McGraw Hill Expands AI Strategy Through Pilot Agentic Tool Programs

McGraw Hill is expanding its education technology footprint by integrating proprietary curriculum data into AI systems. The company currently supports over 7.5 million users across its digital platforms. While the company maintains a base of 100 million paid curriculum licenses, its newer agentic AI tools remain in pilot testing phases.

The firm is running these pilot programs with more than a dozen universities and one pharmaceutical company. These tools perform specific tasks within the educational workflow instead of functioning as simple conversational assistants.

"McGraw Hill’s approach is different from simply placing students in front of a chatbot. He argued that education content is heavily regulated and often controlled at the state and even zip-code level, so accuracy and local compliance matter as much as speed," said Philip Moyer, Chief Executive of McGraw Hill, during the Goldman Sachs Communacopia + Technology Conference.

Balancing Compliance and Innovation

The education sector presents challenges for AI deployment that differ from general-purpose enterprise software. Because curriculum standards are often set at the state or local district level, McGraw Hill must ensure its models adhere to specific regional compliance requirements.

This regulatory environment dictates the company's restricted approach to data. By using its own proprietary curriculum as the primary source material, the company aims to mitigate the hallucination risks that plague generic large language models.

Monetization and Pricing Models

As McGraw Hill transitions from standard digital content to AI-driven services, it is experimenting with several revenue models. According to a transcript from the Goldman Sachs conference, the company is evaluating:

• Per-question pricing: A usage-based model for specific student interactions.
• Token-based pricing: Charging based on the volume of data processed by the AI.
• Premium access upcharges: A 5% to 10% premium for customers requiring Model Context Protocol (MCP) access, which allows the AI to interface with external data sources.

These pricing strategies reflect an attempt to align the cost of compute-heavy AI operations with the value delivered to university and corporate partners. By testing these models within its current pilot programs, the company is determining which structure best supports the integration of AI into the classroom environment.