A teacher in Lyon spends their evening grading a stack of handwritten math tests. It is a universal ritual, measured in hours of personal time. A Paris-based startup named Ed.ai is betting that ritual is about to change with a practical tool for the person holding the red pen. Its AI is designed to read a student's handwritten steps, assign partial credit, explain errors, and summarize class-wide knowledge gaps [Ed.ai website] [FrenchWeb].
The Wedge: Handwriting and Partial Credit
Ed.ai's initial focus is high school math, French, and history-geography. The technical wedge is the combination of handwritten text recognition and multi-step reasoning. Assessing a handwritten algebra proof requires the AI to follow the chain of reasoning and allocate points accordingly [Ed.ai website]. The company claims its tool can cut grading time by as much as two-thirds [franceinfo]. For a pilot involving 4,000 students across 40 schools in the Lyon region, that translates to a significant return of personal hours [CentraleSupélec].
Pilots, Funding, and the Road to 100 Schools
The early traction in Lyon has provided a foundation for the company's €1.7 million seed round. The capital came from a cohort of French investors, including Ring Capital, AFI Ventures, Ventech, and the venture arm of CentraleSupélec [CentraleSupélec] [Ventech VC]. The team, reportedly 12 people with half dedicated to AI engineering, is now using the funding to scale its commercial efforts. The stated goal is to be in 100 schools by the 2025 rentrée and to begin an expansion into the United States [EdTech Actu].
The Incumbent to Beat
The most credible risk for any edtech tool is the status quo: the teacher's own pen, gradebook, and ingrained routine. Ed.ai's answer is to frame itself as an assistant that handles the repetitive lifting. The feedback it generates on common errors is meant to fuel more targeted classroom remediation, a concept the company calls "industrialiser la remédiation pédagogique" [FrenchWeb].
On paper, the unit economics of teacher time are compelling. If a teacher grading 30 papers spends 10 minutes on each, that's a 5-hour block. Cutting that to a third saves over 3 hours. Their bet is that with enough precision on partial credit and enough empathy for the teacher's workflow, it can.