Subscribe now

LC: Before we get into tutoring and AI, tell me about DREAM. What’s your school model?

AFW: We currently support about 3,500 young people across three campuses with seven schools, an early childhood program, and an alumni support program in Harlem and the Bronx. About one quarter of our students have an IEP and we have a significant proportion of students who are multilingual learners.

One of our goals is that our graduates become adults that are attentive to their own physical and mental health. We also have a lot of extracurricular options. We started in 1991 as a Little League, so the legacy of extended day and extended year programming lives on. Nearly 70% of our students participate in our summer and after-school programming.

LC: Where does AI come into the picture for DREAM?

AFW: We started our journey with AI about three years ago. We joined a consortium called AI for Equity and built our own AI task force. We included members from all of our back office and central office teams, all of our schools, and began to experiment with a number of pilots. We’ve probably tested 30-40 different educator-facing and student-facing AI tools. In fall 2024, we were the first charter network in the country and the first school district in New York state to adopt ChatGPT Enterprise for all our staff.

LC: What are the most common ways your folks are using those AI tools?

AFW: The place we really started was around assessment and data because that is one of the most important ways in which we drive student achievement in our schools and also one of the most time-intensive activities for our educators. We co-designed a system for grading and scoring student written responses in ELA and math, which allowed us to roll out AI scoring for 3rd-12th grade on every type of assessment. Quizzes, unit tests, and interim assessments–we estimate this saves our teachers about 50 hours per year. Then we built tools through ChatGPT that take the data and analyze it, determine key gaps that students have, and help our leaders prepare for data meetings with their teachers by automating that creation process.

LC: Was there pushback from any of your educators?

AFW: Before we scaled anything, we spent a lot of time thinking about how to get our people invested in AI. We landed on a central message around equity that has been very powerful for our educators. We now have an AI vision statement, which is that by 2035, we want our students to exceed their more affluent peers, not only in using AI responsibly and effectively, but actually in building AI tools and businesses to allow them to accumulate intergenerational wealth and promote the societal good. It’s a really forward-thinking vision that goes beyond what a lot of schools are thinking about right now.

Every one of our educators is deeply passionate about making our kids succeed long term. So that message of centering student success first and us, as adults, having to become truly expert in these tools to allow our kids to thrive has been something I think every educator, no matter how resistant, has bought into.

LC: What does student usage of AI look like?

AFW: First of all, we are not using AI in elementary school in any student-facing way. We believe to use AI really effectively, you really do have to build the foundational skills offline in elementary.

In middle and high school, we have been piloting ways to teach AI literacy. For example, in 9th-grade, kids were learning about ancient empires in their world history course and studied those offline, no AI tools, for a few days. Then they actually got a paper-based lesson where they got AI-generated claims about those ancient empires, and they had to evaluate, “are the claims accurate? Are they biased? Is this relevant to what we were researching?” Only after that did they go on to an LLM that we use, which is called Magic School, and actually start to create a cross-comparison essay of different primary and secondary sources about those empires, and writing with AI, but using a lot of the skills around evaluation and the knowledge they previously built to make that AI-generated output really strong.

LC: Are parents on board?

AFW: We have not seen the tech-lash in our community the way I know some others have. We did a lot of work in the beginning to keep our families in the loop as we rolled out AI initiatives, and tried to get them to buy into the vision of getting kids to succeed in this new economy. And we try not to use AI in class all the time. Keep screen time to a minimum. Learning about AI involves a lot of competencies and habits that have really nothing to do with the tech tools themselves, but how kids are thinking about approaching those tools.

LC: What is it you’re going to explore under the Accelerate grant?

AFW: We are building our own AI tutoring assistant. Before AI, we started building a data pipeline and warehouse to keep all the data in one place. It lets us see about 65 different data points for every single kid in our network, dynamically updated throughout the year.

Now we are asking, “What are kids learning in the next few weeks, and how do we use tutoring to get ahead of that and give kids prerequisite skills and access to grade-level instruction?”

Imagine a 7th-grader who is about to learn percentages next week. They do not have a deep understanding of equivalent fractions from fourth grade, and therefore will never be able to solve a percentage problem because they can’t cross-multiply and understand fractions. Our AI tutoring assistant takes the knowledge that we have from their i-Ready data, and comes up with a standards-based tutoring plan for them for the week of what they need to learn.

We also personalize the materials for every kid. Even though everyone’s learning 4th-grade equivalent fractions, Jinad is going to be learning that with a set of problems about basketball because we know from his extracurricular data that Jinad is a top-tier basketball player and loves basketball. The student right next to Jinad, Taisha, is going to be learning about something that she’s interested in, penguins. But Taisha also has an IEP, and it says that she has an oral communication impairment and needs sentence stems to be able to really describe her thinking. And so at the top of the worksheet are a series of sentence stems to support her.

AI is not actually teaching the students. It’s not giving feedback to the students. It’s just generating the materials and then telling the educator on the back end how to support that student. It’s an educator-facing tool that creates materials for kids, but still has the human as the primary deliverer of content because rapport and relationship is probably 80% of what allows a student with low motivation and low skill to succeed. That’s been borne out by so many AI pilots that we’ve done on the student-facing side, where we’ve given an AI tutor to kids and they’ve shown no growth, especially for our low-motivation, low-skill kids, because they have no relationship to this faceless bot. They’re just clicking through, not learning anything. On the flip side, we know that to differentiate and personalize to that level for kids is something that no human has the time to do. By combining the human relationship with the AI ability to analyze data quickly and create materials, we feel like we’ve hit the sweet spot.

LC: What does your research design look like?

AFW: We’re partnering with Mathematica to look at the bottom 50% of our 3rd-10th grade students in math. Half of those students will get tutoring starting in September. The other half will get tutoring starting in January. We’ll see whether the tutoring program works by comparing students who get the full year compared to half the year. Then we will also have some students who work with tutors using the AI assistant, and others who work with tutors who create their own materials. There we will ask, does the AI tutoring assistant have an impact?

LC: What is the future of tutoring?

AFW: I think the better question is, “when is the future of AI tutoring?” I think we’re a long way off, but I think there’s a potential to get there. Right now AI tutors are trained on data sets of how we get the right answer. AI is pattern-recognizing the shortest path to correctness, but that’s not how we learn. We actually learn through struggle, through transferability, through lots of practice and metacognition of that practice. We have to retrain the AI on that, which is totally different. We need to sit and watch the best human tutors with the most struggling kids, understand the non-verbal cues they’re giving, and train the AI in a way to better understand how kids, especially kids furthest from opportunity, are learning and mastering material. I think that’s a years-long project.

Subscribe now

Subscribe now

LC: What is Catch Up and Read?

CG: We’re a non-profit that was founded in Dallas, Texas 18 years ago by a woman who was volunteering in a school. She realized the children need more help than a volunteer can provide, so she asked, “What are other systemic supports that we can help schools with?”

She developed our model that grows both students and teachers. We pay stipends to teachers who stay after-school for professional development and tutor students. We started in just one school, and now we are in 25 schools across three districts working in grades 1-3. Students attend two days a week, for two hours after school for teacher-led tutoring.

LC: How would you describe the opportunity and the challenge of working in the after-school space?

CG: The opportunity is that there are more minutes in the day, more time on task, more explicit time to work on their reading goals. It’s also a time that supports working parents, gives them more child coverage.

Attendance is always a challenge after school. Both inconsistent attendance and just availability. Some families aren’t available after school and some teachers aren’t available to stay after school.

LC: How does a student end up in your program?

CG: Usually we partner with a district leader who sets an overall guideline at the district level about who they want us to work with. We want to be in alignment and in concert with the district and the school. Typically it’s based on beginning-of-year data. We partner with the principal to go through the student data, and select students more than a year below grade level. The school then invites the children to participate.

LC: How does the teacher recruitment and training side work?

AJ: The first priority is to identify a lead teacher. This is the person who can facilitate if our literacy coaching staff aren’t on campus that day, the one ensuring the program runs smoothly and that the professional development has taken place. Then we move on to building cohorts of teachers, preferably in those lower grades, PK-2.

CG: We’re growing both teachers and students in our program. Our teacher cohort is usually a mixture of veteran teachers and newer teachers, and they learn as a cohort and take best practices back to the classroom. We’ve seen from the research Johns Hopkins has done on our program that not only do the tutored students outperform non-tutored students, but the students in the classroom of a Catch Up and Read trained teacher outperform other students.

LC: What do you think makes your training so good?

AJ: I’ve been on both sides. I’ve been a lead teacher and now I work at Catch Up & Read. It’s not your typical tutoring program. Catch Up & Read works through campus teachers that students see every day or have as a homeroom teacher. It’s a relationship that’s already cultivated, and they are able to be part of the learning together. Our coaching team also provides in-the-moment feedback that teachers don’t always get. And we get to see their reactions in real time, see improvements made as teachers transform their practices. It’s so powerful when you’re tapping into the skill sets of teachers that are already there, and you’re just enhancing them.

We also provide all materials to the teachers. There’s no, “I have to think about what I’m going to talk about, I have to prep.” We have them role play and rehearse before they get into the tutoring with the students, all the way down to the type of turn-and-talk you’re going to do with students.

CG: Someone said to me, “Oh, you’re the driver’s ed of teaching. You teach the lesson, you role model, and then you coach directly while they’re tutoring.” That’s why it’s an active practicum. Like we’re in the car with them doing that embedded coaching right after a lesson. We’re also there every week for the whole school year. It’s a lot of time.

LC: How do you develop your materials? Or do you have a fairly set curriculum at this point?

AJ: While we have a core set of materials, they are definitely being continuously upgraded because education is ongoing. We have to keep up with innovation and with research. My second year on board, we sat down with one of the writers of the Reading Academies that teachers in the state of Texas have to take. She gave us feedback on our lesson plans, which we incorporated.

LC: Tell me about a school you’ve been working with for a long time.

CG: Marcus may be our longest school, we’ve been there for ten years and it’s been a great relationship. They have a very tight teacher cohort but they have new teachers every year as well. We continue to see growth with that campus, which is exciting. It’s a heavily bilingual campus. We run our program fully in Spanish as well as in English. They’ve even piloted using our program for some of the high Spanish readers to move over to English.

LC: How does grouping work with the students in the program?

CG: We serve 48 students at each campus with six teachers. Those six teachers do a whole group session with phonics and a read-aloud with eight students. Then they divide into two small groups of four. The grouping is based on data, not on grade level. A first-grader could be with a third-grader. It’s all about developmental need identified with the beginning-of-year data.

LC: I want to talk about impact now. What are the learnings from the external research on Catch Up and Read that have influenced how you continue to evolve the program?

MB: We’ve experimented with different dosage levels, especially for our in-school tutoring model. And then we’ve looked a lot at parent participation. Some schools have better parent participation than others–why is that? And then we began tracking very tightly to figure out what are the different ways we need to work within these schools versus these other schools.

CG: Catch Up and Read has believed in the value of external research from the very beginning. Our research partners at Johns Hopkins have done quasi-experimental studies on both our afterschool model and our in-school tutoring program. We thought maybe we’d want to lean into the in-school tutoring, but the pendulum is really swinging back to after school. I’m glad we have an 18-year history with afterschool!

LC: What’s the hardest part about making a program like yours successful?

CG: The engagement and attendance after school. The principal has to really be on board, and reinforce how important this is, that it’s a special opportunity that teachers get to earn extra income right there on campus and that the school is investing in them as well. And the same with families. We invest a lot in attendance calls, trying to let the parents know how important attendance is in our program.

But when we do get the attendance it pays off! Another great finding from our last research study was a statistically significant difference in attendance during the school day. Children who come after school to Catch Up & Read have higher attendance during the school day compared to children not in our program, by two full days more than the comparison group.

LC: Can you say more about how you engage with families?

AJ: One engagement we did was walk parents through all the different reports that they would see throughout the year, whether it’s the [NWEA] MAP report, the [Amplify] mCLASS report, other reports they’re going to see that they may not know exactly what to do with. We walk parents step-by-step on how to read the report, what each of the color bands means, where their student actually is, where we’re trying to grow them and get them to achieve.

Another thing we did was invite parents into the classrooms to do a gallery walk of the read-aloud. We introduce them to our system and how we leave feedback for teachers. We asked them to focus on what they want to learn. The parents were able to sit in their students’ classrooms with their teacher and got to learn how to do a great read-aloud with their students. They also wrote teachers feedback notes. So at the end, the teachers got high praise from parents but they were also able to see exactly what parents are thinking about as they see this process with their child. It was really a full circle moment. All of those parents walked away with a copy of a read-aloud book to use with their child, that provides them all the places to pause, with questions to ask to get the child to engage. We wanted them to learn a new skill they could use at home with their child every day.

CG: We also meet with the principal once a month to review progress monitoring data. It’s another overlay of our program and the whole school. We’re sharing great news about what’s happening with Catch Up and Read teachers and students, and they’re shouting it out in their school newsletter. Our program becomes embedded in the school culture.

LC: What is the future of tutoring?

CG: The future of tutoring is bright. Students will always need extra personalized support. And it’s important that it’s aligned to great work already happening in the classroom.

MB: The future of tutoring has great potential when it’s human-centered and data-driven.

AJ: The future of tutoring is hands-on. I don’t care how much you have AI, the robot, the computer, this is truly hands-on. And if you are not hands-on with that child, that teacher, you’re going to lose them.

Subscribe now

Subscribe now

LC: Why did you build Royo?

YB: Most of my background has been as a classroom teacher. I started teaching kindergarten, went on to second grade, then got my master’s in education policy and was able to see all the research behind the challenges I was experiencing as a teacher.

Then I went to Success Academy, the highest-performing charter network in New York, really looking at the strategies we’re using to get 20,000 kids reading on grade-level every year. I went back into schools, teaching at a Success school in Harlem, and got to experience those strategies in person as a teacher, seeing the way Success teaches reading in a scalable way. All of this inspired Royo. I wanted to explore how we can use technology to fill some of the gaps for teachers that might not have access to all the best resources. And then I serendipitously met Lara along the way.

LS: I started as a third- and fourth-grade teacher, and then became a literacy coach. Then my nephew turned out to be dyslexic and I wanted to be able to support him as a literacy coach, but I couldn’t. I went to my first Orton-Gillingham training, which opened the whole world of structured literacy for me and I started my own tutoring company. I met Yasmin through one of the families I was tutoring, and here I am with Royo, which is a supplemental product for teaching and tutoring. I use it every day in my own tutoring practice.

TL: I came in more recently, and have been more engaged in the traditional high-impact tutoring world. I’m here bridging the gap between high-impact virtual tutoring and the work Royo is doing. We need more states doing what Arkansas, Louisiana, and New Jersey are doing, funding HIT at the state level, to help get tutoring to more students because it definitely makes a difference. The challenge right now is budget for most districts, and how to scale this expensive but effective intervention scale. We don’t want to replace the humans, so that’s where Royo comes in.

LC: What is Royo?

YB: Royo is a supplemental literacy platform where we bridge the science of reading practice and what we call expression-led engagement. We want students to see their future possible selves in foundational literacy skills. We want them both learning how to read and understanding what they’re reading. Another goal is to supplement classroom instruction, based on what the teacher is teaching and what the students already know.

LC: How does it fit into a typical instructional day?

