What Stanford Didn't Teach Me: Rebuilding Vietnam's Hottest EdTech

with 40,000 users and entering 23,000 schools

AIducation is quickly becoming the hottest edtech in Vietnam

but the journey wasn't always clear

what do you do when your startup flops in the field

and only have five months to rebuild

we had to

basically take a very hard look at both our product and ourselves

to see what were our wrong assumptions

and how do you make something that people actually want

when we looked at the data

we realized that most of the stress actually came from education

so it's a all in one

personalized platform that take the students as like the core of it

and then build all of the tools around to support their ecosystem

we believe that the journey of education is not something that the

the children or the students undergo all by themselves

the part that matters is distribution

took me a year to fix that mistake

and don't be ashamed of the product that you're building

just show them even if it's like really bad right now

quick thing before we get started

we have a lofty goal this year of hitting 1,000 subscribers

in order to help more people build really great companies

so if you enjoy the content

learn something new the best way to support us is by subscribing

okay let's get into it

hung is the founder of AIducation

Vietnam's first AI powered learning platform

built by Vietnamese for Vietnamese

hung welcome to Founders in Motion

hey Thea happy to be here and to be sitting on this cozy chair

this is a super fun one cause we've got friends in the studio

and I've known hung for like five

six years yeah

it's a long time a long time

let's kick things off in simple terms

tell me what AIducation is

um so you can call it AIducation or you can call it education

what we are is a platform that supports personalized education

so basically it knows everything about a child uh

their uh

tendency for learning their strengths in learning

which areas they are underperforming in and more specifically they

the platform knows how to get the student to

become better in areas that matter

like in mathematics in English or

you know any other subjects that you're choosing

the platform is a all in one platform

so we have a wide variety of tools

um and the tools are designed for parents as well

for schools as well and also for the teachers

because we believe that the journey

of education is not something that the

the children or the students undergo all by themselves

they go with the parents they go with the teachers

they go with the school so

it's a all in one

personalized platform that take the students as like the core of it

and then build all of the tools around to support that ecosystem

I love that and talk to me about the ecosystem approach

how are you helping them in a comprehensive way

we provide a platform that helps teachers keep track of their students

allow them to design materials

design tests and homeworks

and then be able to automatically host that on a platform

let them let the

the students do those homework

and then automatically grade each of those homework personally

and yield out the strengths

the witnesses and write detailed reports back to the parents

and to the children themselves

for the parents we have a

kind of like a learning portal

that allow them to basically know which area

the children of the child are strong in

which area they need help in and um

they also can know more about like the

the physical and the psychological

health of their children through that

so it's it helps them understand their child on a deeper level

and understand how they can best communicate with their child

in order to help them grow

yeah

and how do you assess the physical and social front

um for the physical and the social

we have uh

the ability to input in your stats

you know your uh

sleeping time your weights

all those attributes that can help calculate

you know your uh uh

BMI and all that yeah

um and for the uh

psychological side we have a

it's called like the AI Journal

basically

it allows them to write in them or like just speak out their thoughts

it'll record all of that and keep track of a kind of like a um

a memoir or like a grimmer of all the interactions that they have

all of those get synthesized

and basically they get analyzed and

come back in the form of a report for the parents

yeah and if that's anything

like

if the system detects that this child may be in danger of anything

like suicidal thoughts

then it will immediately contact the authorities

contact the teachers contact the parents to let them know about that

so that we can take actions before anything bad happens

my first attempt was actually a

like it was called Personal AI and it focused on the mental side

it was basically a service that lets you talk to different like

pals or like friends of yours

similar to character AI at that time

like your

like your friends'friends or like

are these like random strangers

they are not like random strangers

but like they are AI and like

it's specifically for the new generation who prefer to talk to like

their idols like

you know anime characters huh

uh kind of like that

and um

at that time we did have uh users

um but then I realized that it could be like much more than that

because when we looked at the data

we realized that most of the stress actually came from education

so I realized that a full fledged platform that took care of

not just the mental side

but also the education side would be the perfect solution

um we are also um

you know I think growing like quite a lot

and I think it's a testament to the

