Parental involvement in homework can significantly impact a child's academic success. But what if parents struggle with the material themselves? Researchers are exploring how AI, specifically large language models (LLMs), can provide real-time support to parents as they help their kids with math homework. Imagine an AI tutor not for the student, but for the parent! This new research focuses on building a system called the Caregiver Conversational Support Tool (CCST), which integrates with existing math tutoring software. As the student works through problems, the CCST analyzes their progress, including correct answers, errors, and hint usage. Then, using this real-time data, the AI suggests helpful messages for parents to send to their child through a chat interface. These aren't just generic encouragement messages; they're tailored to the specific problem the student is tackling, offering guidance on next steps, prompting the student to explain their thinking, and providing positive reinforcement. Early testing with middle school caregivers revealed interesting insights. Parents particularly valued content-specific guidance, especially when they were unsure about the material themselves. They also liked messages that encouraged their child to explain their thought process, providing valuable insight into their understanding. While the initial results are promising, there are challenges. Some parents found the AI-generated messages a bit lengthy and impersonal. Also, the speed of generating these messages needs improvement to keep up with the flow of the tutoring session. This research represents an exciting step towards creating more dynamic and supportive learning environments, leveraging AI not to replace human interaction, but to enhance it, empowering parents to play a more active and informed role in their children's education. Further development could see this type of AI support expand to other subjects, creating a powerful tool for at-home learning.