LS: A teacher teaches a phonics lesson, it could be in a small group or one-on-one or whole class. After teaching the skill, the students are able to read those words or practice those skills in the context of a story. We don’t want them guessing words. Reading is often invisible. We’re trying to make the process of reading visible to teachers. The voice recognition technology helps students as they’re reading. If they make an error, first a blue highlight shows the student the word they had an error on. If they make an error a second time, the word will be split up so they can sound it out. If they still have difficulty, they can press a button to hear the word. It’s like having a teacher’s assistant with the child, and all that information goes to a dashboard where the teacher can see how each child reads.

LC: What is the teacher doing while kids are on Royo?

LS: Sometimes the teacher is pulling a small group while other students are on Royo. Sometimes the whole class is on it following a lesson. We also see it being used at home so kids are practicing the skills they’ve been working on in class.

TL: One question I get a lot is, “if we’re doing thirty-minute sessions of high-impact tutoring in a 45-minute window, what can the kids do for 15 minutes after tutoring?” Or, “what can kids do to practice on their own?” Not everybody has a reliable adult who can sit next to them to read and practice at home. So 10-15 minutes a day, a couple days a week is a small time commitment, especially compared to 90 minutes a week of tutoring.

LC: With your teacher or tutor hat on, what do you like about Royo?

LS: I used to write my own personalized decodables. I would take the skills my students have learned and write a story for every single child. It would take about 45 minutes each. Now I get the same effect, and there’s such deep knowledge of structured literacy in this. There are pre-read skills. It includes vocabulary up front. It could be rhyming, like phonemic awareness activities to get them warmed up. We have nonsense words. All these things I do in my practice, these kids are now getting without a person in front of them.

LC: Is it best if it’s used every day? Is there an optimal dosage?

LS: I don’t think this should be used every day. Three times a week for ten minutes is a healthy amount. We really believe in a healthy diet of different types of books and we don’t want students only reading decodables. I’m also not someone who is for screens so much of the time.

LC: I was going to ask about this. One thing we talk about in the tutoring word that applies here is the idea that kids need more time on task practicing reading. So I love how this solves a really clear problem. But also maybe it’s a terrible time to launch a program that is a screen-based tech solution for early literacy. How do you think about that right now?

LS: I think about this all the time. With my own kids, I’m very strict about screen time. But I think that’s why we are so teacher-forward. I believe that intentional short bursts of reading practice are so important. This targets exactly what each child needs. You can also print the books.

TL: We really try to focus on things that teachers cannot do. Teachers see the benefit in a tool that helps them do what they can’t on their own, like giving every student immediate feedback on reading. Or if a kid is obsessed with dinosaurs, find him books about dinosaurs, now imagine being able to do that in real time for every kid in the class.

LS: If you look up “decodable texts” on Amazon, you will get a lot of texts that aren’t really decodable. As a literacy specialist, I find this really frustrating. I have very few books that I can recommend that are out to the public. Royo solves this.

LC: What’s the role of parents? Do parents get data or information from Royo?

LS: We have a parent dashboard that is a really important component to us. The parent can see exactly how their child reads each book, which words they had trouble with, and which skill they’re working on. It gives parents tips on how to work with their kids at home. The parent-teacher connection is so important but sometimes parents come to a parent-teacher conference and they’re completely shocked by what they’re hearing. This is a way to open the door for more communication and make sure parents can be as involved as they want to be.

LC: What’s the research you’re going to explore through the Accelerate grant program?

TL: We know high-impact tutoring works, and we know virtual high-impact tutoring is an excellent option, too. We also know that we typically want about 90 minutes of tutoring a week per student, for 12 to 24 weeks. That’s a lot of time! So the idea is, can we get students on a tool that is much cheaper for 30-45 minutes a week. It’s a third of the time and pennies to the dollar on what high-impact tutoring costs. We want to understand how much students can gain or learn with this approach. We also want to focus on multilingual learners, students with different learning needs, just a broad representation of students.

Subscribe now

For one glorious day in May, for each of the past four years, I’ve had the privilege to sit in the Arrillaga Alumni Center, listening and learning from the folks who built and continue to power the high-impact tutoring movement that now serves hundreds of thousands of American schoolchildren. Yes, I am biased. I love this conference.

So I am not even going to pretend to be cool about watching Steph Curry say “high-impact tutoring can help set [students] on a path for lifetime learning” in the kick-off session of the 5th annual National Student Support Accelerator conference at Stanford this week. Was it a pre-recorded video? Yes. Was it still amazing to hear a world-famous celebrity say the phrase “high-impact tutoring” as if it’s a basic idea that everyone understands? Absolutely.

Enough rhapsodizing, here are four takeaways from the day:

1. LAUSD is excited about how tutoring can continue helping students catch up, and get ahead. Dr. Karla Estrada from LAUSD says they are providing high-impact tutoring to 75,000 students, and those students are getting twice the growth of similar students who are not in tutoring programs. Particularly noteworthy is that Estrada says the district is beginning to think about how to use tutoring to allow for more acceleration. Acknowledging that there are many kids who are ready to go further and faster, Estrada said this will be a big focus of LAUSD moving forward. Will they pull it off? How will they pay for it? Those questions weren’t answered but it’s encouraging to see districts, especially large districts who catch the media’s eye, starting to ask what else can be accomplished for kids with tutoring.

2. The global research community has entered the chat. Two of thirteen research presentations were on international studies, both with potential relevance to US context. One study, “Investing in Human Capital During Wartime: Experimental Evidence of Online Tutoring in Ukraine” suggests that relatively low-cost tutoring programs can make a real difference for students when school is disrupted. I don’t anticipate Russia shelling the US anytime soon, but the next pandemic or localized school closures due to weather/climate change is something that we’d do well to prepare for. Even more exciting was a study from Germany on how tutoring by college students for struggling students in grades 8-10 made an impact in labor market outcomes for the students who got tutored. Now, before you come at me with all the reasons that Germany is different from the US, I get it. What’s interesting is that when they provided relatively low-cost tutoring to struggling students outside of school, the students who chose to participate were more likely to be lower-income and were then more likely to really benefit. Maybe creating more high-quality tutoring programs for motivated secondary students who want to opt-in is something to consider.

3. Humans remain at the center. While AI was part of the conversation, it was secondary to the human tutoring taking place right now. Three of the thirteen research presentations involved AI tools, either student-facing or tutor-facing. I don’t think anyone has seen an AI tutor that has completely figured out how to get student motivation where it needs to be for the tutoring to really work. (If you have an AI tutor that does this, please let’s chat!) More broadly, the title of the conference was “Human Connections in a Digital Age: The Next Chapter of High-Impact Tutoring.” Most of the conversations focused on humans involved in implementation or research on human-led tutoring. I did take my first Waymo while in Palo Alto, but I did it with another human so we could keep talking about tutoring! Human connections in a digital age.

4. There’s a lot of tutoring data available to folks who want to use it. Between the National Tutoring Observatory (they say they are building the world’s largest tutoring dataset! and they’ve launched a new tool to help providers get better insights from their tutoring transcripts) and Pearl (where my friend Nate Casey is always dreaming up how leaders can better use data to improve tutoring efficacy) and others, the high-impact tutoring movement remains driven by data and outcomes. As Susanna Loeb said in her intro remarks, the field has consistently raised the bar for quality tutoring. Based on what I heard this week, that work absolutely continues.

Until next year, NSSA. What will we have learned and accomplished by then?!

Subscribe now

LC: Tell me about Learner.com.

MB: Learner.com has been around for a little more than five years. My background is mathematics. I taught math at the university level, and I built an executive education business, and I always wanted to build a consumer business. So in the middle of a pandemic, I started a new business. We started with five tutors and now we have more than 400 tutors.

LC: Is there a profile of a typical tutor that works for Learner.com?

MB: There’s not a specific profile but there are a few things we do that are very different. I probably hired the first 100 tutors myself, just entirely focused on how they have to meet the student where they are, and build the student up to make them understand the material. Parents in the world we operate in have high expectations. If you’re paying $60-$90/hour, you expect certain things.

First, you have to know the American school system. If you interview with us and you use the word “maths”, you’re out. You won’t think like our parents, our customers. The second thing is no college students. You have to be out of school and to have done a variety of things. A third thing is that you must have experience managing parents. Parents are so involved these days, they have questions, they want to know things. If you haven’t had to deal with parents, you’re not going to succeed. We get candidates who say, “I just want to focus on the student.” But no, that’s not going to work. You bring the parent along. You have to write updates to make sure they understand what’s happening with their child. You can be the best teacher in the world but if you don’t have experience directly managing parents, this isn’t going to work for you.

LC: Did you have a profile of a student in mind when you thought about who you wanted to serve?

MB: I thought it was going to be mostly high school, but it’s all ages. Middle school is common, down to kindergarten. I have three kids that are out of college so I know the struggles they went through. I realized that at times, outside help is needed. When you’re arguing about math with your kid, and it sounds a lot like the arguing you have at dinner, it’s not productive. And there’s a whole world out there of people in need, and a lot of pain around academics.

LC: Tell me more about how you select tutors that can meet these high standards.

MB: Every tutor has to go through this one role-playing exercise. It’s an SAT question and the way the question works, it could be answered in a variety of different ways. No matter which way the tutor starts the question, we stop and say, “I’m having a really hard time understanding this. Can you please explain it in another way?”

And if you, the potential tutor, says, “No, Mike, this is the right way to do it and you learn it this way”, that’s a red flag. You’re out. In today’s world, you need to meet the students where they are and build them up from there.

We also treat our tutors with a tremendous amount of respect. We promote from tutors to full-time employees. We have regular meetings for tutors to share experiences. We have communication tools, Discord groups, for tutors to connect and get involved. Anytime we make big changes, we bring people together. We’re working on something now called Next Level Math Summer. It’s important that the tutors can be part of these new initiatives. If they’re not feeling loved, why are they going to go above and beyond for students?

LC: How do you evaluate the strength of the tutor-student relationship?

MB: There are two quantitative pieces we look at. Our model is a trial model. You come to our site, you have to answer 25 questions. There’s an intake form that some people might think is crazy but we think it’s so important because the tutor answers the same questions, and it helps with our proprietary match process. So then the trial itself is a 60-minute session for $25. By the way, we used to offer it for free but people don’t show up as often when it’s free.

We have data on how many people who take the trial then sign up, and the average is 70%. But for some individual tutors, it’s 90% and some are 40%. We put the databases of different tutors into an LLM and ask, “what are the common themes between this tutor and this other tutor?” So we’re learning what works across tutors. We also track renewals by tutor, month to month.

LC: Are parents hiring you for their kids because their kids are behind or because they’re looking for acceleration?

MB: We have two groups of students. We call them Academic Olympians and Drowning Students. The Academic Olympians are about 20% of our audience. Those parents come in with, “We’re thinking about next year. We want to create a whole structure.” The drowning students often show up right now, in the spring time. “My child has a D, needs to get a B, there’s a month left of school.” Sometimes I call this urgent care. People are coming in and they have unbelievable pain or injury that needs to be solved right now.

LC: Can you help those drowning students who show up at the end of the year asking for help?

MB: This is urgent care. We have to just start and get as far as we can. But then we also suggest, “hey, we have a summer program called Next Level Math. We can help your kid once a week over the summer and get them ready to succeed in next year’s math class.”

LC: This is not a typical question I ask, but because you are focused on math and come from teaching math, I’m struck by a statistic in the 50CAN Educational Opportunity Survey that of kids in tutoring, 80% are in tutoring for math. It’s by far the most popular subject for tutoring in the survey. I wonder why this is – are we just bad at teaching math? Is there something about math in particular that needs more attention and support than a classroom teacher can give?

MB: Part of it is self-inflicted wounds. People tell themselves over and over again that they’re not good at math. Then it becomes self-fulfilling. When I taught at Penn State, I can’t tell you how many people came to office hours and used those words, “I’m just not good at math.” That’s not how it works. Someone has told you over and over again, or you started telling yourself that, and it stuck. Early on, what’s important is getting wins and building confidence. Tutoring is a big part of that support. Lots of students are embarrassed to speak up in class, especially if they’re not confident in math. But then if you have a couple wins, all of a sudden you raise your hand to a question, and it spirals in a positive way.

LC: Changing topic! One way that parents in some states are paying for tutoring on Learner.com is by using Education Savings Accounts. Tell me about your journey into the world of ESAs.

MB: We’re approved in 15 states to provide state-funded tutoring. We’ve also built a tool called the Education Funding Finder, based on something the folks at EdChoice had started a couple years ago. You put in your state, it asks you about 12 different questions, and it tells you what’s available. It’s very clear, either “you’re not qualified” or “click here to apply or get more information about the program you qualify for.” We’re seeing there’s a real need for this kind of help, and we want to help lots of different families.

LC: Are the families using ESA dollars to pay for your tutoring any different than your other customers? Are the kids younger or older? More behind? Do they tend to have fewer tutoring sessions or more tutoring sessions?

MB: Students using ESA funding look meaningfully different from our traditional self-pay clients. They tend to skew younger, and over 90% are homeschooled, compared to less than 10% on the self-pay side.

We also see a much higher prevalence of students needing additional support - more than half have some form of learning or developmental need. The most common include ADHD, dyslexia, anxiety, and autism.

LC: How do you think about growth and scaling? What does the future hold for Learner.com?

MB: For us, growth is about preserving quality while increasing access. The biggest risk in tutoring is that it doesn’t scale well if quality breaks - one bad tutor experience can undo everything. So a lot of our focus is on improving the matching process and learning what makes a tutor-student relationship work.

We’re using data from thousands of sessions to understand patterns - which tutors drive engagement, which lead to renewals, and why. That helps us get better over time, not just bigger.

We’re also thinking a lot about distribution. Historically, tutoring companies have relied heavily on direct-to-consumer acquisition, which can be expensive. Programs like ESAs and partnerships open up new ways to reach families who already have intent but need guidance.

LC: What is the future of tutoring?

MB: I think tutoring is moving toward a more continuous, embedded part of a student’s learning - not just something you use when there’s a problem.

Historically, tutoring has been reactive: “my child is struggling, I need help now.” That will always exist, but we’re seeing more families use tutoring proactively to build confidence and stay ahead. At the same time, AI is going to change expectations. It will handle a lot of the repetitive, on-demand support. But what it won’t replace is the human relationship - the ability to build trust, adapt in real time, and motivate a student.

The future looks like a blend: human tutors supported by AI tools, creating a more personalized and consistent experience for students.

Subscribe now

“Save us from what?” is my first response. Can it save us from consigning millions more kids to a life without the academic skills they need to thrive? Can it save us from the ceaseless fighting between adult interest groups that make it so difficult to consistently deliver on the promise of a quality public education? Can it save us from doing the hard work of fixing schools, improving instruction, and debating what makes a life well-lived?

There are no silver bullets. So no, tutoring itself won’t save us. Not to mention it’s unclear whether we want to be saved!

At the same time, tutoring does offer a open doorway through which we can consider many of the thorniest issues that impede student outcomes. How does learning actually work? What do we know about how much can be learned in 20 minutes? How does 15 minutes of learning in a 1:1 tutoring session compare with 60 minutes of group instruction? Great teachers know what to say or ask to get a student over the hump of confusion, but what do we really know about how to help the average teacher or tutor ask the best questions? Or to ask the right question in the right moment? What are the limits of learning in a classroom with one teacher and 25 students when every student in the room starts the year below where they should be?

