AI-Based Learning Going Mainstream (Part 1)

AI-Based Learning Going Mainstream (Part 1)

By Sam S. Adkins

Metaari defines AI-based Learning as education and training technology that enables personalized learning via Natural Language Processing (NLP), Machine Learning, and Deep Learning. Conversational AI is a specialized form of Natural Language Processing (NLP). Conversational AIs adapt to individuals and engage in natural communication via text and speech with humans. They learn as they interact with users. Conversational AIs are the foundation of AI-based Learning.

Metaari published a new market report called “The 2019-2024 Global AI-based Learning Market: Learning in the Bright Air” in January 2019. According to the report, “The international AI-based Learning market reached $1.4 billion in 2019. The five-year compound annual growth rate (CAGR) is a robust 28.4% and revenues will surge to $3.8 billion by 2023.”

In our report, we identified over 600 AI-based Learning companies. The vast majority are startups that have launched in the last two years and new companies continue to come on the market at a rapid pace. The barriers-to-entry are now quite low since developers can connect to commercial cloud-based AI engines dramatically reducing the cost of development.

In this article, Metaari reveals the most interesting applications of AI in education, healthcare and corporate learning and the motivations for adoption.

The Grail Has Always Been Personalized Learning

Despite claims to the contrary, true personalized learning via technology was never possible before the advent of AI. One of the best-known cognitive computing platforms is IBM’s Watson and a growing number of developers are building AI-based Learning products on top of Watson’s cloud-based platform. Pearson, Apple, Blackboard, Scholastic, Sesame Street, Edmodo (now owned by China’s NetDragon) and Houghton Mifflin Harcourt are selling new educational products built on Watson.

Other AI-based Learning developers that integrate IBM Watson include Blupears, Cognitoys, ThoughtFocus, Tencent, and Circadence. They develop fundamentally different products. Blupears is an early childhood learning app. Cognitoys develops physical robots (smart toys) that teach kids to code. ThoughtFocus develops virtual teaching assistants for the higher education segment. Tencent develops tutoring bots for consumers and Circadence develops virtual assistants used in corporate cybersecurity training. There are now hundreds of suppliers developing learning products on Watson.

AI-based Learning developers are also using other AI platforms. Pearson launched their Longman Welcome to English curriculum in Asia in February 2018 on Microsoft’s AI. Cerego’s adaptive learning platform runs on Amazon’s AI. Montessorium’s popular apps for young kids integrate Google’s TensorFlow AI.

In February 2018, IBM’s Watson division announced that it was providing the “brain” for the AI-based robot called CIMON (Crew Interactive Mobile Companion) built by Airbus and the European Space Agency (ESL) for the International Space Station (ISS). The robot was launched into space in June 2018. “CIMON’s digital face, voice, and use of artificial intelligence make it a colleague to the crew members. This collegial working relationship facilitates how astronauts work through their prescribed checklists of experiments, now entering into a genuine dialogue with their interactive assistant.”


CIMON Courtesy of NASA and the European Space Station

China’s Liulishuo is the first AI-based Learning company to file for an IPO. The company went public on the New York Stock Exchange in September 2018. The company uses a conversational AI called Dong Ni Ying Yu (billed as “your personal English teacher”) to provide AI-based English language learning to subscribers. In late November 2018, the company reported that “Our net revenues grew by approximately 265% and our gross billings increased by approximately 181% year-over-year.” The company had over 80 million users across 20 countries by January 2019. Clearly, AI-based personalized learning appeals to users.

Massive Capital Flowing to AI-based Learning Companies

In our latest annual whitepaper on global edtech investment published in January 2018, we reported that “Over $2.90 billion was invested in 197 AI-based Learning companies in 2018.” This a dramatic increase over 2017 and the previous two years.

Just over $1.81 billion was invested to 124 AI-based Learning companies in 2017, a record at the time. Investor interest in AI-based Learning companies is very new. Barely $122.4 million was invested in this type of company in 2016 and only to seven companies. Only $2 million went to a single AI-based company in 2015 and there was no investment in this type of company prior to 2015. The highest round in 2018 went to China’s Squirrel AI Learning that raised $150 million in October 2018.

Yet, AI-based Learning is the one area where the US still has a commanding lead over China. The majority (61.5%) of all global investment in AI-based Learning companies in 2018 went to 102 US-based companies that raised a combined total of $1.78 billion. In stark contrast, $299.7 million went to just twelve Chinese AI-based Learning companies, a mere 10.3% of all global investments to AI-based Learning companies.

AI-based Learning Effective for People with Special Needs

AI-based methods have proven to be quite effective at modifying the behavior of people with special needs. Educational robots and avatars, both physical and virtual, are now relatively common methods of intervention. The IBM Foundation and the National Center for Learning Disabilities (NCLD) announced in August 2018 that they were collaborating on a platform called Teacher Advisor With Watson “a suite of classroom resources focused on supporting teachers who work with students with learning and attention issues.”

