Mark Angel and the Problem of Being Heard

The library kept turning up. Mark Angel’s family moved frequently when he was a child, and his parents worked hard. Public libraries became a place to spend time, with books providing some continuity when the address changed. Long before he was selling reading software to schools, he was a reader who had somewhere to go.

There is a pleasing complication in that beginning. Angel now leads Amira Learning, which puts an artificial intelligence reading tutor on a screen. His own early refuge was a public library. The distance between those two arrangements explains something about his work: he wants more children to have access to the attention that makes reading possible, even when the available adults have other demands on their time.

He has taken an unusually circuitous route to that ambition. Trading workstations, customer support software, speech recognition and school assessment all appear on his résumé. Each involved making information useful to somebody. At Amira, the somebody may still be learning to sound out the word.

A scholarship before the software

Angel studied economics at the University of Chicago. In 1978, he received a Truman Scholarship; the foundation’s register lists Indiana as his selection state. That detail belongs to a period well before the present arguments about artificial intelligence in classrooms. His career has roots in an older computing world, when getting useful information onto a professional’s desk was itself a substantial undertaking.

In 1987, he founded Papyrus Technology. Its software supported real-time market analysis for professional traders, including users at investment banks and brokerage firms. Ernst & Young acquired the business in 1996. The customers had money, expertise and urgent questions. Helping them interpret information was a business with an obvious commercial purpose.

The later move into literacy would change the customer and the stakes. A child learning to read cannot specify a software requirement, negotiate a contract or explain why a system has misunderstood a hesitation. Angel’s experience building for professionals makes the eventual change of audience particularly interesting. The person paying for the technology and the person speaking into it are now different people, with different ways of deciding whether it works.

The answer has to arrive in time

His next company, Kanisa, began in 1997. Angel served as chief executive until 1999 and continued as chief technology officer. Its territory was knowledge management and customer support: finding the information that would help someone resolve a problem. After Kanisa merged with ServiceWare in February 2005, he became chief technology officer of the combined business.

In a 2006 article about enterprise search, he argued for systems that could respond to different kinds of queries. The practical question was whether people could find an answer. Software buyers might debate the machinery; a customer with a broken product had more immediate interests.

That older work gives his education career a recognizable thread. Useful assistance requires more than possessing information. It has to identify the problem and offer something appropriate while the person still needs help. A search result delivered after the customer hangs up has limited charm. A reading hint delivered without regard to the child’s difficulty has a similar problem.

Angel’s subsequent roles included leadership at KANA Software and Nuance Communications. His professional biographies credit him with more than ten patents in natural language processing. These were years spent around software that had to interpret language, with all the inconvenience that actual human speakers bring to an otherwise tidy engineering plan.

The reader on the other side

Education became the setting at Renaissance Learning, where Angel served as chief technology officer and led research and development. The company’s reading products included Accelerated Reader and Star. He was working inside an established school technology business, close enough to see both the reach of classroom software and its limitations.

When he explained Amira’s beginnings in 2022, Angel described watching children read less and seeing shortcomings in the way schools assessed early reading. He wanted a different approach to the early elementary classroom. Individual tutoring offered a compelling model, but supplying a qualified adult to every child posed problems of cost, staffing and scheduling.

Amira, co-founded with Pete Jungwirth in 2018, brought that commercial experience together with research from Carnegie Mellon University. Jungwirth had also worked at Renaissance. The company was drawing on an existing body of work rather than announcing that reading instruction had begun with its incorporation papers.

At Carnegie Mellon, Jack Mostow and colleagues had spent years developing Project LISTEN, a tutor that used speech recognition to follow children reading aloud and respond with help. A 2003 demonstration described hundreds of children using it across nine schools. By then, the underlying idea had already traveled beyond a laboratory demonstration and into the complicated territory of regular school use.

“Reading is personal.”

