The best products grow because they make things easier. They remove friction. Tech giants like Apple and Amazon understood this early on.
Apple removed friction between people and computers with intuitive interfaces
Amazon removed the friction between wanting something and having it, with one-click ordering and next-day delivery.
AI startups like Cursor ($2B+ ARR), Lovable ($200MM ARR in 12 months) and Harvey AI ($190M ARR) are growing faster than anything we’ve seen before. These newer product led growth strategies are centered around the same core idea, but now AI removes friction under the hood.
Cursor and Lovable make coding accessible - anyone can code an app in 2 hours without needing to learn engineering for 4 years in undergrad.
Harvey lets lawyers move fast. Removing friction in their daily workflow by drafting paperwork and accelerating legal research so precision doesn't get sacrificed for speed.
While tech entrepreneurs are long at work trying to make things easier, teachers are left with a greater challenge - how to incentivize children to go through the necessary friction of learning new things… instead of relying on chatgpt for the answer?
Children are keen to look for short cuts. If writing an essay line by line takes 5 hours, but asking Claude to write one takes 5 minutes, then they are often quick to gravitate to the latter. AI can quickly be abused and used as a tool for cheating.
Childproofing AI - outside and inside of the classroom - is paramount. While I could lament on the importance of AI policy and regulation, here I’d like to focus uniquely on how AI is used within classrooms. Edtech products sit at this exciting and innovative intersection - hoping to accelerate learning and provide more access to understaffed schools across America!
To spotlight one of many, Goblins, is an AI tutor for K-12 math. It lets students draw, speak or type on any device to guide them to learn math. The application has various models but my favorite is their cartoon avatar - it lets teachers clone themselves into cartoon form, mimicking their voice and teaching style, so they can individually teach every one of their students. (check out more on them here → https://lnkd.in/p/ePrudjwt)
Edtech products like this save teachers time, and help students who are falling behind catch up. Instead of using AI to cut corner, programs like this engineer AI models that can do things like spot where kids are skipping required readings to divert their attention back or link children back to lesson modules they need to review based on what they are answering incorrectly. The model doesn’t just accept a correct answer without first vetting the student shows their work.
But how do we actually distribute these products into more schools for more children?
While growth hackers in tech are cracked out on PLG 3.0 agentic strategies, education lags far behind. Traditional product led growth strategies pose a unique challenge for edtech products.
When growing a product it’s important to consider the user and the buyer. The easiest products to sell are when the user and buyer converge into one avatar - think selling a skirt to a twenty something working gal. she’s got her own money and her own buying power to purchase the skirt. Edtech products can’t sell or really advertise to children because of regulation protecting children. Math teachers also use the product, but math teachers aren’t the buyers… the product has to be approved by the school administrators, aka the principals who are usually not using the product daily.
I sat down with my old high school vice principal (shoutout new york city public school system) to talk through this unique problem.
Eleanor Roosevelt Highschool
Yeou Jey Vasconcelos is an nyc school all star. She was the vice principal of Eleanor Rosevelt High School during my time, principal of Fiorello H LaGuardia High School, and now principal of Washington Heights K-8 Charter School. When I first started chatting with her about AI in schools, she smiled concerningly… saying that even using the word AI in school is like some sort of taboo. Regarding scaling edtech products she spoke with me a bit about the constraints teachers face.
Many schools receive their annual budget in February. So by August when we spoke, they don’t have room to spend thousands of dollars. As Mrs. Vasconcelos put it, “If they could give us 5 licenses for $200 we’d take it.”
Potentially exploring a top down, tiered growth motion would be helpful?
top down: principal implements into a few classroom → expands to all K-12 math classes in the school
tiered: free demo → low cost 2-5 license pilot with a commitment to revisit at budget season → full sell and classroom rollout timed to the budget cycle
Maybe one detail worth building in could be, requiring that 1 license of the bunch be gifted to a classroom at another school. That turns each sale into a built in school to school referral engine.
Or maybe its a bottom up growth motion?
bottom up: individual teacher adopts and champions the product → builds grassroots demand among peers → pushes upward until administrators formalize the purchase
Entering teacher communities as a an early wedge (maybe even Teach For America corps since those members are young, sold on AI’s impact, and rotate through under resourced schools every year so brand evangelists created here don’t stay put at one school, but they carry the product with them into new schools and districts)
Bottom up growth motions are slower to convert (waiting on teachers to build enough internal credibility to move a budget decision)… but cheaper to acquire and stickier once it lands.
How do school budget constraints align with the unit economics of running a profitable edtech startup? And how do we run an AI product smoothly for hundreds of thousands of students when an entire school district logs on at 9am without crashing the school’s network?
I am hiring product engineers for Goblins to help solve this exact problem. Reach out to [email protected] to learn more :)
see you at the next side quest (!)