Imagine a student in a rural town exploring the dynamics of the solar system with a multilingual artificial intelligence tutor that feels more like a friend than a computer. This will be the experience of 10,000 students in Israel this May as eSelf.ai deploys a large-scale test AI agents in the guise of its digital tutors.
As AI tools integrate into education, they’re evolving from chatbots to conversational avatars with distinct personalities, transforming student engagement. This transition reflects a change in role and form. It suggests a psychological shift in how students are expected to interact with AI, which is becoming a digital companion accompanying them on their learning journey. Three notable examples–eSelf.ai’s digital tutors, Tavus’s Conversational Video Interface, and Khan Academy's Khanmigo–highlight this trend in agent-centered AI and its potential to transform student engagement with digital learning.
AI Agents: Digital Tutors Redefine Personalized Learning
At the forefront of this trend is eSelf.ai, an innovative AI company specializing in highly interactive digital avatars explicitly tailored for education. Unlike traditional chatbots, eSelf.ai’s avatars mimic human tutors, using real-time visuals, virtual whiteboards, and natural voice conversations, which results in a very similar experience to a typical online tutoring session.
A key feature of eSelf.ai is its natural voice communication capability. “Students text friends but prefer voice for tutoring—it’s more engaging,” says eSelf.ai CEO Alan Bekker. The avatars can engage students through fluid conversations, creating an intuitive and personal environment. The technology seamlessly integrates multilingual functionality, instantly adapting to students’ preferred languages for input and output, thereby breaking down linguistic barriers often impeding learning.
The AI tutors are designed with distinct, varied personas, resembling candidates from a casting call for young professional educators. This variety of personas enables educators and students to select avatars that best resonate with their personalities and learning styles, further enhancing student comfort and engagement.
MORE FOR YOU
‘Major’ U.S. Dollar Fed Warning Braces Bitcoin For A BlackRock ‘Megaforce’ Price Shock
Tesla, Looking To Boost Sales, Has A New Killer Model 3 Lease
Trump Approval Rating Tracker: Majority Don’t Approve Of Trump’s Handling Of Inflation, Border And Trade
AI Agents And ESelf.ai’s National Pilot Program In Israel
Marking a significant milestone, on April 22, 2025, eSelf.ai announced a groundbreaking collaboration with Israel's Center for Educational Technology to deploy its personalized AI tutoring system nationwide. Beginning in May, the pilot program will engage over 10,000 Israeli students, representing the first large-scale national implementation of AI-driven tutors globally. Harvard University will serve as the program’s academic advisor, providing guidance on educational frameworks and rigorously assessing its impact.
Israel’s tech-forward education system and teacher shortage make this project an ideal testing ground. This pilot aims to validate the effectiveness of personalized AI tutoring at scale and see whether it can produce the same significant improvement in student learning as a good personal tutor can. If successful, this pilot could significantly improve educational equity and access worldwide, offering personalized tutoring to millions.
AI Agents From Tavus Go Face-To-Face With Students
While eSelf.ai focuses on natural conversational speech interaction, speech is not the only way AI agents interact with students. Tavus’s Conversational Video Interface allows real-time interaction between students and digital agents, complete with voice and video capabilities.
Hassaan Raza, CEO of Tavus, describes the “human layer” AI agents need to realize their full potential as teachers. Agents need to be relatable and able to interact with students in a manner that shows not just subject-domain knowledge but empathy. A robust interface for these agents will include video, allowing students to look the AI in the eye.
Significantly, such technology facilitates highly realistic human-like behavior in spoken interactions and imbues the AI with the ability to observe the user and engage based on what is seen. It can recognize emotional cues from students and adjust its responses dynamically, further enhancing the authenticity and depth of interactions. If a student frowns, the avatar might say, “Let’s try another example,” adapting in real time. This type of ability can be essential when a student might be confused and unable to articulate their confusion. In such cases, a good tutor will observe the confusion and ask probing questions until it becomes clear what the confusion is or the lightbulb goes off and the student understands.
AI Agents And Student Companion
In contrast to the anthropomorphic avatar approaches of eSelf.ai and Tavus, Khan Academy’s Khanmigo uses a minimalist geometric figure with expressive eyes, like a friendly cartoon character. It can interact textually or via speech and successfully captures a unique personality that resonates deeply with students. Rather than providing straightforward answers, Khanmigo is designed to be a Socratic guide, steering students toward discovering solutions independently, fostering critical thinking, and deep engagement with the material.
Feedback from students has been overwhelmingly positive. Many describe interactions with Khanmigo as akin to working with an actual tutor. Khan Academy reports that students learn more effectively with Khanmigo, with student satisfaction notably higher than traditional learning methods. The personal tone and thoughtful interaction style help students feel supported and motivated, despite lacking a visual or auditory persona. Khanmigo’s thoughtful guidance makes it a faithful learning companion, not just a tool.
AI Agents In The Uncanny Valley
One challenge presented by systems embracing anthropomorphic agents is the phenomenon commonly called “the uncanny valley,” where near-human likeness causes discomfort. Moderate anthropomorphism, as seen in Khanmigo’s simple design, builds trust without unease. The challenge for developers is balancing the quality of the simulated tutors with the level of personality required to make the agent engaging. Being too human-like without making it to the other side of the valley can quickly shift perceptions negatively, undermining the effectiveness of otherwise advanced technological solutions. Thus, AI developers must carefully calibrate the realism of their avatars, balancing the benefits of relatability with the risks of evoking the uncanny valley.
AI Agents Are The Future Of AI In Education
By 2026, AI agents could be as standard in classrooms as textbooks. By offering features such as visually rich interactions, emotionally intelligent video conversations, and supportive textual dialogues, the tools move AI from disembodied knowledge to embodied personalities that approximate the types of learning interactions characteristic of the best teacher-student interactions.
As these tools evolve, developers must remain mindful of the psychological dynamics. Addressing challenges like the uncanny valley will be pivotal in ensuring AI agents become widely accepted and genuinely beneficial tools within education rather than just being the source material for blooper reels and memes.