If we let ourselves ask these questions—the complicated questions, the ones that are hard to answer, the ones where many of us disagree—we might actually make some progress. Nothing can save us; we might just save ourselves.

Subscribe now

Or will AI save us?

Always a tech-forward conference, every possible kind of AI was on display in San Diego this week (Rebecca Winthrop, who is much more versed in broader AI applications than I, has a good write-up here!). Here are three takeaways about AI and tutoring from what I saw, heard, and discussed this week.

AI for high-impact tutoring is still not all the way there.

It seems reasonable to believe that in the next five years we will have AI tutoring programs that, for at least a good chunk of students, can help them catch up, fill in foundational gaps, engage them positively in learning, and cost much less than human tutoring. I’ve also been told fifty times this week that I must believe this. But today, in April 2026, these programs don’t exist. Some are being piloted right now, and we’ll see how they do! The tech itself is rapidly evolving, so is it possible that my grandchildren will come home to a speaking teddy bear that can review a quick phonics lesson while my robot housekeeper prepares dinner? Maybe. I never aspired to live in the Jetsons, though, so let’s focus on kids in classrooms today. We still need human tutoring. Human teaching. Human relationships. I’m not the only one saying it—Sal Khan recently announced that his AI tutoring program Khanmigo just didn’t work (amazing write-up by Dan Meyer here).

Relying on teachers to change their behavior has…never worked.

The AI tutoring programs that aren’t high-impact—they are on-demand, or designed to be occasionally leveraged by teachers after a challenging lessons, or just don’t meet the design characteristics of high-impact tutoring (especially on dosage)—might work well for some kids. But the ones that involve teacher use seem to rely on teachers wanting to do things differently. Wanting teachers to do things differently is at the heart of every policy idea in education—manage student behavior differently, use curricular materials differently, teach faster or slower or deeper, etc. Guess what? The average teacher does not want to do things differently. We spend a lot of money on a wide array of education policy initiatives solving for all the pieces that aren’t what teachers do precisely because that piece is so difficult. So I’m skeptical when the message from ed tech is, “We just have to help teachers see the value in this, and change how they spend time.”

The world of builders and creators is invigorating.

You don’t have to like ed tech to acknowledge this. You don’t have to be all-in on AI to see the creativity, energy, innovation, excitement, and belief that almost vibrates through the halls of ASU-GSV. I love it because it’s the future I want all kids to see, that they can grow up and decide to just do something. Build something. Be something. Are there layers of money and power that I don’t understand? Sure. Do I think every product on display is good for kids? Nope. But the big picture idea of America as a place of creation and possibility? That is worthwhile.

There are people who believe we can do things better and change the future. Thousands of people who believe this is possible in education! I’m a skeptic on many things but it’s honestly quite fun to spend a few days letting people paint pictures of a different world. I had many conversations about the dangers of cognitive off-loading to AI, about how to educate kids to be critical thinkers. Adults should be critical thinkers too, and that means entertaining the idea that the person with what seems like a crazy vision might be right. Entertain it, imagine the weight of it passing back and forth between your hands like a juggling ball, commit fully to the thought exercise. Ask tough questions about evidence and the absence of evidence. And then return next year to see what else you haven’t begun to imagine.

Subscribe now

Subscribe now

LC: Tell me about ALP.

SH: The Augustine Literacy Project (ALP) primarily works in Charlotte-Mecklenburg public schools, and has been around for 21 years. But the story of the last four years is one of rapid growth and student success. We are serving 850 students this year, which is triple where we were four years ago. We’ve also decreased the per-student cost by 50 percent.

What we’ve really been building over the last four years is not just a tutoring program, but a system that allows schools to deliver tutoring at scale. It’s been a tremendous ride for an organization that started with a handful of volunteers on the idea of using the Orton-Gillingham approach as a basis for tutoring. But even several decades ago, we thought every child should have access to high-quality tutoring. And we’ve seen that people respond to things that work.

LC: Ayush, your turn.

AA: I’m the co-founder and CEO of Sierra TMS, a full-stack operating system built for high-impact tutoring. We power it with our own virtual classroom called Pencil Spaces, which we designed from the grount up for tutoring, not bolted onto something built for video calls. I’m a kid who grew up in Nepal with access to passionate, kind, loving teachers. But at the same time, the environments around us were challenging in ways that made everything harder. Education didn’t just open doors for me. It changed the entire trajectory of my life. I’m so grateful that I get to spend the rest of my life making education a little bit better every day.

LC: ALP was obviously around long before the pandemic. How did the pandemic shift either the sense of urgency you have around the work, or how you actually operate?

SH: In 2022, I realized we had data from an accidental experiment and control group, of sorts. We had 180 tutors in 2021-2022 working virtually because some schools did not want volunteers back in the building. But then I also had data from our summer program, which was in person. I looked at the data, and it was identical.

That’s when we started to realize something important: the barrier wasn’t whether tutoring worked but whether schools could operationalize it consistently during the school day. So we asked a fundamental question: What if we found a platform that made for this? What would happen? At the same time, we also pursued standardization, or scripting of tutoring. But the big picture was technology and systems.

LC: Can you tell me more about the scripted curriculum approach?

SH: This is key. We use volunteer tutors who tutor twice a week. We can’t afford to train Orton-Gillingham tutors. We need to train system deliverers. So that’s what we do. So now our tutoring program is 100% scripted. Our core program is working with first and second grade students. Most of our kids come in knowing about eight letter sounds. We’re teaching the fundamentals of reading. We use DIBELS to diagnose which of the three levels we offer each student should be in. At many Charlotte Title I schools, as many as 70% of kids are not at grade level literacy in our core grades. So we serve on average almost 60 students per school that we are in.

LC: Did you have to sell your school partners on the move to fully remote tutoring?

SH: Very little. One school didn’t want to do it. I said, “Well, we can’t get tutors to you because of where you’re located.” They changed their mind. Technology totally changes a big problem with segregation, which is geography. On average, our tutors were willing to travel 2.8 miles from their home. But Charlotte is hyper-segregated. When we moved to virtual tutoring, we took all our tutors and put them in one school. That changed everything. We met our goal of serving 70% of students. And when you do that, when you serve that many students, no one cares anymore. The technology, the systems, and the processes have allowed us to get two things, and that is speed and volume.

LC: It sounds like you are doing this interesting thing of taking a kind of old-fashioned volunteer tutoring idea and then meeting the technological moment of 2026. Let’s talk more about that.

SH: There are other reasons besides staffing that using technology and digital devices makes this work. It’s cheaper because I don’t have to print anything. And it’s immediately updatable. I don’t have to run a once-a-year cycle of revisions. I can run one every five minutes if I want. And it gives us the ability to track what’s happening in real time. Most importantly, it allows us to run a system with consistency. That’s what ultimately drives results.

LC: When you first asked that question of “what if there was a platform that could do this for us”, what did you initially think the key characteristics of that platform needed to be?

SH: It was a three-legged stool. The first thing was audio and video embedded. And it’s been three years-plus with Ayush and his team, and I can’t explain why their platform is so smooth. There’s just no friction between the student and the tutor. The second leg on that stool was an interactive white space, truly interactive. And the third thing was tracking and gathering data. We looked and looked, and couldn’t find anyone with all three until we found Pencil Spaces.

LC: Ayush, I know you have spent time in your career building tools for sectors outside of education, the technology sector more broadly. I’m curious what has surprised you about working in education and also what’s been challenging?

AA: Our intention is always to solve for the student. For us, the best technology is invisible. When you plug into a power socket, you aren’t thinking about the power plant. You’re just focused on what you want to do. That’s how we want our platform to feel. What has been most rewarding is the sheer joy of watching a child work well with a tutor. There’s almost no way to describe it. To know that we contribute towards making that a little bit better, it’s just magical. I don’t think anything that I’ve ever done in my life before this comes even close.

What has surprised me, though, is how often people in education settle for “good enough.” We’re in education. Shouldn’t we hold ourselves to the same standard of curiosity we hold our students to? I expected more urgency to improve. That part still catches me off guard.

LC: That’s an interesting point. I’ve been talking a lot recently about this idea that I think in education we do this thing where we’re afraid to say what it is we would really want because we think we can’t get it. I want people to feel comfortable saying what it is that we want so we can talk about what we would need to everyone the things that work. So I agree that we do a lot of implicit settling. That said, Steve, you didn’t settle when you wanted to do a study of your program! What did that study find?

SH: The study was done by UNC-Charlotte, another community partner, which I think is really important. And the headline is this: a first-grade ALP student was nearly twice as likely to get to the benchmark than the control group. Second-grade also had strong results, but not as pronounced.

We gave them three years worth of data, and the results from the first two years weren’t statistically significant. In the third year, it goes off the charts. What happened? That’s when we went to scripted lessons and we introduced Pencil Spaces as our platform. It was the pursuit of scalability, and the scalability had to include standardization and technology. It’s all inextricably linked, in my opinion.

LC: Because you have all the sessions inside the platform, do you look at things like consistent use of scripted materials?

SH: Yes, and we want to get better at it. To give you a picture of it, if you go to Newell Elementary School, there’ll be 20 kids getting tutored in a room. As you walk around, you’re going to see every child within one activity of each other. They could be getting a new concept but they all take their break within 30 seconds of each other. This is not random. We demand that the script be followed. If we hear you’re not finishing the lesson, you’re not doing something right. The only deviation should be for behavior or error correction. And that’s where we are so excited. Pencil Spaces keeps working on their AI tool, and it’s really good, especially at the behavior and error correction parts. We’re having less and less behavior issues because the kids find the program engaging.

Velocity is very important. We have to keep pushing. The tutors often want to slow down because the kid doesn’t have it perfectly. But getting through the lessons is actually more valuable, we find, than performing perfection.

LC: What else are you thinking about these days when it comes to technology and tutoring?

SH: I’ve been around continuous improvement my whole life. If you try to just take a best practice from someone else, you’re not in the right position. You want to be right at the tip of the spear, and it’s dangerous and uneasy. With AI, I’ve never seen anything move faster in my life. But there’s no place I’d rather be. And there’s no one else I’d rather be with than Ayush and his team.

LC: What is the future of tutoring?

SH: The future of tutoring isn’t just better programs. It’s systems that schools can actually run. My mission, my passion is to serve as many kids as possible in Charlotte-Mecklenburg Schools, and now it’s broadening to the state of North Carolina. I see the future of tutoring as figuring out the most efficient delivery modes, figuring out what the different roles are. DIBELS is just one measurement. There’s also the relationship, and we’re not as good at measuring that. The future of tutoring is that we want to present a system that only needs one thing, which is funding. It says, “I need $100,000 to launch in a school, reach 70% of kids, and bring them up within two semesters.” And we know exactly how to deliver that.

AA: The future of tutoring is bright. It’s one of the few interventions that consistently proves effective. That creates space for innovation. I believe the future of tutoring will be more AI-augmented. Time is precious. We should be using technology to free up human time and focus it where it has the greatest impact. That lets us move from thinking about quantity to thinking about quality.

Subscribe now

Subscribe now

LC: Take me back to the first time you remember thinking that tutoring was something you should be thinking about.

PS: It goes back to 2020. We were talking to leaders, particularly executive level leaders of ed prep programs who were grappling with the shuttering of schools, impacting clinical experiences that teacher candidates were required to have.

The question became, “We have folks who are set to graduate in six weeks. They’re not going to fulfill their state mandated requirements because they can’t get into classrooms. And there’s prohibitions on virtual teaching or non-school based settings. What do we do?”

Around that time, everyone is thinking, “how do we shift, how do we pivot as quickly as possible to ensure students are learning?” On our end, we are thinking that teachers who will enter the classroom in six months are going to experience classrooms that look very different. And they deserve some opportunities to engage with kids before then.

When we asked our partners what they were doing, several said they were deploying tutors to provide additional adults in zoom classrooms during virtual learning. And we thought, “Well, that’s fascinating.” Maybe virtual tutoring is something we should be thinking about as a tool in the teacher training toolbelt. So that was really the first time I remember thinking about tutoring.

Around June 2020, we wrote down on paper an idea about how to recover from COVID, and at the center of that plan was this idea that teacher candidates should serve as tutors. We looked at the population of available tutors and said, “there’s 600,000 people enrolled in any given year in teacher training programs. Let’s find a way to get them in front of kids at a time when schools really need support.”

LC: When you think about the work you’re doing on tutoring now, and when you think about where you started, what does that feel like? What are your reactions to that?

PS: The kernel of that original idea is still there. We are actually now even more confident that, when deployed right and trained right, tutors can be successful in virtual settings with kids. But we also know that schools are open again, that both tutoring and teaching is deeply human, deeply relational work. To the extent possible, having in-person tutoring adds such a richness to school communities and to our teachers-in-training that is a really awesome evolution of that original idea.

LC: My understanding of what you’re doing at DFI with tutoring is two main things. Big picture federal advocacy work around potential legislation and funding, and then this coalition of the willing, which involves teacher prep programs, tutoring providers, and other folks trying to build out and test these models. Is that right?

PS: Yes, those are the big pieces, though we’re also working to shift state policies, too. Programmatically with that community of the willing, we are thinking about the specific skills that we want tutors who plan to be teachers to take from their tutoring experience into their classroom. Part of the work we do with teacher prep programs is to think about the design of the actual coursework they’re delivering and the training that tutors get to help them bridge the tutoring experience into practice.

LC: Do we know what those skills are?

PS: One of the areas we think tutoring can help really deepen a teacher candidate’s understanding of the instructional power they have is around how to identify and effectively use high-quality instructional materials. Most tutoring initiatives have a structured curriculum that provides a lot of scaffolds for the educator or tutor to deliver specific instructional interventions for kids. We want every new teacher to be able to do that work for themselves when they get in the classroom on day one. What better way to do that than actually do that work much earlier in your teacher training experience?

Also, DFI’s seminal report, The Science of Learning, lays out teacher actions grounded in principles of cognitive science, which we believe is essential for effective instruction and we think any opportunities to train teachers, including mobilizing them as tutors, should be grounded in this science. For example, when using program-provided curriculum, how are we training tutors to ask questions that focus students’ attention on the "to-be-remembered" to promote effortful thinking.

LC: Ok, real talk. Teacher prep has always been a complicated issue in education reform. It sometimes feels like the thing no one can change, or the thing that never changes. Your whole thing is the teacher prep side. Can you share specific examples of programs that you’ve seen doing things differently? I know there really are teacher prep programs thinking differently about what it means to prepare teachers in 2026. What do you see in the field that makes you feel excited about the future of teacher prep, or teacher prep and tutoring?

PS: I think we are finally seeing teacher prep and tutoring programs commit to the science of learning as foundational to core areas, especially in reading and math. At the top of that list for me is a program at the University of Notre Dame called Tutor ND. They don’t have a formal school of education at Notre Dame, but they have an education degree pathway that students who enroll in then partner with the Alliance for Catholic Education to move into a certification pathway while teaching at Catholic schools. But the tutor training is run out of a cognitive science learning lab led by Nicole McNeil, who is a cognitive scientist who focuses on how students think, and has spent a ton of time in PK-12 classrooms to connect cognitive principles to teaching practice. And they have taken what we know about cognitive science principles and instruction, and how to build that into a tutoring program, including the tutor training. There’s a lot to learn from Nicole and her team as a tutor-teacher pathway grounded in learning science.

