#168 - FORVIS MAZARS : Comment passer à l’IA en RH sans oublier l’humain

·1 hr 9 min·5 clips

Mathilde almost quit HR after discovering it was a function 'not loved by others' and seen as a 'bullshit job'.

1. This episode of Yaniro's Human Factor podcast features Mathilde Lecoze, DRH at Forvis Mazars and president of Lab RH, on the topic of HR tech and AI applied to HR professions—not to the HR function itself. 2. Host Alexis Yves notes Mathilde is a rare guest who can speak as both a Lab RH president and a hands-on DRH practitioner at a mid-size firm competing against the Big Four. 3. Forvis Mazars France counts approximately 6,000 employees with an average age of 29, 1,500-2,000 recruits per year, and a 20% annual turnover—considered low for the audit and accounting industry. 4. The company has doubled in size since COVID and is growing rapidly, making scalable HR operations a survival requirement rather than an efficiency preference. 5. Mathilde arrived in HR from finance and consulting after a business school education, initially managing audits, and nearly quit HR two years in after discovering the function was associated with authority and isolation. 6. Her reframing: apply a marketing and client-service mindset to HR—define who your clients are (candidates, collaborators, managers), design experiences for them, measure satisfaction, and select tools based on client value, not HR team convenience. 7. Mathilde explains her distinction between integrated SIRH and best-of-breed: she deliberately chose best-of-breed because she couldn't find an integrated system that gave candidates the follow-up, CV sorting quality, and onboarding experience she needed. 8. Forvis Mazars now operates an IT architecture of 23 startup solutions, interconnected via People Connect as a Core RH layer; two-thirds of these tools focus on recruitment. 9. The foundational principle she states: AI requires clean data and data-driven processes—'if you don't know what you're doing, AI will let you go faster into the wall.' 10. Her process design philosophy has evolved: rather than designing processes and retroactively adding data collection, she now starts from required output data and designs processes backwards so that data quality is embedded in daily professional gestures from day one. 11. JIG, a recruitment scheduling automation tool, is cited as a concrete scaling example: it plugs into recruiter calendars, sends candidates an SMS, enables self-booking in three clicks, and pushes the appointment to the ATS—replacing the need for multiple HR assistants. 12. Forvis Mazars received 35,000 applications per year when JIG was introduced; it now receives 220,000 per year, handled by the same 15 recruiters. 13. The HR assistant who previously handled manual scheduling retired naturally; no positions were eliminated, and freed recruiter time now goes to interview redesign, candidate closing, and operator debriefs. 14. Forvis Mazars is recognized as a pioneer in employer branding on social networks, including by large agencies and CAC 40 companies—an outcome Mathilde attributes to constraint: with a small employer brand budget, they produced viral content ('a little trash, finally funny') that outperformed sponsored campaigns from larger competitors. 15. On generative AI, Mathilde distinguishes between individual efficiency gains (faster content writing, LinkedIn message benchmarking, business agreement drafting) and true HR process scaling—she considers generative AI still in the individual efficiency phase. 16. Her forward bet is on agentic AI ('IA Agentique') as the only way to implement a real Skills-Based Organization at scale: dynamically matching 6,000 employees' skills to roles, helping managers give feedback, setting objectives, and validating skills—all at scale. 17. She is in active discussions with providers including Maki on talent marketplace and agentic AI implementations. 18. Mathilde acknowledges a common fear that AI agents aren't human, but argues: 'Today these things aren't being done at all because we can't scale them. So it's better to do it than not to.'—while flagging the importance of bias controls and code review. 19. The episode's tone is highly practical and specific—Mathilde consistently provides exact numbers, named tools, and named providers rather than abstract principles. 20. Best suited to HR leaders and CHROs at mid-size companies facing rapid growth, recruitment volume challenges, or early AI strategy decisions; less useful for HR generalists in stable, large-enterprise environments.

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Yaniro - The Human Factor