News and resources
Enhancing salary surveys with an AI-driven approach
This video was recorded during the ALL IN 2024 event in Montréal. François Marcoux, Director of Product Management at IVADO Labs, shared the stage with Étienne Boucher, Partner at Normandin Beaudry, to delve into the project they have been working on together over the last year.
Understanding Normandin Beaudry’s mission
In compensation management, staying ahead of market trends and ensuring competitive salary practices are essential for any forward-thinking organization. Normandin Beaudry has consistently prioritized providing clients with innovative solutions to design competitive rewards packages.
Normandin Beaudry, a leading Canadian consulting firm with nearly 400 employees, provides expert guidance to hundreds of organizations annually. Their core focus lies in optimizing total reward strategies, covering pension and savings plans, employee benefits, and compensation structures. Central to their advisory capabilities is a comprehensive database of salary surveys, meticulously compiled each year through data collected from numerous organizations.
In the past, participating companies submitted their employee salary information via an Excel-based survey. Normandin Beaudry experts would then validate and map these data points to industry benchmarks, a process that consumed thousands of hours annually. This manual job-matching task, while essential for ensuring data quality, presented a significant bottleneck.
The AI wave and the search for solutions
The call for change came five years ago. A client from a major tech firm expressed concern over losing top engineering talent to smaller competitors offering more competitive salaries. The incident highlighted a growing challenge: the emergence of AI-driven competition and the intensifying war for talent. As large organizations began acquiring smaller ones, the demand for accurate, real-time compensation data became increasingly critical.
This realization prompted Normandin Beaudry to explore AI as a transformative technology. Despite their initial foray into machine learning, where they tested classical machine learning techniques, the results for job-matching fell short. Their models struggled with new or evolving job titles, and their team still preferred manual checks to ensure accuracy. Additionally, the early attempts to leverage emerging language models were met with limitations, as these technologies had yet to reach maturity.
Partnering with IVADO Labs for a customized AI solution
A year ago, Normandin Beaudry began a new journey with our IVADO Labs team. This partnership, focused on flexibility and innovation, provided a customized solution to address the specific needs of this unique challenge.
At IVADO Labs, our development process was meticulously designed to maximize impact:
- Blueprinting: Assessing the feasibility and business case.
- Proof of concept: Testing initial models to ensure viability.
- MVP development: Building a Minimum Viable Product and refining it based on real user feedback.
- Production deployment: Fine-tuning and preparing the solution for full-scale use.
Leveraging tailored language models
The result of this collaboration is a sophisticated AI system that integrates multiple specialized language models, each trained to handle specific aspects of job mapping. By avoiding generic APIs like ChatGPT and instead fine-tuning models on proprietary data, IVADO Labs and Normandin Beaudry achieved far superior accuracy. This approach also combines these custom language models with traditional machine learning techniques to augment the solution and address common language models gaps (e.g., handling numerical data such as salaries).
A user-centric experience
The final product significantly enhances Normandin Beaudry’s workflow. Clients now upload data while the system handles automated matching and validation. An intuitive interface flags job titles that significantly deviate from established benchmarks—such as unexpected entries like ‘zookeeper’—enabling users to make swift, informed decisions on inclusion or exclusion. This innovation preserves data integrity and positions Normandin Beaudry to scale their survey capabilities efficiently.
The impact is clear: Normandin Beaudry expects to save thousands of hours annually and has laid the groundwork for future growth, such as conducting quarterly surveys rather than annual ones.