I guess like the correct vision or like the

the correct um

way of you know

mapping out and building our products

but that is a long story because we did like

fail for quite some time before we finally

were able to find out what we should be building yeah

yeah I

I love the Segway so before you release a product into the market

how do you think about testing if people actually want it and need it

you should build something

and then get it as fast as possible to your customers

and then let them give you their feedback

either it's good either it's not good

either they need something more

either they don't think that this feature is needed

and also

one very important thing is that even if they tell you that it's good

if they don't buy from you

then it's still not good I so agree

everyone will tell you to your face

that they love what you're building

they love everything about it

but it's the moment when someone pulls out their credit card

that is the turning point of like

oh like

does this actually have a fit in the market

cause like trolling it is cheap

but to actually pay for it is a whole other thing

so we had a mini little debrief before this

and you said something that really struck me

that what AIducation is now

is very different from the initial kind of mental health focus

and you actually spent a whole year pushing a totally different angle

only to realize people didn't want it and re

develop your code from scratch

so crazy story

um

why if you have your product out in front of people and then they say

yeah this is not what I want

or if they they say it's cool

but then they don't use it

then I think it's really a problem

we had to basically

take a very hard look at both our product and ourselves

mm hmm to see what were our wrong assumptions

what were the learnings that we had

from basically a year of going out into the field

and actually staying and working and understanding with our customers

yeah and at that point we just decided yeah

you know what let's just throw all this away

we have the ability to rebuild quickly

so why shouldn't we

we now have all the learnings and understandings that a year ago

we did not have now we have actually talked to schools

to parents to teachers and to students

so we actually know what they would want

so why should we not do it

so it it was like um

it was really plain like just painful it

it wasn't like yeah

I had to delete the code or anything

we created like a an entire new repo

but then we had to start from scratch

so we quickly rebuilt it

the four week run that we had basically two sprints

we had a uh

workable platform that we went back

give the users gave the teachers

the students the parents in the schools

and this time it it actually got a different response

okay so I don't wanna dig into your trauma

but if we have to go back to that moment where you were like

oh like this is not working anymore

were there any tickers or signs that like made you feel like yikes

like we need to do something different

like is it

I don't know monthly active users

is it like people not paying

like conversions

like talk to me some of like the like analytical thinking behind that

first warning sign that we had was even though they did sign up

there was no basically no activities on the platform

they they sign up and then they just yeah

I sign up now what

and then they they didn't care hmm

um the second one is that there were like you said

there were no paying users

and the monthly active number of user was too embarrassing to like

discuss right now

um yeah

so I think those would be the two major like

tickers or signs that we really needed to change

we talk about this

but founders get attached to their product very easily

but sometimes across the entrepreneurship journey

it's a patch of struggle and you push through

but how did you figure out that this was a deeper

fundamental issue that needed a pivot

rather than just like a a patch that you needed to like

slap over with like more sales and marketing efforts

like more education like how do you differentiate between like

it's a dry patch versus like shit

like things are wrong everything that is good

so also takes time when we looked back at education at that time

I think we realized that we started from the base of personal AI

and the tools that we had

and we had a great set of tools and solutions

but they were not addressing the specific problems

that our customers had

I really love the story because a lot of people think when you pivot

you start from the beginning

but you really don't like

you start from such a deeper understanding of first

what you're capable of what you're trying to achieve

but the most importantly like what your customers actually want

and then you can kind of make a little bit of a different segment

segue but you have the relationships and you understand them way more

okay so this time

how are you gonna make sure

you're doing something that the market needs

and the customers want uh

what we did were when when we started like rebuilding all of that

as soon as one feature was ready

we ship that out to our testing server and then we had our partners

you know the schools and the teachers and the students and the parents

yeah try that out immediately and then we got the feedback

we also had a kind of like a like a drawing board

and then we had all of the features and then all of the designs like

um which pain point were we solving

and which feature would be best to solve this specific pain point

we communicated that very clearly with our partners and testers