Mark Angel, 2025

A small voice, a complicated assignment

Amira’s basic interaction is easy to picture: a student reads aloud, the software listens and offers targeted help. Behind that plain description are several separate judgments. The system must follow the reading, recognize where support might be useful and respond in a way that lets the child continue. Teachers receive information intended to guide their own instruction.

The attraction for Angel is the possibility of making that individual exchange available repeatedly. The child gets practice; the teacher gets another view of the child’s progress. His company now presents the product as a learning agent operating under teacher direction, connecting assessment, instruction and tutoring. The ambition extends beyond scoring a passage at the end.

Even the most elegant account of the technology must survive its audience. Children are wonderfully indifferent to product road maps. They need the experience to make sense. A teacher needs it to fit the school day. A school leader needs to know what the purchase accomplishes. The founder has to persuade all three, although only one group will attend the sales meeting.

Getting into the classroom

Recognition helped make the idea visible. Amira & the StoryCraft, a reading app aimed at families, appeared in TIME’s Best Inventions of 2021. Angel discussed it that December with Neil C. Hughes, alongside the company’s efforts to bring voice technology to reading practice. The home audience added another setting in which the product had to earn its place.

Distribution mattered too. Houghton Mifflin Harcourt became a partner, giving Amira a route through an established education publisher. In 2024, Stand for Children announced a partnership with Amira focused on early literacy. Angel’s project was accumulating relationships with people who already worked inside education’s institutions.

The larger organizational change came in June 2024, when Amira merged with Istation. The announcement described a combined reach of four million students across all 50 US states and 17 countries. Those were company-reported reach figures at the time of the merger, a measure of distribution rather than a finding about learning gains.

Angel remained chief executive. Istation’s Dick Collins and Amira chairman Amit Patel became co-chairmen, while Jungwirth continued as chief technology officer. The arrangement brought Amira’s tutoring work together with Istation’s content and school relationships. For a founder trying to move from a promising interaction to routine classroom use, those relationships were part of the product’s practical future.

Five million readers, and questions that remain

By August 2026, Angel said Amira worked with almost five million students worldwide. He was also answering questions from New Mexico officials about how the software operated and used student information. The growth had brought a different kind of attention. Parents and policymakers wanted explanations about a technology listening to children in public schools.

In his August 28 letter, Angel defended the company’s practices and said education technology should be accountable for educational outcomes and appropriate data use. His position was that Amira’s purpose was to support teachers and readers. The obligation to explain the system now accompanied the effort to expand it.

New Mexico had allowed alternative assessment arrangements that month following objections to the software. In September, San Francisco families and educators also raised concerns about their experiences with Amira. Some questioned the interaction itself; others questioned the evidence behind adoption. For Angel, these were challenges to the everyday experience his company was selling, as well as to its research claims.

His stated standard gives the debate a useful point of reference. He has argued that classroom technology requires rigorous independent testing and that changing a product creates an obligation to keep testing it. A research ancestor can establish a reason to investigate. A school buying today’s software still needs evidence about today’s software.

Another turn of the page

In March 2026, Boston University’s EVAL Collaborative selected Amira for its inaugural EdTech Evaluation Challenge. The selection opened a path to an independent randomized controlled trial. The university described a research team spanning computing, education and other disciplines, with attention to secure data handling and differences in outcomes across student groups.

An October account of the planned study set out three comparison groups: children using Amira, children receiving no supplemental instruction, and children receiving live tutoring from trained university undergraduates. That makes the question unusually concrete. The comparison includes a human alternative that schools might plausibly organize, as well as the absence of additional support.

Angel’s enthusiasm for the evaluation is consistent with his public emphasis on results. The findings will have to do the persuasive work. For now, the trial is a next step, and its selection is an opportunity to test a proposition rather than a completed verdict on it.

The child who spent time in libraries has become an executive asking schools to make room for a listening machine. His career supplies reasons to take the attempt seriously: decades of language technology, experience inside schools and a willingness to build partnerships. The classroom supplies the harder question. When a child closes the session and opens a book, what has become easier?