Rohit Agarwal, founder of ed-tech company TenMarks, remembers the moment he came up with the idea for the startup. The daughter of his friend brought home a C+ on a math quiz when Agarwal was visiting. He had taught math to pay his way through grad school, so he offered to help. “She said, ‘Don’t tutor me on what I already flunked, help me for the next test,’” Agarwal says. “And I was like, ‘Honey algebra doesn’t work like that, it builds on top of each other.’” She put her head down on the table and started crying, and the moment sparked Agarwal’s idea for TenMarks.

K-12 math teachers can give out homework assignments on computers through TenMarks, which grades students’ questions as they finish them. When students struggle with a problem, the system can prompt follow-up questions to figure out what part of the concept the student doesn’t understand. Once the machine learns that, it will offer additional readings or videos for the student to watch that explain the concept.

The idea behind the program is that TenMarks will help the students who miss fundamental building blocks, perhaps because they didn’t understand it the first time around. And it will help them understand those building blocks in the process of practicing, instead of them receiving the feedback a grade back after the homework is already turned in — a grade that is likely stuffed into their backpack and not looked at again.

“Instead of grading students and telling them what they failed on after the fact, expecting them to fix it, why don’t we help them the first time around?” Agarwal says.

Last week, TenMarks rolled out “common core” assignments and questions that meet the national standards many states are adopting. On the surface it’s an exciting venture. It could cut down on the hours and hours teachers spend grading papers, while simultaneously giving students a personalized learning experience.

But like many ed-tech ventures, the system has an inherent bias to benefiting privileged children and wealthier school districts. TenMarks runs on a computer, meaning every single student in the class needs to have access to a computer to do their assignments with it. In middle to higher income areas, this probably wouldn’t be a problem, since most homes would have at least one, if not two or three computers. But for students and teachers in a low income area, TenMarks isn’t an option.

Agarwal says that schools have computer labs, and students could complete assignments there. But in order for TenMarks to roll out to every grade at a school where students don’t have guaranteed home access to a computer, that would require computer labs to service the whole student body population. It’s unlikely that a cash-strapped school would be spending money outfitting itself with a bunch of computers.

This is the problem inherent in many ed-tech solutions besides TenMarks: They presuppose the money for technology exists. And for the schools without the money — well, the gap in student achievement just gets wider when privileged students have individual lessons tailored to them by a computer, and less privileged students do not.

That’s not really TenMarks problem though, and for what it offers it tries to keep the playing field level. The program is monetized on a freemium basis, where teachers can administer assignments to students through the system for free. It’s TenMarks’ way of grassroots sales: Get the teachers to love the software first, and then offer the personalization of lessons to districts for an added cost of $20 per student per year. The assignments are created and vetted by former teachers, so there’s a human brain behind the machine algorithm.

It’s a sensible endeavor, one that most parents would logically want for their kids. It’s like having the tutor go home to help with math homework. Or as Agarwal puts it, “We want to be the teacher’s best friend.”