services to the technician (from smart voice skills) in multi-screen/voice mode: the screen is used as a touch and visual, the voice is used to speak and listen.

The multimodal interfaces proposed by SIMSOFT INDUSTRY respond to CNIL’s recommendation number 4: adapt to voice media (audio and visual) (CNIL, page 67). Visual interfaces are most often provided by the support software used in the industry (SAP, IBM-Maximo, 3DS Apriso, Infor, …) with which SIMSOFT INDUSTRY’s voice support services are integrated.

Finally, in the industrial world the information sought is most often standardized, validated or even certified. For example, when a quality inspector wants to access a torque of a mechanical element to validate compliance, he does not want a choice of answers, he wants the right value at the right time. SIMSOF INDUSTRY’s voice support services refer to the company’s repositories in which they are used. The information provided to the technician is therefore accurate and not ambiguous.

Personal giving, voice giving

The use of voice raises two main questions: the use of this data identified as personal by the RPGD, and the use of data from voice processing.

The majority of consumer assistants massively use learning algorithms to improve the quality of the services they offer. This learning can use the characteristics of the user’s voice allows “many systems to perform an automatic [de proposer] end-to-end transcript of speech”(CNIL, page 17). However, the use of this voice data, mostly on remote servers, is often unclear to the user.

The most common examples of current use of voice support services by the general public are mainly to make a request for information or immediate action: request of the weather (78%), up to listening to the information (64%) (CNIL, page 22). These uses do not involve the retention of the data generated vocally.

Uses that require persistent use of voice generated data come at the bottom of the list: set an alarm (62%),make a phone call (43%), or make an online purchase (34%) (CNIL, page 22). In these use cases, the voice generated data is “retained” by voice support applications at least for the duration of the requested action. One can mention the creation of events in an agenda, which requires the retention of information over an even longer time.

Solution for industrial use

The use of voice support services in an industrial context must meet industry constraints. Of these, the simplest are noise and safety. Noise is present everywhere on industrial sites, indoors and outdoors. Industrial data security is a constant concern for IT services.

An on-site operator, quality controller or maintenance technician will use voice support services to access relevant information (see previous point) or to trace information from their actions. This type of information may consist simply of completing a task, a measured value, or qualifying the state of a system. More complexly, the user can generate an intervention report with the help of his intelligent voice assistant.

SIMSOFT INDUSTRY offers this type of service, respecting industrial constraints, and rules related to the protection of personal data. To do this, all processes related to the data generated by voice are carried out in “embedded”. That is, the data processing algorithms are embedded in the technician’s PC, tablet or smartphone. No voice data passes through a local or global network. This point responds to CNIL’s recommendation number 2: to favour the local to the remote (CNIL, page 67).

Good practices for the industry

The use of intelligent voice support services is growing in our private uses. The accelerated digitalization of the industry in Europe will make it possible to penetrate this type of interface for technicians or workers, who are faced with the use of complicated software. The objective of “simplifying and fluidizing” production or maintenance tasks will be met.

In its report, the CNIL proposes 4 tracks in the form of “good reflexes” for the development and acceptability of voice support services. SIMSOFT INDUSTRY offers its industrial vision of these recommendations:

  • Maintaining desirable friction : rather than focusing on implementing an absolutely seamless user experience, take advantage of moments of friction (i.e. moments of choice, settings requiring the user’s attention) to present the reality of data processing to users in an appropriate way (CNIL, page 67)

SIMSOFT INDUSTRY offers its industrial customers a comprehensive “Spixify Your Industry” approach to get them to consider all aspects of the introduction of voice support services at the workplace. In addition to the points related to the acceptability of operators, this program addresses data management constraints and interfaces to existing tools (MES, GMAO, ERP).

  • Focus on the local to the remote: as much as possible, implement data processing modalities and capabilities directly within the devices, which gives the user a good control of them and is a factor of trust and acceptability (CNIL, page 67).

All SIMSOFT INDUSTRY intelligent voice support software is embedded on devices already used by operators and technicians: PC, tablet, smartphone. Voice data does not pass through local or remote networks.

  • Ensuring the means of control: allowing the user to understand and control the uses that are made of his data and to set the operation of the device according to his choices (CNIL, page 67).

Awareness and training of industry technicians and operators is part of SIMSOFT INDUSTRY’s “Spixify Your Industry” program. This period of taming voice support is a key point in the acceptability of voice support in a work environment.

  • Adapting to voice media : relying on exclusive audio interfaces poses significant challenges in presenting information to the user, collecting consent, or implementing controls. It is therefore necessary to reflect on the means to be deployed (CNIL, page 67).

The interfaces developed by SIMSOFT INDUSTRY are by nature multimodal: voice interface with voice support services (Spix.SKILLS), visual interfaces of support software (MES, GMAO, ERP). Specific thinking is taken when technicians do not have access to screens.



Press Contact: Laura PALACIN,, 05 31 61 85 10 / 06 26 84 55 58


SIMSOFT INDUSTRY develops the first Intelligent Vocal Assistant 100% dedicated to Industry 4.0 technicians. The