For a more traditional four-year undergraduate pathway, I go to the Eastern Shore of Maryland, to Salisbury University. It’s a rural community of folks serving predominantly rural K-12 schools. They are participating in DFI’s Learning by Scientific Design Network, which is a national network of folks thinking about how to incorporate learning science principles into the design of courses and clinical experiences within their teacher prep program.

The dean there, Laurie Henry, is also just really committed to tutoring. Early on, she worked to embed virtual tutoring opportunities partnering with urban schools so that her candidates could have access to different types of classroom contexts. They’re studying in rural Maryland but they could go teach in DC or Baltimore. So they leverage virtual tutoring to expand and diversify the classroom context that teaching candidates are exposed to.

LC: Is it working? Is virtual tutoring enough to give them a sense of what those schools are like?

PS: She’d have to give you the details, but she has been able to grow her network of partners in this work, which seems to signal that there’s something there, this idea that it’s a satellite completion of student teaching to go into different classrooms as a virtual tutor. A different kind of structural innovation is North Carolina A&T. The dean there, Paula Price, partnered with the math department to recruit candidates as tutors, as a sneaky way to expose folks who may never have thought about teaching as a viable career option for them. She is seeing folks who have completed that tutoring experience either add education as a secondary degree and in some cases, even switch their degree to education.

LC: One of the things that first struck me about the way you were framing the opportunity with teacher prep and tutoring is just the sheer number of people in teacher prep programs and the way it could help address the human capital needs of tutoring. Do you still feel like that is something we should be thinking about?

PS: Definitely. We’ve been advocating at DFI and as a collective coalition of ed prep programs committed to mobilizing aspiring teachers tutors for the federal work-study set-aside to increase to 25% from 7%. But even 10% of all that work study would only get us about 20,000 individuals working as tutors. The question becomes what are the other levers to pull on to shift the system. One is policy. Kudos to the US Department of Education for putting out guidance on Title IIA this week around strategic staffing. We should be having a conversation about how to use those dollars to pay candidates for their field experience if they serve in a tutoring capacity that aligns with evidence.

The other lever is shifting mindsets, beliefs, and practices of our teacher prep partners and programs. We did a survey and asked, “What types of pathways are you offering to teaching? Is it traditional bachelor’s, graduate degree, grow-your-own residence, registered apprenticeships? Are you offering a tutor-to-teacher pathway?” In a representative sample of 200 ed prep leaders, one person said this year they would offer a tutor-teacher pathway. But around 20% of leaders said this is at the top of their innovation list.

LC: We’ve been talking about future teachers and tutors – all humans. Where do you see the evolving role of technology and AI in this work? I’m outspoken that I’m bullish on tutoring as a human endeavor. But can you prepare teachers for the classrooms of the future without technology?

PS: First and foremost, we see teaching and, in turn, tutoring as a deeply human endeavor. The question we ask is, “how do we leverage technology as a tool to both effectively train teachers and where possible, support student learning? What’s the interaction between those two virtuous cycles?” We’re asking questions and forming hypotheses around how AI can drive access and even affordability to higher quality, richer practice experiences, including in tutoring spaces. How can AI be incorporated to provide prompts to help a novice educator serving as a tutor to ask the deeper-order thinking question in the moment? How can AI provide detailed feedback reports to the tutor after every session? How can AI effectively reduce the administrative burden in both tutoring and the teacher prep space, matching candidates with appropriate students and the right mentor teachers. I don’t have a fiery response because we’re cautiously optimistic about the future of teaching and training teachers to effectively use tech. But the core of what we want future teachers to be able to do is build relationships with kids and deliver as strong instruction as is reasonable for a first, second, third year teacher.

LC: I love that. Big final question: What is the future of tutoring?

PS: The future of tutoring is that tutoring is part of the fabric of how we serve K-12 students and train teachers in this country. It is both a learning acceleration and a learning remediation tool. It’s an educator workforce strategy and we need to start having this conversation with folks in the educator talent and development space to situate tutoring alongside other big bets in educator workforce development.

Subscribe now

LC: What is Canopy Education and how did you end up running an after-school tutoring provider organization?

WM: Canopy existed before COVID, but the pandemic shifted things around for us. Our tutoring work started out with a program called Real-Time Access to Early Literacy (Louisiana’s REAL program), but then that morphed into the Steve Carter program. We’re now the largest provider of services in Louisiana for the publicly-funded Steve Carter program, named for legislator Steve Carter, who was a champion of education and tragically passed away early in the pandemic.

LC: Why was getting into tutoring appealing for you?

WM: I’ve been in education my whole career. I started out as a Teach for America Corps member working in the South Bronx. That was in 2006, so coming up on 20 years ago. I’ve done many different things in education beyond teaching. I spent a year traveling around the world writing about top ranked schools. We did systems-level consulting, we did early childhood work, etc. And the guiding star for all of this is really how to provide the most impact for the most number of students. Part of that is knowing what you’re good at and following what you believe, and part of it is knowing where the winds are at any given time. With the rise of tutoring and public money for tutoring, it became pretty clear that this was an area where if we could figure out how to do after tutoring at scale in a really high quality, consistent way, that would be the most powerful thing that we could do.

LC: How much of a risk did you feel like you were taking? Were you pretty sure you could figure this out?

WM: It was an extraordinary risk. We invested a lot into this early on to try and stand it up. I don’t know if I would do it again but we were enthusiastic. There was a sense of possibility that we needed to create something new while everything was falling apart. So we went for it.

LC: And your model uses teachers as tutors in an after-school tutoring program, right?

WM: Right. Last year we tutored over 1900 students across 298 schools. We hire teachers who then tutor kids from their own school or their own community. Not the students that they teach, but other students who are at their schools. On average, our students participate in tutoring for 5.8 months and they grow 11 months worth of growth.

LC: How many days a week are they meeting with their tutor?

WM: It varies. We have everything from once a week to four times a week. And this comes down to the scheduling that works for the family. We encourage high-frequency tutoring but sometimes the student has other obligations and can only do once a week. Our data shows these students grow more slowly but are able to participate for a longer period of time.

LC: Did you immediately know you wanted to use teachers or did you look at other options?

WM: We knew we wanted to use teachers and we knew that we could make it work if we paid them well. We have spent our whole careers in school working with teachers, coaching teachers, supporting teachers, organizing teachers. We know the realities of what the day looks like. We know what they’re willing to do for zero extra dollars versus $60 an hour. We already had a network of hundreds of teachers through pre-existing partnerships. And then we’ve been able to grow organically because teachers like to just talk about how much they like tutoring with us.

LC: Why do you think teachers like tutoring? I know you pay well and that’s important, but what are the other parts of it that they like?

WM: What they tell us is they love watching the kids grow. So much of the teaching day, you’re doing full group instruction, and it’s really hard. When teachers are able to stay after school for a couple hours, work with a small group of students who are motivated to be there, and the teacher sees them grow in those moments and sees their acceleration, it’s just incredibly rewarding. We pay $30-$60 per hour depending on group size, so that helps with enthusiasm. They will go the extra mile to reach out to families and develop those relationships.

LC: What makes a student eligible for the Steve Carter program?

WM: They need to be at basic or below on the Louisiana state assessment. Basic is struggling. It’s not great and doesn’t count as on grade level. Although interest in the program has been so high that this fall even basic students weren’t eligible. It’s only kids who are approaching basic.

LC: Talk to me about curriculum. How does that work? How do the teachers know what to cover in the tutoring sessions?

WM: For the Steve Carter program, you first have to apply to become a vendor. All of your tutors must go through a comprehensive science of reading training. You need to use Tier 1-aligned curriculum materials that must be reviewed. So our curriculum is derived from those materials. We’ve reworked the materials for small group tutoring sessions.

Then we have a placement screener, which assigns students to a level. There’s early, mid, and late for kindergarten, then the same three for first grade, and so on. Within each of those levels there are nine lessons. Each lesson is some phonics work, some review work, text comprehension, etc. So we have both a placement screener that shows which lessons students should start with and a separate pre-assessment that’s used to measure growth.

LC: What’s the difference between this and what gets used in a classroom?

WM: This is designed for small groups and it’s designed to focus on remedial gaps, to go deeper on these like foundational skills. The way the curriculum is organized is meant for more of a conversational back-and-forth for the teacher and the student rather than a whole class curriculum.

Making it turnkey and very simple for tutors is incredibly important. We would not be nearly as successful if we were asking teachers to find their own materials for tutoring or use lessons designed for whole class instruction. Doing the reorganization so it’s tailored to small group is critical so tutors can focus on supporting students as opposed to organizing or interpreting materials.

LC: Can you explain more about how students are identified for the program?

WM: The state and districts notify families. But our own outreach work goes beyond sending a flyer home in the mail and hoping parents pick it up. It’s tutors reaching out and saying, “Your kid is eligible for this. Would you like to participate?” We have different schedules, different frequencies. Some kids do the program intensively for two months, some do it once a week for eight months.

Even our lower frequency kids, over a longer period of time, are almost doubling their expected growth. Based on DIBELS data, the more condensed tutoring more than triples expected growth.

LC: The typical critique of after-school programs and why, initially, after-school tutoring wasn’t the emphasis of the post-pandemic tutoring movement, is twofold. One, the logistics seem more complicated because you need people who will tutor after school and get the kids there. And then the second issue, resulting from the first, is that you might not get to the kids who really need the most support. Can you talk about how you see the role of afterschool tutoring?

WM: After-school tutoring is extremely effective when it’s done well. I didn’t realize how much of a riddle we had solved in Louisiana until I started looking around for other states who might be doing this and learned after-school hadn’t worked out as well in other places.

There are three key pieces to making it successful: quality, communication, and consistency.

The quality piece is that our tutors are all professional educators and they’ve all completed the science of reading training that the state requires. The interesting thing about the research you mentioned is that, the studies that say after-school isn’t as effective don’t look at programs where certified teachers are tutoring. They are programs with volunteers or parents tutoring. So the lower impact is more to do with who is tutoring than when. Having professional educators is the most important thing.

Then communication. A lot of the time school systems say it’s enough to make families aware, to check the box that we “informed” families. But we partner with schools, and we empower our teachers to reach out to families and do a lot of hand-holding. The communication is coming locally from trusted institutions and people that they know.

The flexibility in scheduling is what helps for consistency. We don’t watch kids disappear when their own schedules change - we adapt. About 80% of our kids get at least 13 hours of tutoring or more. And we see that once you cross the 12-hour threshold, that’s where we start to see significant impact from tutoring.

LC: It seems like your secret sauce is that you’re using teachers, you pay them well so they’re invested, you have this great curriculum and a lot of communication with families. I’m sure it’s not as easy as you make it sound! My final question is, of course, what is the future of tutoring?

WM: The future of tutoring is in our hands. It’s about people who are providing tutoring, families who are participating in tutoring, all of us telling our story and the story of how successful tutoring is and can be. The future of tutoring is also tied up with funding. That comes down to the decisions of elected representatives, whether on school boards or state legislatures. And those elected officials really do care about what they’re hearing from people. We have to make sure that the powers-that-be understand how important this is for their constituents. We can’t be shy now.

I am also very optimistic about the new federal education tax credit and scholarship granting organizations that are specifically focused on tutoring. In a couple years, it will be possible for anybody who owes taxes to say, “I want to take ownership over where my tax dollars are going and I want $1,700 of my tax burden to go to a tutoring scholarship for a child in my state.”

We know tutoring models that work. It makes sense to publicly fund tutoring to make sure all kids are confident, capable readers and students. The quality of our education is the future of our country. The future of tutoring is helping strengthen the quality of the education students can access.

Subscribe now

My work on tutoring has focused almost exclusively on human tutoring. In other words, not AI. My primary response when I’m asked about the impact of AI on tutoring and education more broadly is that I’m working hard to stay open to a world I can’t yet imagine.

Subscribe now

I also believe that the rapid spread of tutoring in public schools over the past five years is precisely because it’s a human endeavor. Kids across America are hungry for meaningful relationships with adults, especially when it helps them take a more engaged role in their own education.

But I live in the real world, so I’m thinking and reading about AI a lot. I’m particularly intrigued by the work coming out of Eedi in the UK. They call their approach “humans-in-the-loop” tutoring, essentially a human supervising a student interacting with an AI tutor. In a randomized controlled trial this past fall, human tutoring and “humans-in-the-loop” + AI tutoring got pretty similar results. Dan Meyer did a great write up if you want more details.

Bibi Groot, Eedi’s Chief Impact Officer, published a recent update with a surprising and unusual finding: “students who received [the supervised AI tutoring] were 5.5 percentage points more likely to answer the first question correctly on the subsequent topic than those tutored by humans alone.”

Something about the AI tutoring, it seems, made students start thinking differently about the subject area more broadly. Groot is clear that they don’t really know why this happened, what it means, or if it’s real. The leading hypothesis is “the supervised AI prompted students to articulate their reasoning through extended dialogue…the supervised condition kept students engaged in the problem for longer.”

What I most appreciate is that the Eedi team seems remarkably grounded. There’s a big RCT underway and the Eedi team seems as curious about what they might learn as the rest of us. Usually, companies message more certainty and excitement about their product, even during a research project. I’m even more intrigued by Eedi’s human-in-the-loop tutoring because of their measured tone.

Meanwhile, I’m also thinking a lot about tutoring in the out-of-school space. More on that in a forthcoming Q&A, but I am quite bullish about the opportunity to build effective tutoring programs run through or in partnership with schools, that take place out of school time. In a few weeks, my organization will publish the second edition of the State of Educational Opportunity, our nationally representative survey of 20,000 parents. Sneak peek: parents are procuring a LOT of tutoring! Lots more to come on that soon.

Subscribe now

In recent weeks, I got to participate in a fun discussion about tutoring through Mike Petrilli’s great substack Schooled, did an awesome webinar with the Campaign for Grade-Level Reading, and started prepping for book events in Nashville, Portland, Honolulu, and more. People want to talk about tutoring!

The future of tutoring is big. Onward, 2026!

Subscribe now

LC: What does tutoring look like for Pearl right now?

JF: There are three big pieces. The first is the shift we’re seeing to in-house tutoring by districts. The second is around the greater need for data transparency when working with vendors. And the third is our approach is AI. You can’t be in tutoring right now without being able to point to something about AI.

LC: Let’s start with this in-house shift. What does that mean?

JF: Pearl provides infrastructure for the tutoring industry. We’ve had more districts reach out to us about the need to bring their tutoring in-house in the past four months than in the past four years. It’s a simple equation for them. Tutoring works, no one can deny the evidence of how high-impact tutoring works. But paying vendors at the level they were paying is unsustainable. How do we keep doing tutoring in a sustainable way?

We’re seeing districts and states ask for help around operations and data collection for tutoring. And more of them call it an “intervention” than tutoring, by the way. At the same time, the idea that a district is going to be able to stand up a program to support every single student in the district is an optimistic one, to put it lightly.