Luxemburg’s LuxAI sells a small robot called QTrobot (pronounced “cutey robot”) designed “to mitigate the social challenges faced by kids with autism.” LuxAI claims that their robot is the first scientifically validated AI-based autism intervention product.


QTRobot Courtesy of LuxAI

Brain Power’s Empowered Brain product “aims to empower children and adults all along the autism spectrum to teach themselves practical life skills and assess their progress numerically.” Empowered Brain “is based on neuroscience from MIT and Harvard and draws on the newest in technology including artificial intelligence, wearable computers, and augmented reality.”

AI Takes Root in the Academic Segments

There is a growing use of AI-based Learning products in the academic segments across the globe. The products used in the PreK-12 segment are now common in language learning and STEM. Virtual avatars are also being adopted in higher education institutions. They are used as virtual teaching assistants, wellbeing coaches, student advisors, and virtual tutors.

Pearson hired Milena Marinova, an AI expert from Intel, to run their new AI and Personalized Learning group in July 2018. This is the first dedicated AI-based Learning department formed by an educational publisher.

Finland’s Utelias sells an AI-based Learning robot called Elias. It is used in schools across Finland to teach foreign languages. “The robot recognizes the pupil’s skill levels and adjusts its questions accordingly. It also gives feedback to teachers about a student’s possible problems.”

The Auckland energy company Vector hired New Zealand’s Soul Machines to create the “artificial human” teacher called Will in September 2018. Soul Machines develops remarkably lifelike avatars for clients. The company describes Will as a “convincing and engaging digital teacher.” Vector provides the avatar to the schools to teach children about renewable energy.


Will Courtesy of Soul Machine

In May 2018, Tencent announced that they were working with China’s central Ministry of Education to develop the AI English Education Teaching Aid System, “which can support adaptive learning and personalized teaching through artificial intelligence.” This will be the world’s first national AI-based Learning deployment.

The Japanese government announced a program in August 2018 that will deploy AI-based robotic English language tutors in over 500 schools in 2019. The robots will be NAO robots purchased from SoftBank Robotics.

In September 2018, Amira Learning launched an AI-based reading tutor called Amira. “Amira is the first intelligent reading assistant able to listen to children read out-loud, assess their oral fluency, and provide AI-driven tutoring. Amira is based on technology licensed from Carnegie Mellon University (CMU). In studies done by CMU and DePaul University in public school classrooms, Amira produced reading growth comparable to giving each student a human tutor.”

Rapid Adoption of AI-based Learning in the Corporate Segment

Corporations are now using AI-based Learning products for customer education, personalized employee training, intelligent knowledge management, and performance support. According to Metaari’s recent market research, corporations across the globe are currently spending $584.9 million on AI-based Learning and will be spending $1.57 billion by 2023.

IPsoft’s Amelia virtual avatar “observes, learns, and remembers anything you ask. She can read emotions and context during conversations with colleagues and customers. If Amelia is unable to resolve an issue, she will escalate the task to a human colleague. Amelia will then observe how the colleague resolves the issue, learn from it, and apply that knowledge to any similar tasks.”


Amelia Reads Emotions, Courtesy of IPsoft

A startup called Worxogo has offices in Bengaluru (Bangalore) and Mumbai. “Mia, Worxogo’s personal cognitive AI engine, learns, predicts behavior, and coaches employees for improved business impact. It helps them improve their productivity through actionable advice using principles of neuroeconomics and behavioral science.”

New Zealand’s FaceMe markets their virtual avatars as digital employees. The avatars are extremely like-like and realistic. They market the product to a range of verticals including healthcare and education. “Solutions to pressing problems related to health and wellbeing, education, and other domains can be found by capitalizing on the unprecedented quantities of data and recent progress in emerging AI technologies.”


Courtesy of FaceMe

Inexpensive Conversational AI rapid authoring tools, chatbot templates, and so-called pretrained AI models are now on the market. AI-based Learning developers are now able to get products to market very quickly. A basic Conversational AI can be built in the Microsoft Bot Framework in a matter of minutes by simply uploading a FAQ file.


There is now rapid global adoption in all six buying segments across all seven regions tracked by Metaari. AI-based Learning is now going mainstream in developed economies and gaining traction in the developing economies, particularly in Southeast Asia. AI-based Learning truly is the answer to solving Bloom’s 2-signma problem. Look at the newest AI-based Learning applications for proof.

For more on this subject, view “How AI-Based Learning is Solving Bloom’s 2-Signma Problem” by Sam S. Adkins.


About the Author

Sam Adkins is Chief Researcher at Metaari. Metaari is a market research firm that identifies revenue opportunities for learning technology suppliers. Metaari principals have refined a sophisticated learning technology product categorization schema. Their research taxonomy is the backbone of their quantitative data repository.  Contact the author at

Part 2 of this 2 part article can be found HERE.

Leave a reply