At a time when schools are investing heavily in data, many educators still lack timely, actionable insight into how students are actually learning. This is particularly true in math, where misunderstandings can compound quickly. Teryn Thomas is the CEO and Founder of EdLight, an organization focused on analyzing student work to identify patterns in reasoning, misconceptions, and strategy use in real time. By turning everyday classroom artifacts into meaningful instructional insights, EdLight aims to help teachers respond more precisely and improve outcomes. In this 5 Questions interview, Thomas reflects on the limits of traditional assessment systems, what teachers really want from AI, and how better visibility into student thinking can reshape instruction and equity in math classrooms.
What is the nature of your work?
What’s often overlooked is this: schools aren’t short on data, they’re short on usable insight into how students think. We’re rebuilding the math instruction infrastructure to change that.
Most school systems rely on test scores and averages to tell us how students are doing. We focus on something different: how students are actually thinking. By analyzing the math students write, draw, and explain, we surface patterns in reasoning, misconceptions, and strategy use in real time.
Here’s the thing: teachers already collect extraordinary evidence of learning every day. The problem isn’t access to data. It’s the inability to analyze it at scale. EdLight turns student work into actionable insights so teachers and leaders can respond quickly and precisely.
And it works. In a recent study, EdLight participants scored 11.8 points higher on the IAR math exam than peers with similar prior achievement, with the strongest effects for students starting furthest behind.
We’ve been partners with Valley View School District in the suburbs of Chicago for four years, and they’ve transformed their student work protocols, professional learning communities, and ability to do error analysis at scale. When teachers can clearly see how students are thinking, instruction becomes more responsive, more equitable, and more effective.
Why is this work important?
Here’s what keeps me up at night: The way schools use data right now is fundamentally misaligned with how learning actually happens.
Today, schools invest heavily in assessments that deliver delayed, aggregated feedback. By the time results arrive, the instructional moment has passed. Teachers are left making high-stakes decisions without timely insight into how students are actually reasoning. That lag has real consequences. In math especially, misunderstandings compound fast. When students repeatedly experience confusion without targeted support, they disengage, and identity gaps form early. I’ve seen it as a teacher, as a school founder, and now as a CEO building tools to fix what isn’t working for students and teachers.
This isn’t just about efficiency. It’s about equity and return on investment. Our research shows EdLight’s greatest impact is with students starting furthest behind, with meaningful gains across groups, including up to 6.5 points for students at the lowest starting levels. When you improve high-quality, evidence-based classroom instruction for all students, you decrease the number of students who need costly interventions. That saves districts real money while reaching more kids.
Superintendents are making tough decisions about how to spend right now. We’re not adding another program. We’re making the instruction that’s already happening work better.
What’s been the biggest surprise so far?
How much teachers want tools that respect their professional judgment instead of trying to automate it away.
There’s understandable skepticism around AI in education. But what we have found is that teachers aren’t resistant to AI. When technology is positioned as augmentation rather than automation, adoption changes dramatically.
That insight fundamentally shaped our product strategy. We built EdLight to surface patterns and insights while keeping instructional decision-making firmly in the hands of educators. It changed how we design, how we sell, and how we partner with schools.
Another major surprise has been the predictive power of aggregated student work. One assignment feels small. When, instead, you analyze thousands of pieces of work across classrooms and schools, patterns emerge that can anticipate where misconceptions will form before they fully surface. That has implications not just for classroom practice, but for curriculum design and system-level planning.
Where do you see your work in five years?
In five years, I see EdLight as the standard for formative insight in math classrooms nationwide.
Teachers and leaders will walk into planning meetings with immediate visibility into how students are reasoning. Time won’t be spent debating what happened, but deciding what to do next. With evidence generated via our platform, the conversation shifts from “did they teach it?” to “did the kids learn it?”
Beyond classroom impact, I expect our aggregated data to contribute meaningfully to research on how students learn mathematics. We’re already partnered with Brown University and multiple school districts, and in five years, that research footprint should be shaping curriculum development, teacher preparation, and district strategy at scale.
From a business perspective, success means becoming the trusted partner for districts serious about improving math outcomes while spending smarter. Not adding programs. Making existing instruction more effective.
But here’s what I actually measure success by: student identity. If more students see themselves as capable mathematicians because instruction consistently meets them where they are, we’ve done the job.
It’s significant that we are doing this work in community. Through the AIMS Collaboratory, we’re partnering with districts, researchers, and organizations who are asking similar questions about how AI can actually support teaching and learning, not just measure it. What I appreciate about that space is the shared commitment to doing this work responsibly and in service of real classroom problems. It’s not about building in isolation. It’s about learning together and making sure what we create really works for teachers and students.
What else should people know?
The future of AI in education will belong to companies that earn trust.
In my experience, change moves at the pace of trust, relationships, and school calendars. But the upside is durability.
Technology doesn’t transform schools by default. It works when it aligns with teacher workflows, protects student privacy, and strengthens instructional integrity. For us, responsible AI isn’t a feature. It’s a design principle.
Every product decision starts with classroom realities: What decisions are hardest? Where is visibility weakest? What costs teachers the most time? Listening isn’t a phase for us. It’s how we operate.
This work is also deeply personal. After years in classrooms and instructional leadership, founding schools, leading networks, sitting in the data meetings that didn’t work, I’ve seen how strong math instruction opens doors for students. And I’ve seen how fast those doors close when kids fall behind.
We’re building EdLight for durability. Long-term district partnerships like our four-year relationship with Valley View, rigorous research alignment with Brown University, and infrastructure schools can count on. Not experimental tools that disappear when a grant runs out. Our communities don’t have time for that.
Education doesn’t need more noise. It needs better insight and better results. If we can help teachers reach more students earlier and more precisely, that impact compounds for years to come.