so that they knew why this feature was there

and how that interacted with the

you know the entire ecosystem as a whole

we have gone through 124 like different like um variations

like alternative versions of education in order to make the

like the small changes the investor side of me is coming out here

but how is AIducation actually differentiated from other AI

personalized learning tools

we are not fragmented we are a coherent

cohesive platform with different tools for all players in education

for teachers students

parents and schools

and we provide actual real time personalized education

not just something that you can you know uh

get out of the box with uh chat GPT

and if we zoom out a little bit

talk about broader trans are broader themes

um now with vibe coding people can create apps in seconds

then people are like shipping AI agents

like printing paper these days

how do you think about general strategies to build differentiation

and what are some companies that you think are doing it very well

that is a great question and I think one that yeah

the investor side of you will be really interested in uh

two ways to build differentiation

uh the first way is uh

you actually have to really understand your customer

so in our situation

the reason why we flopped was partly because what we did at the time

yeah could be achieved by Chat GPT

so one of the questions that we were asked quite a lot when we first

got our product to the market was

how is this different from Chat GPT

and why should I be paying you instead of paying $20 to chat GPT

sometimes you may think that you understand their pain point

but then when I actually got the product out

and then I actually talk with them

I actually got to know their pain

got actually got to like

it's like a day in the life of my customer kind of thing

only by working with them

can you actually build something that they want

and something that is differentiated from the market

the second thing that you can do is

if you have something that is similar in um

you know in terms of qualities and

um in terms of features to your competitors

yeah then the part that matters is distribution companies

even tech companies they are resistant to changes

in the sense that if they have been using one product

and that product is really good

like 9 out of 10 then it's really hard to convince them to change

like a 9 point 25 agree out 10 product

the other one is integration um

Fireflies AI were founded by a two uh students from MIT

what they do is they are like a smart note taker

the reason why Fireflies AI

were able to grow so fast

was because they had a lot of deep integrations

they knew that their cut

their main customers were going to be the big Corp

you know they were going to sell b to B

and the thing with b to B companies

is that they already have their own CRM

they already have their own communication tool

and they realize that if they can

integrate themselves deeply into this ecosystem yeah

by allowing the user to just click one button and have their meeting

uh you know

data be basically smoothly transition and float into their ecosystem

and that's how they actually survive even though um

video platforms like Google

Meet and zoom can turn out all of that too

but they don't have the integrations

to go with all of the different tools that um

you know enterprises are now using

the cost of development has gotten so cheap

and it's become so easy to deploy applications

and to write code

what becomes more fundamentally important is how do you distribute

how do you create a brand

how do you create image

and that's where you kind of get those rise of like

influencer brands like Chloe Kardashian with her popcorn or rode

um to selling for $1 billion

yeah I know right incredible

incredible um

read to Hailey though

amazing to Hailey um

so yeah like

I think distribution is such an important thing

and like now since we're on the topic of rappers and GPT rappers

what are your thoughts on them

people are giving GBT rappers kind of a bad vibe

yeah like I

at our at education

we do train our like own models yeah

um but just because we have our own models like um

I don't think that like

you don't have to train your own models to be like different

and you can be a GBT rapper without like having all of the like

the bad like marketing names to it

because if you really think about it in the past

like companies like Oracle and Google

like they have softwares that are customer facing right yeah

but at the heart of those softwares

they are just databases

instead of thinking about it as like GPT wrappers

you can think about it as like

a function in the code that allows your platform to be intelligent

everything is a rapper of everything yeah

everything raps like a W s yeah yeah

right

GPT and Gemini um

they are general purpose AI

yeah so sometimes for specific tasks

you would want to train

so that your model gets a deeper understanding of that specific task

training for us was important

because we really wanted

a model that was capable of conversing in a way that is

you know according to the standards of pedagogy in Vietnam

yeah but also

the model has to really understand and really be able to solve the

math and the English

and the physics and the chemistry problems correctly

I don't think being able to train your own model

necessarily matters that much

in some cases in some cases

it provides you that extra like

um accuracy

yeah um

but for most founders and most companies

I don't think that would be needed

um an example would be cow AI

like they are really successful with uh