Instead, districts and states are bringing programs in-house and then also approving a set of vendors to work alongside them. So we built a new module called the Data Hub where we integrate directly with vendor systems. And then we’re building a large language model to sit on top that will allow state and district administrators to chat directly with their data. They can immediately know which school has the worst tutoring attendance, what’s the trend, and what’s the recommendation for how to fix it. We’ve leveraged a lot of the knowledge base out of [Stanford’s] National Student Support Accelerator to help bring the time to insight from what’s often six months to just a few minutes.

LC: You’ve been busy! Let’s dive into the last thing you just mentioned, the language model. Can you say more about your AI work?

JF: You can’t look at Google alerts in the morning without seeing a hundred new AI features or tools released in the tutoring space. Our whole perspective is to focus on building and delivering AI tools that enhance the human connections, amplify the human power in our platform, not replace it.

The admin agent I mentioned is focused on three things: automating mundane tasks, giving advice, and reducing the time to insights. Importantly on advice, we know that people want to know where recommendations are coming from. We’ve taken all the research from NSSA, Accelerate, Proven Tutoring and fed into our model, so when it makes a suggestion to an administrator about how to improve their program, it will also reference the actual research study behind it.

LC: I love the emphasis on data in the tutoring movement. Do you think this is part of a broader movement of districts looking at data on lots of things, or is there something about tutoring itself and the way the movement has evolved that lends itself to being data-driven?

JF: Students are in a tough spot right now. I believe states are seeing their test scores and saying, “Hey, we really need to solve this problem and we need to invest in things that work.” I think that’s broadly driving the interest and desire for meaningful data sets that can be easily analyzed.

LC: Your company recently published a Trends in Tutoring report. Were you surprised by any of the trends that you found when looking at the data for the report?

JF: One of the surprising pieces to me was the existing emphasis on high-impact tutoring because it just isn’t easy to do. I was expecting to hear more states or districts say “we’re going to keep tutoring, but maybe do larger group sizes so that we can make it more cost effective.” The insight we learned is that instead of trying to be more broad and cover more students, they’re picking a specific group of students to give it to. Let’s do it for 20% of the kids and do it right as opposed to doing it for 80% of the kids and trying to dilute it.

One thing that wasn’t surprising is that there’s an almost binary perspective about tutoring right now. Either you’re doubling down or you’re not doing it at all. Some states tried tutoring, it didn’t work, and they said, “we’re done.” But they didn’t even try to figure out why it didn’t work.

This gets to the big hairy problem Pearl is trying to solve. If they had the right data infrastructure from the start, they’d be able to provide continuous improvement to tutoring.

Subscribe now

LC: I want to unpack even more what it means for districts to move tutoring in-house. What does this really look like?

JF: It’s both state-dependent and district-dependent. The majority of districts we see going in-house are in a state with some level of tutoring legislation motivating them. Tennessee, Texas, Louisiana, Virginia. That legislation often defines who is allowed to get paid as a tutor. We’ve seen that retired teachers are an easy first call. Paraprofessionals are another group we’re really starting to see take hold. The New Jersey Tutoring Corps, for example, got their program approved as an apprenticeship, which should let them tap into Department of Labor money.

LC: It sounds like one of the things you’re saying is that maybe a year ago you wouldn’t have thought we’d be where we are right now.

JF: We made a massive bet that post-ESSER, districts would want infrastructure as much as they want vendors. We are very pleasantly surprised to see how much of the market is leaning towards that. But of course, vendors will always be there.

LC: I think this all makes sense. I think the ESSER money, with tutoring anyway, made people less afraid to try. The money created a runway to stumble, fall, and get back up again. What’s fascinating about what you’re describing isn’t just that districts are continuing tutoring, but that they are willing to rebuild their programs.

JF: I’ve had so many calls lately with districts that say, “We don’t see things with the level of evidence tutoring has. So we’re not going to throw it aside to look for something else.” It’s also true that districts within each state are so different. There’s no cookie cutter or silver bullet here, it just requires incredible configuration and flexibility and systems or data tools to be able to meet the community where they are, not force that community into a box.

LC: What is the future of tutoring?

JF: It’s a cohesive, blended infrastructure that can be personalized to the student. Imagine a world where there’s a platform helping a student get tutoring by a local teacher. But the student could also get matched to a vendor. I think tutoring will end up a hybrid of in-person, online, and AI work. With AI, imagine the world where Joey goes to the platform and says, “I need help with trigonometry for a test tomorrow.” Joey uses the AI tool for 30 minutes, and it sends a report to Joey’s tutor saying, “Joey worked on this for 30 minutes, and here’s a question you can ask him to check for understanding.” The future will be an orchestra with all these different instruments supporting a student’s learning, with humans leading the work.

Subscribe now

Subscribe now

LC: Who are you? How long have you been working on tutoring?

SR: I’m a founder of Carnegie Learning, which we started 26 years ago, and prior to that I was part of the research team at Carnegie Mellon that developed the instructional approach that became the basis for the company. I wouldn’t necessarily say I’ve been working on tutoring, it’s really been a focus on instructional curriculum.

A big part of that was developing an intelligent tutoring system, which is an AI program that acts as an individualized instructional coach for students. It’s kind of like tutoring, but it’s not the same as human tutoring. It uses that tutor word. We actually talked about that a lot in the early days. I think the word originated in the UK, where they talk about tutors as being more like instructors.

The question we keep asking is, “why do personal human tutors have such a big impact?” and is followed by, “what can we do in this program that will help us approach that kind of impact on students?”

LC: How do you answer that question of why? I’ve wondered about this a lot - why does tutoring work?

SR: There are at least two aspects of it–cognition and motivation. My background is in cognitive psychology. For many years, I focused on cognition because, frankly, I had a better handle on it. But I now think that motivation is at least as important as cognition.

One thing that’s clear is that personalization responds to students’ particular state of knowledge and misconceptions and instructional needs. So a tutor will naturally know whether the student understands the concept, are they ready to move on, or do they need more instruction? Good tutors are flexible about providing students with explanations and activities and instruction that responds to what they know.

The worst thing that can happen to a student in school is the student telling the teacher, “Here’s how I think about this problem.” And the teacher says, “Well, that’s not the way we do it in this class.” Motivationally, that is a really clear statement to the student that you don’t belong here. It can be burdensome on a teacher to listen to a student’s explanation and understand whether that’s a reasonable approach. But that’s really how you produce learning.

LC: That’s interesting, because it also seems like it could push students to a more shallow sort of cognitive approach. Instead of thinking deeply about how they understand this problem, the student is just told “do it like this.”

SR: If you’re building understanding in a certain way, and the teacher says, “forget that”, the new approach may not connect to your prior knowledge. In a one-on-one tutoring situation, the tutor has the time and social imperative to take the student’s reasoning seriously. You’re less likely to be dismissive.

LC: What about the motivational side? I’m curious how you think about humans versus technology with this piece. I think the cognitive side is an easy sell for technology but I worry about the motivational side in a technology-first approach.

SR: I agree. I’m a strong AI guy, but I’m not convinced we have the motivational side quite right. In school, students are always being asked to do things they don’t inherently care about. Nobody wakes up in the morning and says, “I wish I knew how to solve quadratic equations.” They’re being asked to do things on the basis of trust–that a trusted adult is giving them things to do that are worthwhile. Having a human being playing that role is the essence of the motivation. A lot of it is trust.

LC: I think a lot of people would agree, but would also say that humans are expensive. They argue that if you’re trying to solve for instructional challenges at scale, you have to use technology and AI is going to get us there. Do you agree with that?

SR: Sort of. We can’t give everyone a one-on-one human tutor. What we ought to be thinking about is how we can give students the best possible instructional experience in software, and how we optimize teachers around that. Let’s say you have 30 kids in a classroom working with our math program. The teacher is able to talk to at least some of those kids one-on-one during that session. That personal instruction is valuable. If we can figure out how to best distribute that teacher’s time, which is the most important resource in the classroom, then we are thinking about the teacher and the software as an instructional system that works together. Maybe the software detects which students would most benefit from interaction with the teacher.

LC: Is that what happens? Your software can say, “Hey, you should check in with Steve”?

SR: We did a study on tutoring a few years ago. We gave students unlimited access to an online tutor, and we looked at both why a student would choose to talk to the tutor and what was that interaction like. There were two main things we learned. First, most students never asked for help from a human. And then there are some students who ask for way too much help.

Our software does alert the teacher about which student is most likely to benefit from their direct help. It’s not necessarily the student who is struggling at the moment, though. The struggle can be good. Students are solving math problems. Maybe they’re making errors, but they’re also improving, they’re learning from those errors. We don’t want the teacher to interrupt that moment.

On the other hand, there are kids who will never pull the teacher over, they’re just flailing and they’re not making progress with their errors. And that’s when you want the teacher to come over. We’re trying to make that distinction.

Subscribe now

LC: It feels like we’re talking about the typical kid in a classroom. But a lot of tutoring is directed at kids who are significantly below grade level. Does this approach work the same way with a student who is really behind?

SR: I’d love to say it’s the same thing, but the truth is there are important ways that it’s different. A lot of those kids come to school everyday thinking, “I’m so far behind, I’ll never catch up.” Software saying that’s not true or AI saying that’s not true just isn’t going to cut it. For students who are really behind, it’s about a trusting relationship. Maybe they’ve never had someone who really believed they could succeed. For those students, hopefully, one-on-one tutoring can really work if they’re with a tutor who they can develop a relationship with, who really cares about how they’re doing, and is able to connect.

LC: When everyone started talking about tutoring in the pandemic’s aftermath, what was your reaction?

SR: Honestly, I had some real fear that the opportunity would be wasted because people would spin up these programs really quickly without thinking about the issues of making them work well. This relates to what I was saying about motivation, about the relationship between the tutor and student. It would have been nice if there were more controls on steering people towards models that really emphasized attendance, connection, those things that we know work, but it didn’t always work out that way.

LC: Do you think tutoring will stick around?

SR: If you look at upper-middle class communities, there’s a lot of demand for tutors. And if they want to pay for them, it’s not surprising that they’re effective. The question really is how can we afford to provide some of that for people who can’t afford to pay for it from their own pocket? There’s a big tendency to say, “well, we shouldn’t pay for that human tutoring because we have this AI chatbot now.”

And certainly AI is better than nothing, and chatbots can be very effective in some cases. But I think they don’t solve the problem in the way that people think they solve the problem. They provide adequate fill-in instruction. They’re not providing good guidance on what you should do as a student, what you should be thinking about what you do next.

Overall, with AI, we have to really think about the question, “why are you learning this?” We owe it to students to have an answer. There’s emerging research that students use AI for things they feel are busy work or useless. We have to be more explicit about why they are learning something.

LC: Will classrooms 20 years from now look different than classrooms today?

SR: My immediate response is that classes will be the same as they ever were. But we’re at the point where we are thinking about more flexibility. And in fact classrooms really aren’t the same as they were decades ago. Personalization is the biggest challenge. I think AI and advanced technology will help with that. Without personalization, you’re not going to get the motivation. And the school system as a whole has to have accountability.

LC: What is the future of tutoring?

SR: It’s not any one thing. It’ll be a range of things. One-on-one personal human tutoring will always continue because people will pay for it. In the short term, I think there will be more reliance on AI. And then some inevitable backlash.

We’ll come back to a reasonable balance where we figure out better holistic models of thinking about tutoring as part of the instruction, resources and methods that we can give to students and ideally figure out what role it plays for each student.

Then it becomes less expensive for everyone and we can figure out where AI does substitute so that we can save some expense there, without completely getting rid of human interaction. I think that’s where we end up. But it’s going to be different for different kids and in different situations.

Subscribe now

LC: Do there continue to be new contracts and new opportunities in tutoring? Are there districts who want to continue or new districts who say, “now we are ready to try this thing”?

JS: We’re picking states to operate in based, in part, on

where funding is available. California has ELOP funds [Expanded Learning Opportunities Program], and other states funding tutoring from the legislative side. People are still committed, we’re seeing a lot of renewal. I don’t think the overall market is growing beyond where it’s been since ESSER, but I also don’t see it shrinking either. There’s been contraction but now I think it’s leveled off. We’re certainly not seeing a lack of interest in tutoring, but the district financial conditions can be a barrier.

Subscribe now

LC: Is Littera a platform or a full-service tutoring provider?

JS: It’s both. Right now a majority of our business is full-service tutoring, but the data side of the business is becoming more and more interesting. There’s a new opportunity for a state that’s trying to wrap their head around the question of “can I get information about if this money is being spent well.” We can now provide a data layer to help answer that. For example, right now we are working with tutoring vendors at various degrees of technical expertise and making it possible for Florida to get reliable data on their relatively new tutoring initiative.

LC: Is there consensus within the field as to what core data pieces are that we should be collecting?

JS: We’re trying to help drive the conversation towards some level of standardization, in partnership with Accelerate and state fellows. We’re not focused on collecting outcomes data like state assessments because states and districts already have that data. What they need is the ability to compare how one vendor performs compared to another vendor, compared to a third vendor. You can’t understand differences in outcomes without more information than just test scores. A lot of it is implementation data. We can easily generate a list of the kids that missed tutoring yesterday. But then we start thinking about the school building. Who in the building should get this list every morning? Who in the building needs different information to keep the program working well? It’s important to move past collecting data to distributing it to drive better implementation.

LC: Do you think we’ve gotten better the past few years at knowing if tutoring is working?

JS: When I started Littera in 2020, tutoring programs were being run the same way I would have run them at a Kaplan center in the mid-90s. All on a spreadsheet, calling tutors, guessing at availability. We couldn’t answer the question “is tutoring happening”, forget about “is it working.” Now we’re further down the track on “is it happening”, putting definitions around attendance and dosage. A lot of people, at least in our little space, are talking about these things in a consistent way.

The next logical question is, “is it working?” That’s where outcomes-based contracting (OBC) is helping to drive the answer. But the requirement there on a provider is much higher. And then the next question we need to get to is what are the cost economics of these different profiles of tutoring programs. With our work in Florida, the state will have data on 15 providers and can actually see what’s worked better than others and what the outcome per dollar is, if they choose to look at it that way.

LC: Why is there still such a range in cost? Why hasn’t the field converged more on a standard pricing model?

JS: There’s a lot of reasons. The economics are different for for-profit and non-profit providers. For-profit needs a model that the market will ultimately fund and it has to work within a school district’s economic structure. A lot of non-profits, including some we work with, don’t have that constraint. Their constraint is “how much funding can we raise to serve how many kids.” That’s a different dynamic that drives a different pricing attitude.

Tutoring also remains a very fragmented market. You and I could start a neighborhood tutoring company with almost no overhead, for example, and charge whatever we value for our own time.

But unlike a consumer business, districts aren’t really buying on price. They have a budget and it’s a question of “how many kids can we serve with this particular solution.” There are a lot of factors that get weighed in the evaluation of an RFP and price is often less a priority than you might think.

LC: If you were going to give advice to districts making these purchasing or contract decisions, what advice would you give?

JS: They should talk to people with similar contexts. Don’t just look at a program that says “hey, we did a great job in this other city” but really identify similar contexts broader than academic need. How will tutoring really fit into your district? How will it fit inside the school day?

I think outcomes-based contracting is critical because it really helps define how the work goes. Almost never have I been in a situation where the district feels like all the kids that need tutoring are getting tutored.

