Compro Communications Inc. is a leader in the field of on-hold messaging design. Founded with the goal of transforming communication management within organizations, Compro offers a comprehensive range of services designed to enhance efficiency and connectivity. This preferred provider for corporate voicemail messages in Quebec creates voicemail messages for approximately 2,000 clients, from text drafting to voice recording.
While the production frequency varies from client to client, most clients update their messages according to seasons and special events, such as the holiday period. At each cycle, Compro inquires whether clients wish to record a new production, validates target audience information, collects ideas they would like to highlight in the new production, and drafts the associated scripts. The entire process is labour-intensive for the writers, especially during peak periods, which create significant pressure.
With our intervention, Compro can now contact its clients every month to check if they would like to record a new message. Artificial intelligence has accelerated two renewal processes:
Low-value activities, such as sending emails, following up with clients, and drafting notes, have been eliminated, allowing Compro to focus on its core expertise: writing and translating texts that convey its clients’ messages.
The company wanted to accelerate the production cycle by minimizing the need for human intervention while maintaining its high-quality standards. The solution enabled the sending of text change requests for a given period and the logging of numerous reminder notes in the ERP.
Semantic analysis is advancing at an impressive pace. The emergence of LLMs, such as ChatGPT, has transformed content generation practices. These innovative technologies have accelerated the development of these applications. In practice, adopting an approach that quickly delivers business value and short-term ROI is crucial.
Beyond updating technological maturity and managing change within the company, the challenge for Videns was to select and combine several text generation models (French, English, multilingual, pre-trained or not, supervised or not). It was necessary to account for the fact that the field of NLP (Natural Language Processing) was underdeveloped at the project’s start, and many issues still persist with multilingual “Transformers” models or their French versions. Thus, it was necessary to build a model capable of handling complex aspects of the French language, such as metaphors and words with dual meanings.
The automation of the most time-consuming tasks in the voice message production process has significantly reduced the time required for each project.
The integration of generative AI models allows us to focus on what truly matters by reducing low-value administrative tasks and improving the quality of the final texts.
Adopting LLMs has resulted in a quick return on investment and more efficient, precise message production.
In this project, Videns integrated generative AI models into the voicemail message production process to automate the most labour-intensive parts. Through meticulous process mapping upfront, we ensured all necessary behaviours to represent the complex process. This achievement aims to reduce pressure on writers, particularly during busy periods, and allows them to focus on creating content. It is a deployed project that is currently operational.