you know documenting your food and calories right yeah

um I don't uh

I I think one of the founders mentioned that they had like

some tricks with that I don't think they um like

um train it uh

too much but they had kind of like a look up to know

which food correspond to which like

number of calories so I think

you know smart tricks and smart engineering

yeah is what is needed

not necessarily the ability to like train your own model

you know that's like so important

cause everyone thinks that you can just call GPT and call it a day

but like the exercise of fine tuning it and like

training it on your own database

like fine tuning it and then tweaking it

is actually such an arduous process

yeah because sometimes it hallucinates

without you understanding why it hallucinates

and it becomes like you're kind of working

with something that you're not

quite understanding the architecture behind it

and you're trying to make it as accurate as possible

and so some companies especially like

um in very specialized spaces like healthcare or yeah

things like that you have to be very accurate

cause you're dealing with like patient records

yeah um

and founders spent like years in development

just trying to figure out like

how to make it as accurate

um and people just think that they

they can just call a GPT rapper and call it a day

I think that's where the rappers get a bad rap

yeah yeah

I think I think so too so um

one topic that we don't really touch on too much in the show but um

so hiring

hiring is one of those things that sounds pretty straightforward

but choosing those first few people or first believers

are actually quite difficult

and something that you really

really don't think about when you're just starting out

so yeah like how did you think about your first hire

and what were some of the things that you looked for

I have had experience working with

with some like startup companies in the US before

and I have seen their culture

so um I know it's a cliche but like

I think startup culture is actually really important

you'll be working with them at least like 8 hours a day

and then maybe some way more yeah

way more

so you should at least be able to have fun and you know

have a good laugh with them

now the process at hiring of hiring an education was uh

we go through like screenings and if it got past screenings

then I'll interview them but when I interview them

I will focus mostly not on their like uh

tech capabilities or like their abilities to do the work anymore

I'll actually focus on whether or not this person can vibe with like

the team yeah

yes so that is very important

and now we have a team that is a little bit more than 20 people now

oh yeah and the they are all like great friends with each other

they like today they are actually going out with each other to um

do the support for the

for the kids who are doing the entrance exam to universities

normally in Vietnam you work from Monday to Saturday but for us we

which is a concept that I think is crazy

that you have to work Saturdays yeah

but people do do it here people do it a lot here

but for our company on Saturday

you don't have to work the afternoon shift and on the morning shift

you also don't have to work

but you need to come to do TikTok with each other like

you know basically have fun basically yeah

so I mean make content too

you know yeah

make content too great

and we like to do a mini little segment on our show to keep it fun um

it's would you rather

mm hmm so I'm gonna give you two scenarios and you can tell me like

which is the least dreadful

mm hmm from the two okay

you ready yeah

would you rather pitch on live TV or cold call 100 investors

that's interesting

hmm

that is depressing haha

um personally for me

I think I would rather it's about like

hey if I pitch

then would they actually care about it

or they just turn off the TV and do that

um I guess I'll do the code calling with the investors really

yeah I would have thought that you would set TV cause you're already

you've been on TV before yeah

I have been on TV before but um

it was to showcase how successful we are oh OK

I'm a uh

I'm an introvert at heart

it's a really tough to like go on national TV and then talk about like

if you flop or something like that

it's tough okay

would you hire a genius jerk or a nice under performer

the jerk is person who like disrupts everyone

then I would have to choose the um

under performer but nice

love that um

run out of cash or run out of energy

run out of energy I can try to get my energy back

but I can it is tough to get cash

it's tough to get capital

hmm true

yeah

and what's the most valuable lesson that you've Learned as a founder

each and every one of us has a different like version of reality

I was so immersed in my own view of reality

and what would be a good product

yeah I did not really have a reality check on

whether or not others would also want my product

it took me a year to fix that mistake

and don't be ashamed of the product that you're building

just show them even if it's like really bad right now

but at least you all know

if that's something that they would want and use

thank you for coming on the show

really love hearing about all the struggles

all the pivots and what you're up to now

and I can't wait to see where AIducation takes

like maybe in the form of a different language eventually

who knows we'll see

that's a wrap if you like this episode

please hit the subscribe button

it helps us bring on more awesome guest

level up production and drop new series you wanna watch

see you in 2 weeks