LC: Are you seeing more RFPs (Request for Proposal) for outcomes-based contracts?

JS: Most of our sales do not come from RFPs. Most are conversations we get into through outreach or referrals. When we explain what Littera does, we always say, “one of the things we do is provide contracts based on outcomes, and here’s what that entails.” Everyone’s intrigued. Sometimes we talk through outcomes-based structure and it won’t make sense for a district. But we introduce the idea at every opportunity.

LC: How much of the tutoring you’re seeing or supporting in person versus virtual?

JS: We only deliver virtual but a number of the folks using our platform are in-person. The North Carolina Education Corps was our first customer, and they’re all in-person around the state. I saw something recently saying that tutoring is going from virtual to more in-person, but that’s not what I’m seeing. I see virtual programs growing. Most of the communities most in need of support are human-capital constrained. I can find a lot of tutors that know the alphabet, but I don’t find that many tutors who can teach Algebra I. The degree of difficulty of staffing is high. To think an in-person model is going to serve those needs isn’t realistic.

My prediction is ultimately the outcomes for virtual tutoring are going to be better than in-person tutoring at a system level. What’s important is quality control on the tutor, being able to make sure they’re doing the right things and have the right expertise. If I have to get somebody to a specific classroom on a Tuesday at 3pm, my local sample size to do that is quite small. If I use a virtual tutor, I can find any tutor who is the right fit and have more quality control. That’s what the CUNY Reading study showed, right? It doesn’t matter if it’s on Zoom or in-person. If the tutor is good, they got good results.

LC: What are you thinking about AI these days?

JS: I think the ability to access and use data with AI could unlock insights currently impossible to get to. We have to train ourselves to ask the question that we think isn’t possible to answer. We have to ask it anyway because sometimes it’s now actually possible to get an answer. We use AI throughout our operation - back office, dev, support - but I get the question a lot about whether AI can compete with a live tutor. On content, I think so. On outcomes for tutoring with children, I am very doubtful.

LC: What is the future of tutoring?

JS: I hope we get to a place where districts have the information to be strategic shoppers of tutoring - and other interventions for that matter. There are now states running these programs and they are going to be able to look at all manner of tutoring options and modalities to measure cost-benefit. As more outcomes become available, it’s possible to imagine an ‘outcomes-per-dollar’ measure that could be applied across providers, for-profit and non-profit, virtual and in-person.

Subscribe now

LC: It’s fall 2025. What does tutoring look like for Step Up?

SO: This year we’ll have 3,000 tutors working with 5,000 kids in 2nd-6th grade. There are now 32 colleges where we are work-study approved, plus an additional 11 where we are working through a college corps or similar program with funding where students are paid and placed in tutoring organizations. We also have a CTE partnership, a high school model in which high school students in CTE education pathways are paid a stipend to tutor through Step Up.

We grew out of an LAUSD initiative which was to recruit adults from the community to volunteer to tutor kids over Zoom. When we launched in 2020, we got thousands of people to join and asked for a three-month commitment. Essentially zero people stopped after three months–it turned out to be a sticky relationship. You’re working with one student and you’re working with their family, so most people tend to feel quite committed.

We’ve been working on accessing work-study funds since 2020. That was when colleges first reached out to us because they saw Step Up, with our virtual model, as a way to solve a problem for them. All their community service placements were shut during COVID and they needed 7% of those Work Study dollars to go to community service placement. So Step Up solved that for them. Now, we turned up the volume on Work Study and just did a national expansion outreach to colleges across the country.

LC: What’s the geographic distribution of those colleges? And the high school CTE piece?

SO: We’re in 14 states, and half of the colleges we work with are in California. The CTE program is just with LAUSD.

LC: Do you think differently about how to use high school kids as tutors? Are there different operating parameters?

SO: It’s a lot more high-touch. We’ve had to walk them through how to get fingerprinting done, just get them set up to begin tutoring. The idea was to have this be a true work-based learning experience, and it just takes some effort. We do have an interesting workforce to draw from to support these different tiers of tutor. Many of our adult volunteer tutors become extremely committed - it becomes part of their identity. Not only tutor, but many take on additional volunteer roles. We have several teams of volunteer tutors who do all of our interviewing for prospective tutors, lead our live training - .we have five to eight trainings per week delivered by volunteers, and a select group help with family onboarding and support. These teams of volunteers deeply extend the reach of our program.

We think a lot about how to combine a workforce of volunteers with the little bit more transient workforce of college and high school students to make this something that can grow, keep a very human element, stay relational, but keep our expenses down. Our work now is to build a national tutoring corps. Our vision is to build a national tutoring corps that has a big tent, where you have people from a lot of different walks of life who can come in and find their place.

LC: I understand the tutor profile you work with, but can you talk about how your tutoring actually works?

SO: We’re in four states and over 40 districts, across southern and northern California, and more recently in Arizona, New York and Washington, DC. We work with individual schools to enroll students in tutoring. The school’s role is primarily to identify the students that are most in need of tutoring support. Once they are enrolled, our program is very focused on parent support and parent engagement. We call every parent, onboard them, and match them with a tutor.

We have an intensive focus with parents on engagement and attendance because this is an out-of-school time program. But the parents have very much dialed in. Our spring semester attendance was north of 75% for an at-home tutoring model of mostly low-income, low-tech families. We work with a large percentage of non-English speaking families, for example. They get weekly texts from their tutor about what their children are working on and an accomplishment they can celebrate. We do a lot of regular incentives that drive attendance, and we’ve really focused on bringing parents into the process and giving them a lot of visibility.

LC: I’ve been thinking about the out-of-school time piece a lot. Because I understand all the reasons why we focused on tutoring during the school day and I think those reasons were really right at the time. But if we’re really thinking about the evolution of this work, it seems like the family engagement part might really matter.

SO: We have taken what some see as a liability—which is tutoring delivered outside the school day—and flip that into an asset. The asset is we have the ability to work directly with the family, with the person who knows their child the best. We know the importance of family engagement in education, enforcing learning at home. What we’ve now found is that schools are saying these families are engaging now more with school. They have a better idea of what’s going on with their child’s learning. They’re watching kids doing active learning, having instruction that’s happening at the kitchen table, and they’re coming back into school. Parents don’t necessarily separate tutoring versus what’s happening in school. It’s all just learning. So I think there is a great opportunity for tutoring to help parents get more involved in their child’s education overall.

LC: This seems like the kind of information parents are really hungry for. We know parents are starting to see through grade inflation, and don’t necessarily trust report cards. Parents want real information. Do you mostly tutor reading or math?

SO: Parents can choose either literacy or math, and the student stays in that track. All our tutors are trained in both but the students aren’t switching back and forth. If you choose math, you are doing math tutoring for three months. If you want to switch after that, you can switch to reading and stay with the same tutor.

We’re building a national tutoring corps so we want to make an experience that really serves that. Let’s make a really clear roadmap of what third grade mastery looks like, for example, and then surface that for the tutor and the student. “Here’s where we are going to get to. Here’s the mountain we’re going to climb and we’re going to do it together.” And have built-in opportunities for celebration. A huge piece of what relationship-based tutoring enables is students having regular opportunities to be seen as successful.

One of the things we train our tutors on is the mentor mindset. This is some of David Yeager’s work on high expectations, high support. Building self-efficacy is really believing, having an identity where you feel successful, and that is done most by being successful.

LC: What are you hoping to learn and achieve in the New School Venture Funds’ gen-AI math tutoring cohort?

SO: We’re developing several tools that pair novice tutors with scalable solutions for coaching and support – lightening the instructional lift while making them more effective tutors. We’re building two tools. One’s an AI feedback tool. It generates very detailed transcripts of what goes on in a tutoring session–dialogue, whiteboard activity, gestures, facial expressions–and then analyzes that against a framework of high-impact tutoring practices. Then we generate back for each tutor a bite-size report right before their next session. “Here is one tutor move that you can focus on that will have a big impact.”

And then the other thing we’ve built is a tool called Whisper and that is a real-time assistant or co-pilot. Whisper takes a screenshot of the math problem you’re working on with a student and creates a script of how to explain this problem in a way that gets at the foundational concept. One of the challenges with math tutoring is tutors not understanding the actual content. Tutors need to activate the tool themselves, we’ve learned from user research that this is important. They activate it with one click and instead of stopping to think, “How will I explain this division of fractions problem to this student?” the tool prompts them with an approach. It gives them a few minutes back and lets them go with the flow of tutoring.

Subscribe now

LC: What is the future of tutoring?

SO: It’s a national tutoring corps that exists in every community. It will be the norm for every child to have access to a caring and trained tutor. This does exist right now for families who can pay for it - but we can make it that way for everyone. We have 5,000 students this year and we have a path to 25,000 students. But what does it look like and what kinds of systemic changes need to be enabled for this to exist everywhere? There are 20 million adults and 2 million college students in this country that volunteer over 20 hours per year, 42,000 students on federal work-study that must do a service job, and 580,000 enrolled in an undergraduate or graduate education program.

If we capture just one percent of this population for high-quality relationship-based virtual 1:1 tutoring, that’s 250,000 people. This could make a meaningful dent in access to opportunity for kids in low-income households. The supply and demand is there, we just have to figure out the pieces to connect them. I am fired up about this. I think it’s possible and we are making progress.

LC: How long have you been working on AI and tutoring?

KK: I just gave a keynote talk at the AI in Measurement and Education conference, and my subtitle was “my 40-year journey.” Because in January 1985, I started working on an artificial intelligence tutoring system project. And ever since then, I’ve worked on intelligent tutoring systems.

LC: What were you hoping to learn from that very first project?

KK: We were just sitting around building stuff we thought would be good! I always had more of a scientific and practical than technological orientation. I like the advanced technology but I really want to ask, “What is the nature of intelligence?”

Subscribe now

LC: I sometimes think that one of the questions we don’t totally understand about tutoring or even teaching is exactly how or why it works. Do we really understand how humans absorb information?

KK: I was just reading Kahneman’s book Thinking Fast and Slow, and there’s a sentence in it which is something like, “we don’t know ourselves.” I think that issue is fundamental to the barriers that exist to doing more tutoring in schools. There’s also an expert blind spot issue where experts don’t know what they know, and so they often misjudge what’s hard for students. They think that story problems are harder than equations because it takes them longer to do the story problem since they first translate it to an equation. We found this isn’t true for beginning algebra students. They’re actually better at the beginning story problems than at the beginning equations.

LC: Why is that?

KK: In this study, we had story problems, word problems, and equations. The big gap is between the word problems and the equations, not between the story problems and word problems. So it’s more about English versus Algebra, and you can think about algebra as a second language.

In the end, we learn through experience and practice with feedback in varied situations. Deliberate practice is crucial, so much more important even than the time telling a student how to do something. Doing means I’m actually trying to perform something. We don’t have knowledge to have knowledge. We have knowledge to do stuff. And practice is good because if we’ve been thinking and struggling to do something, we are more attuned to what we’re being told. That’s why tutoring is powerful.

LC: Can you say more about why or what makes tutoring so powerful?

KK: The position of problem solving and instruction is tightly integrated. There’s a paper from maybe 1980 on intelligent tutoring systems in Science, John Anderson wrote that paper. The way these tutors work is not like the typical instruction and then problem solving practice. Because in tutoring, we learn through problem solving with guidance from instruction. Good tutors, whether computers or humans, integrate the practice and the instruction. So then you don’t necessarily need as much explanation.

LC: As someone who’s been thinking about this for a long time, are there moments when you wish school or district leaders understood something differently than they do?

KK: Most teachers have these intuitively-based, experience-based beliefs that telling has to come up front but then the practice is usually at home on homework. There’s no practice with feedback. There’s instruction, instruction, practice without feedback, feedback later, and maybe some discussion around that feedback.

There’s some evidence that if you do preparatory activities which engage in the topic and then you hear a lecture, you get a lot more from the lecture. We’ve been running studies that just do practice with feedback, and the content of the feedback is the content of the lecture. We find the learning is better than it is with a lecture.

LC: Can you explain now what PLUS is and what the value of PLUS is?

KK: PLUS is a loose acronym for Personalized Learning Squared. And the square reflects two things. One is between engagement and learning the content, and the other is between computer tutoring and human tutoring. The squared idea is that you don’t get good outcomes unless you have both engagement and good content-based tutoring. [LC: Essentially, PLUS is an ed-tech math program that uses

human tutors to support student learning and participation.]

Subscribe now

These tools are really effective when they’re used. But we can see in the data that a lot of students aren’t using them. We also see things like gaming the system where the students sort of participate, but really are just clicking their way through. The more students do that, the less they’re learning. That’s an engagement problem. We’ve also learned that if we look at how long it takes before a student logs in at the start of class, that’s very predictive of their engagement and overall learning.

LC: How often are students logging on?

KK: We asked teachers to commit to using it in class at least one day a week. The tutors are college students, and they are working virtually. The tutors see a dashboard that indicates how well the students are doing on the PLUS program from their group of three or four students, and they can decide who to work with.

LC: So if I’m the tutor, I can see their data to figure out who’s engaged or not, who’s succeeding on the task or not. And then I can say, “Ok, Ken, let’s talk about this.” Is that right?

KK: You can also say, “Why haven’t you logged in yet?” or a nicer version of that. We’ve seen in other situations that there are observer effects. It’s kind of like having a workout buddy. We have data that shows that even when the tutors aren’t with an individual student, the students are more engaged during the sessions when they know the tutor is logged in.

LC: How do you think about the relative value of human tutors versus AI? It seems like you’re trying to build something with the idea that the human connection is critical for engagement.

KK: We know so much in psychology about cognition and social motivational issues that we don’t bring to bear in classrooms. Most of our tutors are Carnegie Mellon students who tend to come from tech fields, engineering. They know the math pretty well. Most of our tutoring training is focused on these social motivational issues, building relationships.

LC: What are the big questions you still feel are unanswered about tutoring?

KK: I have questions about how we change teaching practice based on what we’ve learned from tutoring. One of the key ideas of deliberate practice is being clever about the exercises such that they address things that many students are struggling with at the right level of complexity. That’s part of why computer-based tutors are powerful. And then that changes how we think about the tasks that we give to students. If we can build all these insights into the technology, we can really scale it.

It’s still important that humans are involved to gauge progress and motivate students. Peer collaboration is wonderful for that, too.

LC: AI has been around for a long time, as you’ve shared, but in the past few years we’re now in a different, boosted era with large language models. Are you thinking differently about AI as you look towards the future of your work?

KK: What’s new is we can build all these programs much more quickly. With my grad students, we can go to Chat GPT and make up a scenario. Boom. 70% of the time it’s pretty cool, 30% of the time it’s garbage.

LC: How do you help your grad students figure out which is the 70% that’s pretty good?

KK: That is the hundred billion dollar question these days. A lot of programming is test cases, and you see if the AI passes the test case. But if you are missing key edge cases that will happen in the real world, you can conclude it works even when it doesn’t.

I’ll be working with my PhD students and sometimes they’ll bring in data and I think, “wow, those results are weird.” My spider-sense goes off. Now we have an opportunity, even an obligation to figure out what higher-order thinking skills are needed, especially on the evaluation side. My spider-sense is a way of saying most of my expertise is outside my awareness. It’s not magic. There is something I’ve learned over time that allows me to have that sense.

LC: What is the future of tutoring?

KK: Issues around effectiveness are one thing, cost is another thing. It’s going to be easier if we can provide tutoring in a cost-effective way. That’s where science and engineering needs to continue to improve. Over the past year, we wrote new software that is doing a better job than our human tutor supervisors at keeping the consistency of tutor-student match going, for example. And we have better filtering on “dependable” students with respect to coming to tutoring when they say they will come. Every little thing helps. I like to make an analogy to Moore’s Law, with the doubling of computer speed. Is there a discovery that produces doubling every year for tutoring? Obviously not. It’s tons of micro-engineering improvements. We have a national tutoring observatory with Cornell, MIT, CMU, and a bunch of tutor providers to amass more data to work out these micro-efficiencies. I worry the most about what happens if we don’t get a better understanding of how important deliberate practice is.

3 Takeaways

1. On the power and importance of knowledge.

In a session on literacy at ExcelinEd’s annual summit, AEI’s Rob Pondiscio said something I hadn’t heard before: “We think reading is a skill, that comprehension is a skill. That’s incorrect. It’s a condition we create.”

Pondiscio’s point is that creating the conditions for reading comprehension is much more about ensuring kids have requiste background knowledge than we like to give credit for. He explained, “Literate speakers and writers make assumptions about what their audiences know. If we don’t know it, the Jenga tower collapses.”

The (somewhat obvious, but often missed) reason this matters is for the kids Pondiscio described as “school-dependent.” These are the kids who aren’t being read to at home, whose parents don’t have the ability to give them the background knowledge they need to access the full universe of texts and books.

Hopkins researcher Ashley Rogers Berner chimed in to point out that this is exactly why you can advance literacy while teaching history, social studies, and other subjects beyond “reading.” I saw a presentation a few years ago at Harvard about a science curriculum that builds background knowledge that helps kids become better readers. This all seems like common sense when you think about it, and the framing of “creating the conditions” for literacy feels useful.

2. The long-term economic impact of improving math is decent.

If you are breathing in the education world this week, you know that college kids can’t do math. My org published a report that outlines policy ideas for fixing math back in August and I was happy to hear our friends at ExcelinEd share about their new Math Matters report and 50-state scan. The framing statistic that struck me is that “boosting math proficiency can raise a student’s future earnings potential by 5-10%.”

The study behind this statistic came out last year and looks at earnings at age 30 across various demographic groups. Researchers simulated the effect of a variety of measures in both early and middle childhood, including health, parent-child relationship quality, reading test scores, and math test scores. In preschool, early elementary, and middle elementary, the math test scores have the biggest effect on how much money you earn at age 30.

Now, maybe increasing earnings potential by 5%-10% doesn’t feel like a lot. But if you believe that an increasing number of jobs will require some familiarity with math or data science—or that living in the world in the 21st century and beyond will require some familiarity with those skills—it seems worth our time. As Eric Mackey, State Superintendent of Education for Alabama, said at the same math panel, “In the 2030s and beyond we can’t rely on only 5% of people being great mathematicians.”

3. Boys are struggling in the early grades.

Emily Oster was one of the keynote speakers at the conference. Her presentation of data showing how much boys are struggling is certainly something worth noting. Oster said that people often suggest we should make the early grades more play-based as a way to make the first years of education more conducive to boys at that age, and she rightly points out that this is just never going to happen. Another idea is to have boys start kindergarten a year later than girls, but of course that forces parents of boys to come up with (and often pay for) another entire year of early childcare. Oster proposes a transitional kindergarten year be available to kids (mostly boys, I presume) who need more time between pre-k and k.

I actually have a different idea and if you’re wondering if it relates to tutoring, why yes, it does. I think it’s time to consider what learning really entails and to create pockets of time when kids, even in kindergarten, are truly engaged in deep learning, and then to ease up on the amount of learning time overall. If you’re getting a 15-minute individualized tutoring session on early literacy, and you are really working hard during those 15-minutes with a trusted adult using a great tutoring curriculum, maybe that gets you farther than 30 minutes of whole group instruction. The trick is the learning has to be real, which is hard to do already and even more difficult when you consider just how hard it is to get teachers to change their practice (humans are creatures of habit—this isn’t a ding as much as a reality check). But I think we should explore it nonetheless. If you take the same approach with math, you can claw back another 15 minutes of the day. Now you have a 30-minute window for a second recess. In other words, maybe the way we create early learning environments that work for all kids is to use what we know about how learning happens best, combine it with some tutoring, and see what happens.

2 Articles

Both of the following articles are very much worth your time.

“When Grades Stop Meaning Anything,” Kelsey Piper.

“How Public Education Failed in the Liberal Enclaves That Care About It Most,” Andrew Rice for New York Magazine

1 Jumble of Thoughts

I’m thinking a lot of about risks, certainty, and data after reading Gerd Gigerenzer’s 2002 book “Calculated Risks: How to Know When Numbers Deceive Us.” We live in a world even more awash with data and statistics now than we did when the book came out, so of course we’ve heeded exactly none of the warnings about ensuring that doctors, lawyers, politicians, and regular citizens of the world have the tools to understand when statistics are accurate.

But Gigerenzer is also honest that many people don’t want to know that decisions are more complicated than statistics might suggest. When asking a group of physicians why they might recommend a specific course of action, one doctor said, “Patients want to be reassured. They want to be relieved of their anxiety, to be in the right hands, even if they do not feel better than before.” I couldn’t help but immediately think about replacing “patients” with “parents.”

A great point that Emily Oster made her in keynote this week was that data are simply inputs. We still have to make a decision. Do you red-shirt your son from kindergarten? Do you pick your neighborhood school or a Spanish-immersion charter school or a religious private school? Do you choose travel soccer or backyard afternoons? Which choice will guarantee your child’s success in life?

What both Gigerenzer and Oster point out is that there is no certainty. You get the best data you can, you make a choice, and then you have to accept the consequences. I’ve spent most of my career working in and around data, and it is too readily used as a shield from the fact that life involves choices and there are no guarantees. Not in the medical profession, and not in predicting your child’s longterm outcomes. But we do need clear data that parents can access to make the best decisions possible and so parents can know well before college whether their child, in fact, can do math. (And we need everyone to have a better grasp of statistics - the lack of understanding in 2002 was frightening and my guess is has only gotten worse.)

Subscribe now

Subscribe now

LC: What does tutoring look like right now for On Your Mark?

MS: We’re starting out this year with around 3,000 students. One of the things that has changed the most is that the growth game and the scaling game is really focused on states. That’s actually a really positive change because a year ago we were wondering what the heck was going to happen. Where would any schools find the resources to be able to support tutoring work? We’re now in a much better place. Most of our growth is in states where there is a concerted effort from a policy perspective and from the funding side to be able to support the work of high-dosage tutoring.

LC: How many states are we talking about?

MS: It feels like we’re getting close to a third of states, somewhere 15-20. Louisiana’s been in the forefront and it’s been really exciting. But then you have states like Massachusetts, which is doing really targeted tutoring work. It’s different from universal tutoring where the focus is on getting kids who are struggling in all subject areas or specifically in math and reading across all grade levels, which is the goal in Louisiana. In Massachusetts, they are investing specifically in early literacy. They’ve got five providers and the entire approach is first grade reading tutoring.

LC: How do you describe the average impact of your program?

MS: We focus mostly on students who start the school year well below the benchmark. These aren’t kids who are on the bubble or need a little help. These are kids who really, really struggle. We focus on what DIBELS calls their zones of growth to be able to measure our impact. When they are falling into those zones, they are making forward trajectory on closing the gap. Last year we were able to get 54% of all of our students across the board in those zones of growth. Our goal this year is 60%.

LC: What do you think makes the difference between a student who gets into those high growth buckets and one who struggles to benefit from tutoring?

MS: That’s a really good question. I think it’s a hard question to answer with a blanket response because everything is very individualized. What we’ve been pushing ourselves to do in those instances is really break it down into the subtests more clinically on DIBELS to be able to see where the student is excelling or where they are potentially getting stuck.

LC: I saw the study, which your organization participated in, that found 81% of tutor time was used instructionally. That number was higher than I thought it might be, to be honest.

MS: I was pleased with that too! I’m super proud of it. What I’m realizing is there’s not one type of tutoring that’s going to meet the needs of every student. Part of what’s interesting to consider in the long run is how do we gather data to better understand the type of tutoring that will be the best fit for each student?

What we’ve uncovered are students with indications of dyslexia who need a more prescriptive program than our virtual tutors currently offer. It might be that some students really need a one-on-one tutor because there’s an attention challenge and they need that one person to focus on and with them. It might be that there is one student who is really great on an adaptive tech product but needs a little human prodding along the way from the engagement and motivation side. I think what we’re learning is just how diverse these needs are. The more data that we can gather on exactly what works for which students, the better we can give the right tools and resources to schools.

LC: I love the granularity of the research at this point and that we’re learning so much about how to do this well. I think it allows for more thinking about how tutoring and classrooms might look different in the future.

MS: One vision I have is where there are a couple of highly trained tutors assigned to a classroom, and they are ready to be deployed based on student needs. It may mean that there’s one or two students who consistently need 20 minutes a day with the tutor. It may be that our tutors are looking at a dashboard along with the classroom teacher and are able to say “Ah, this student is stuck in their independent work. I’m going to pull them in and do a mini lesson.” And I can do that and have the resources in front of me to be able to meet those kids’ needs. That is a differentiated evolution that it seems like it’s far off but we’re close to getting all the data that we need to be able to operate in that way.

LC: What pieces of data do we need for that vision that we don’t have right now?

MS: We’re working on how to get clearer connecting the tutor to the actual academic outcome. Because we’re pretty good at tracking all the data around attendance and implementation. We’ve developed an AI model that gives us feedback on what’s happening in tutoring sessions. But it’s at the input level, not the outcome level. How are we using that data to actually connect it to the outcomes? We need to go there so we can make good value judgments about how schools spend funds. I want to provide a great service with meaningful outcomes for kids where schools have the data they need to align tutoring interventions with individual student needs.

LC: Why did you start a tutoring company?

MS: I have a strong belief that when the personal and professional align, there is a lot of magic that can happen. I love innovating and start-ups, and I wasn’t even sure virtual tutoring could work. And it was this great surprise to realize, “Oh man, does this work!” And there are tons of tutors who want to connect with students in this way. I came out of working in schools where we were begging to find folks who would come in and teach. We’d often be met with candidates and think, “We have to make a choice here, but none of these are the best choice.”

But it turns out there’s a huge talent pool of folks who are super interested in engaging with students as a tutor. It’s a great way to use them as an asset when they’re not directly in schools, and to think about how to inspire them to want to be in classrooms. The biggest opportunity in tutoring is the pool of amazing, talented people all across this country who want to support children in their learning journeys.

Meanwhile, personally, my youngest child was starting to struggle in school. All of a sudden, the material was inaccessible to him. As much as I argued with the school, they were convinced he just needed to pay better attention in class and not get in trouble. Finally he was diagnosed with dyslexia when he was 13 years old. I had the ability and resources to get him into a summer program to receive the explicit foundational skills instruction he needed.

He’s now 16 and he’s thriving. I share this to say, all kids need to have access to the support they need to read on grade level and they need strong foundational instruction. If they’re starting to struggle, all kids should have private tutorials, not just those who can afford it.

LC: I too was skeptical about virtual tutoring at first. It’s been a wonderful happy surprise that it works so well.

MS: It hit me the first time I got to see it in person. A room with 25 students all meeting their tutor at the same time. We work in the primary grades, so this is 5, 6, 7 year-olds. The amount of joy and excitement they brought to meet their tutor, once you see that in person, you think, “Oh. There is something here that is potentially magical.”

LC: What’s the biggest risk to tutoring?

MS: The biggest risks are around funding to make sure that good models are able to rise to the top and then scale accordingly. The field doesn’t feel as crowded anymore, but how are we ensuring that there’s enough actual financial resources in the ecosystem so that we can continue to try things out, to test it out, prove it out, change it, make it cheaper, figure out how we scale it in different ways, figure out what the best match is for each individual student.

I worry that our work isn’t pervasive enough yet that teachers and schools understand that there’s another way to go about really individualizing, differentiating instruction. If you have an interventionist at the school who’s working her tail off and meeting with small groups maybe one or two times a week, she can work with maybe 60 kids. We can meet with 60 kids in the same 20 minutes and then students aren’t missing other parts of their day.

LC: What is the future of tutoring?

MS: The future of tutoring is delivering on promises we’ve been making to kids and families for a very long time. The promise is that school can really adapt to student needs, meet them where they are so that they can have a bright future. It means flexible instruction that I think we can’t imagine yet. We’re going to learn more and more about what the capacities of technology are to be able to support student learning and then how we really change instruction by matching kids to tutors to make sure that they’re going as fast as they can or as slow as they need in order to cover what is most important for them in their learning journey.

I’ve been in education now for 25 years. We’ve been telling teachers, at least since I started as a brand-new teacher in 2000, that it’s your job to meet the needs of every student where they are. I took that to heart and I taught a whole bunch of teachers over time who do take that to heart and that is a really, really hard job to get done. We are now in a position where we can actually start thinking collaboratively around how we deliver on these promises to kids.

Subscribe now

LC: As a homeschooling mom and a homeschooling leader, how have you generally thought about tutoring as a tool? Do you see it as part of a homeschool experience?

AOE: Being a homeschooling mom is like being a charge nurse. Her job is to match the right patient with the right nurse, and she oversees everything. So in my homeschooling journey, I don’t necessarily have to be the one teaching my kids. I can tap someone else in. I just have to make sure that they’re learning.

I have one child who, at first, just hated reading. And I tried everything to get him to read. So I paid this retired reading specialist, 46 years of teaching, and she found a trick that worked for him–she gave him $5 if he could read the words on a page in a certain amount of time. He is now an avid reader. She took him from being below grade level to two years above grade level in 18 months.

LC: If the idea of tutoring is about giving more individualized learning and more small group instruction, is there an argument that a lot of homeschooling overlaps with tutoring? Where is it the same and where is it different?

AOE: I just came to that realization after reading some of your work! I read it and thought, “this is the case for homeschooling right here.” The difference, I would say, is the heart behind it. In my opinion, if you’re in a traditional school, the child is typically behind when they are put into tutoring. But a big realization for me is that I don’t have to wait until there’s a problem to put them in tutoring. I can let them explore other avenues or harder course material now.

The difference is that when the parent is in control of their kid’s education, you can direct it however you see fit. The way I’ve seen it in traditional public schools is “oh, there’s a problem, we need to hurry and catch them up.” I would like to see more homeschoolers embrace tutoring as a way of giving children access to the best or hardest course material as soon as they’re ready for it.

LC: You’ve hired tutors for your own kids. How do you measure success?

AOE: I look at data, that’s my favorite thing. I like to see where they were when they started, here’s where they are in the middle, here’s where they are six months down the line. Are they improving? But also, how’s my child’s attitude about the course material?

One of my sons is great at math but not as strong in language arts. One teacher had really stifled him to the point that he wrote some things that were not positive about himself on the back of a test. What we discovered with tutoring was that he actually has so much personality that he wants to put into writing. The tutor is now teaching him how he can show his personality and be witty, but also stick to MLA format and other writing rules.

LC: One argument I make in my book is that adding tutoring within public schools could help us rethink how we use time in the school day. And one thing we routinely hear is that homeschool kids aren’t spending as much time in class, but that doesn’t mean they’re not learning as much. How do you think about the value of time and learning?

AOE: I think there’s way too much time in the public school day. Kids need a chance for their brain to recoup from all the knowledge. My eldest was in public school and there were times where I said, “You have too much work. Tell the teacher I said no, you’re not doing this.” We chose a homeschool model with a block of time on dedicated days, so it’s 9-2 on Tuesdays and Thursdays. You have these subjects you need to complete and I need to see the work. Wednesday is a study hall where we go to the library. We go to the coffee shop across the street, they get their treats, and they’re in the teen section of the library, working on what they need to work on. Then they have their tutoring time.

Fridays are for labs, tests, and catch-up, and Monday is also for catch-up. If we have a regulated time where you have to do math at this time, it doesn’t prepare kids for the real world. If you have a professional job where you’re not clocking in, you set your own calendar and you are responsible for your work. We’re not setting kids up to be able to have those skills, but that’s what I think my kids are getting. In public high school, kids never have the opportunity to set their own schedule, have autonomy, figure out time management and how to juggle everything.

LC: How does the way you think about tutoring compare to the broader homeschool community you are part of? Is this typical? Are you an outlier?

AOE: Within the homeschool community, there’s a misconception that tutoring is only for wealthy homeschoolers. The library here offers tutoring for free, for example. There are different options. I think a lot of parents would like to do tutoring but think they can’t afford it.

I’ve also used therapy as a tutor. If you have a child with a speech delay or who has a hard time with handwriting and you qualify for therapy, you can turn that into tutoring. One of my sons was in occupational therapy for handwriting, he worked on writing his letters and understanding the rules of writing. It was very similar to tutoring, in my opinion.

It’s also why I love the ESA program that dropped in Texas. Because you can use some of those dollars to pay for tutoring. It doesn’t have to be expensive. I paid $60 a week for two hours of tutoring.

LC: How does that compare with the cost of other things homeschool families spend money on?

AOE: It’s all over the place. You could get away with homeschooling for free. It’s more work, though. When I’m coaching homeschool parents, I ask, “do you want to spend more money for less of your time, or do you want to spend more time for less money?” I’ve done both.

LC: When you hire a tutor, are you giving them materials from the curriculum that you use?

AOE: With our first reading teacher, I just trusted her completely. She asked what curriculum we were using, and I told her. With this new tutor, I’m also focused on building agency in my 12- and 13-year-olds’ learning. I don’t want them to see education as something spoon-fed by Mom. I test my kids and share those results with the tutor.

I want my kids to use their tools, including a tutor, to solve problems and learn to advocate for themselves. If they are struggling with long division and fractions, I’ll say, “Great, you have a tutor. Go tell your tutor.” I do this not to brush off my responsibility as a parent but to have them troubleshoot a little on their own first. It’s important for them to practice sharing their struggles with a trusted adult rather than solely relying on me. I’ve seen how some young adults struggle after high school because they never learned to seek help or problem-solve independently. Going to a tutor helps them build confidence, critical thinking, and resourcefulness while still having a safety net.

LC: Do you feel like there’s a difference between virtual and in-person tutoring? Or is it just about finding the right person?

AOE: Maybe pre-2020 I would have said in-person is better, but now it’s second nature to be virtual. We’re talking right now virtually, and these kids can work all the technology. They are not adverse to technology.

LC: What do you wish the homeschool world understood about tutoring, and what do you wish the tutoring world understood about homeschool parents?

AOE: I wish homeschoolers understood they don’t have to do all the things or recreate the wheel if their child is struggling. It’s okay to tap in for help, and it’s accessible. You might have to do a little digging. There are some programs through the public library, you can log online and get tutoring for free. It doesn’t make you a bad parent.

With the tutoring world, I think there’s a misconception that homeschoolers are extremely brilliant or they’re just wearing a tinfoil hat and all they’re taught is the Bible. There’s a bias there. I would like everyone to be open to the fact that these are just kids.

LC: What is the future of tutoring?

AOE: Accessibility. If every child had access to tutoring, it would alleviate many of the problems we’re having. If kids have access to individualized learning paths, and one teacher can’t do it all but a tutor could, we could close a lot of gaps.

LC: Two paths to access you’ve mentioned are the new Texas ESA and the library. I love the library. I still go to the library with my kids regularly.

AOE: In our town, the downtown library has three levels. The third level has a teen section with flexible seating for them, and special study rooms only teens can book. Every day there’s a maker space with 3D-printing and sewing machines and painting. They also have a teen librarian who can help with so many things–fill out the FAFSA for a first-generation college student or help someone get tutoring. These are all things that can close gaps.

My eldest was in public school in East Fort Worth where schools were just failing. For four years in a row test scores went down. But she did get after-school tutoring at her school and the tutoring is what closed the gap for her, made her believe she was smart. Because she is smart. She just didn’t have the right resources.

In the eighth grade, my best friend and I often found ourselves laying on the ground, staring at the sky, on magical late spring evenings in 1994 when it felt like the sun would never set and summer vacation was around the corner. We watched the sky, pointing at clouds, feeling the lingering warmth of the sun still shining at what seemed an impossibly late hour but was in fact 6:30pm. I remember the feeling of possibility. The magic of being young, of sensing the possibility of your whole life before you and the possibility that the world is changing more rapidly than any world has ever changed and the possibility that you have some role to play that you can’t yet define in this emerging world that no one has ever experienced. Is there any feeling like it?

Three decades later, my oldest son’s very favorite thing to do for several years was watch the Dark Universe movie at the American Museum of Natural History planetarium in New York City. I’ve sat more times than I can count in the darkened, domed theater, listening to Neil deGrasse Tyson explain that one remarkable fact about the universe is wherever you are in it, that is the center of the universe. Infinite centers means, of course, there is no center of the universe. We are particles and light and each our own universe in an unending galactic performance of energy and movement. I often left the theater feeling so very small but with that same sense of possibility. That the world is more than I can see. That my universe will continue to expand in ways I can’t yet imagine.

I’m a bit older now, and I see how the constraints of real life make it hard to wonder. Who has time to lay on the ground and stare at wisps of clouds that look like dragons when there’s dinner to make and children to drive and Zoom calls to attend and the internet to scroll? When older generations routinely denigrate younger generations, it’s not really because they are so much worse. No matter what you think, it’s not that. But they are young, and now we are not. It feels like our universe is no longer expanding. We cannot feel possibility with a mortgage. We cannot believe that the changing world we live in isn’t the most changed, the most challenging, or the most fantastic. When we can’t feel possibility, we begin competing for prestige. In that competition, sarcasm has more value than hope. Criticism scores higher than kindness. Clever takedowns draw online crowds; the house always wins, and the house roots for failure.

I think a lot about tutoring and about what it means to learn. There are still more questions than answers. Our brains are another universe with the potential for unending expansion. When I consider the future of tutoring, the one question I ask everyone I interview for this site, the word I keep landing on right now is “possibility.” That more kids feel that more is possible. That more adults feel that more is possible for their children, or for themselves. That they are the center of their universe, and there is nowhere to go but everywhere and anywhere they want. Yes, the world today is different from those warm spring evenings of 1994 and the world of tomorrow will change yet more. Constant, however, must be our willingness to wonder at the possibilities. Stare at a sky with no limits and feel your place in the universe awaits. I think tutoring can help us get there. I want to live in a world of possibility, I want my children to dream of possibilities I can’t imagine, and I don’t want to be entirely constrained by what already exists.

More kids need more positive adult connections. More adults need more positive human connections. More kids and adults need to see the value in learning to learn, not merely the transactional process that gets you to college or to a job. More kids need to feel excited about learning. More kids need help with foundational math and reading skills. More parents need to feel secure that their child is on solid academic ground.

I can give you the data, show you evidence that tutoring works. We can discuss yet more evidence of why and how it’s hard to do at scale. Someone will say these resources are best spent elsewhere, and someone else will say we have unreasonable hopes for the education system. Yet another will say I’m insufficiently critical of tutoring’s challenges. These things matter, but they are known quantities in the universe, the solar system of implementation and randomized-control trials and policy arguments. DeGrasse Tyson taught me something else in those planetarium showings, though. He talked about dark matter, which he describes in one of his books as “our frenemy. We have no clue what it is. It’s kind of annoying. But we desperately need it in our calculations to arrive at an accurate description of the universe. Scientists are generally uncomfortable whenever we must base our calculations on concepts we don’t understand, but we’ll do it if we have to.”

The dark matter of tutoring is possibility. It’s why thousands of teachers, principals, district administrators, and state leaders have worked so hard to bring a new version of tutoring to hundreds of thousands of American children. Our current public education system fails too many children, limits their possibilities because of their zip code or family income or just bad luck of getting the worst teacher in the grade three years in a row. The post-pandemic tutoring movement reignited the belief of so many adults working in education that there are more possibilities out there. It’s been humbling and wonderful to watch and write about.

So this week, if you ask me, “what is the future of tutoring?”, my answer is: possibility. The universe continues to expand.

LC: We had all these great randomized-control trial results with tutoring and now we’re starting to have real-world results. A lot of them may seem disappointing. Can we actually scale tutoring?

MK: I still think tutoring represents one of the most promising educational reforms and evolutions of the traditional school day that we have. And it would be a real loss if we were to abandon it because of very mixed implementation under difficult circumstances and unrealistic goals at both scale and impact that we’ve had in the wake of the pandemic.

That said, we have a lot of work to do if we’re going to get anywhere near what meaningful tutoring looks like. And the question is, are we willing to do that work and be creative about how we prioritize and allocate funding in our education system to do it?

LC: I agree! And I think it’s encouraging that researchers are really trying to dig into how we can improve efficiency. How do you see the role of research in this work right now?

MK: The last several years have been a successful case study of researchers working in partnership with practitioners to say, “hey, you’re doing this thing and we should learn from it.” This is an unbelievably fertile moment for knowledge in the R&D space of tutoring, where we have all these districts and states doing their own kind of boutique version of what they think tutoring should be.

The amount of research has really blossomed, and I think policy and practice has a lot to gain from that. But the real question is: is what we’re doing going to translate into the Tutoring 2.0 movement? What are the macro lessons learned, where do we need to be more humble? I’m glad we know when tutoring didn’t work or work as well as we hoped it would to move the needle so we can reflect on what it would mean to change it.

I really disagree with a lot of the narrative about what it means [for a program] to work or be successful. We are finding programs that are improving student achievement on state standardized assessments by half-a-tenth or a tenth of a standard deviation. If we were talking about those types of gains in a different context, we would be jumping up and down for joy.

The kind of catch-22 is that researchers and policymakers, myself included, in order to build momentum and a coalition of folks willing to invest in tutoring, I think we highlighted tutoring’s large potential which may have inflated expectations of what success would look like. We have put ourselves in a tough spot.

LC: What’s the research study you would do on tutoring if there were no constraints on available data or other resources?

MK: We need to know so much more about how tutoring affects the tutors. What does it mean to be a tutor? If you are a middle-schooler, a high-schooler, and you take part in a reading buddy program, is it a way to gain confidence, work experience? Does it set you on a path of employment, experience and opening future jobs? Does it expose you to teaching or education more broadly as a potential career path? Does it bring in a more diverse group of future educators? So there are huge questions that really remain unanswered by the tutors themselves.

We should also be examining what generative AI will do to the potential for cost reductions in tutoring. We should do that with a range of healthy skepticism, frankly, because I think there’s something fundamentally important about human-to-human interactions that could be lost. But I don’t think we should exile generative AI from the experimental phase of how to make tutoring more cost effective. I’m cautiously optimistic.

LC: I think I am too. I think tutoring starts to open the door to a conversation that AI pushes even more open, which is: are there other models for how school looks? Are there opportunities to revisit the assumptions we’ve made about the classroom model?

MK: If someone wanted to make a wager about whether or not 5, 10, 15 years from now the grammar of schooling is going to look radically different because of AI, I’ll take that bet. I’ve read enough history about the unbelievable resilience of the traditional educational model to be unwilling to bet on radical deviation. But I do think that in the tutoring space itself, there is real potential for training tutors and combining tutoring with computer-adaptive learning and AI-powered learning platforms, and all the things that I can’t even imagine at this moment.

LC: What’s your take on how much we understand about the mechanisms of why tutoring works? I really want the relationship piece to matter, even though I’m also open to AI playing lots of roles.

MK: I think our understanding is only just emerging. I’ve written papers about the important role of mentoring and the role that teachers play as natural mentors in their students’ lives. So count me on team “human relationships.” That said, I think there’s enough early emerging evidence to suggest that that is not the dominant mechanism through which tutoring operates.

For some kids, that relationship is why they come to school, they feel cared for, seen, understood. And other kids are just not getting the way the teacher is explaining how to do fractions and they need someone else to hold their hand in an individualized way that’s slower. I sat there during COVID and watched my son play Dreambox Math and he loved it, and I think he learned a ton. And there was no adult or other human involved. So I think there’s definitely potential there.

LC: Well, I watched my son hack through the school security filters to get to YouTube during math class. So that was a different COVID experience!

MK: Well, that could be a valuable skill in another context! Look, teachers impact their kids through a lot of different ways, depending on where kids are and what they need. I think so too with tutoring. There’s some tutoring that is about supporting kids who have major gaps in their core conceptual understanding and knowledge base, that we’re just meeting them where they are and helping to shore up those gaps.

And I think there’s other kids who need to be able to ask more questions, engage more proactively in their learning, and tutoring provides the personalization to do that. Then there’s kids for whom it’s the relationship-based motivation, someone with an expectation that you will show up and do your work. All of those are plausible mechanisms but there’s enough evidence that suggests that when we do tutoring that’s aligned with rich, data-driven instruction and strong curricular resources, that’s on average a better thing for kids. It has a lot to do with how tutors spend their time and what content they teach.

LC: What is the future of tutoring?

MK: The future of tutoring that I envision in my hopeful optimism is that we have democratized access to individualized instruction in a way that it’s not only available to families in the private market who can afford it. It is something that is part of the fabric of our education system and it’s integrated into a multi-tiered student support system.

It’s not a one-off program, it’s part of the basic infrastructure of how we support students, it also provides opportunities for adolescents and young adults to be tutors and make connections with their younger peers, pay back the support that they get, be exposed to the education system as a potential career and it thoughtfully integrates technology in a way that is a complement to human-to-human interactions.

That vision is at risk because of findings that didn’t meet the large expectations out of the gate. With high turnover in school district leadership, state legislatures, and governor’s mansions, there’s a question of whether or not we will stay the course, or we’ll decide that tutoring was just the flavor du jour of ed reform. I think we will find amazing innovators who will stay the course. It may take a decade or more of proven demonstration cases and to say, here’s a 10-year path where the end result is so compelling that people say, “ok, let’s do that.”

There’s a parallel analogy of how kindergarten slowly grew to be part of our public education system over multiple decades in a both bottom-up and then state-led approach. People get into the business of education because they think they can be part of positive change so it’s good to have a proactive vision of what that might look like. Changing this ossified, decentralized system of education we have in the United States is hard, hard work.